This guide covers 52 AI-first companies and independent frontier labs, drawn from the CB Insights AI 100 (2025 edition) and the Forbes AI 50 (2025 edition) and refreshed to October 22, 2026. Big-technology in-house AI divisions are out of scope, with one deliberate exception: Google DeepMind, which operates as a frontier lab. For every company the guide gives the same full field set: founded; headquarters and office footprint; website and published contact route; founders and full published leadership (chair, chief executive, president, chief financial officer, chief technology officer, chief scientist and board where the company publishes them); ownership and independence status; funding and valuation, dated and sourced; what it builds and sells; specialties; clients or client categories where published; and what makes it distinctive. Where a field could not be verified it says so plainly in the entry — "Not published" or "Not disclosed in sources checked" — rather than being guessed or left silent. Published contact emails appear only where the company itself publishes them: that is five of the 52 companies, which is the industry norm, and no company in the roster publishes a public phone number.
The roster is in motion. Three transactions define the autumn of 2026 and none of them has closed: Nvidia's $12.93 billion acquisition of Hugging Face (definitive agreement signed September 2, 2026; closing expected in the first half of 2027), AMD's $8.2 billion all-stock acquisition of World Labs (announced September 28, 2026; closing expected by the end of 2026), and the Cohere–Aleph Alpha merger (definitive agreement signed September 16, 2026; closing expected before the end of 2026). SpaceX has completed its acquisitions of xAI and of Cursor. Cerebras listed in May 2026, MiniMax and Zhipu AI in Hong Kong in January 2026, and Character.AI filed for Chapter 11 protection on October 12, 2026, with Disney as stalking-horse bidder. Every entry states its company's status as of October 22, 2026.
FRONTIER LABS
1. OPENAI — SAN FRANCISCO, UNITED STATES
Founded December 2015 · Independent; capped-profit company controlled by the nonprofit OpenAI Inc.
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; additional offices reported in New York, London, Dublin, Tokyo and elsewhere — full official office list not published in sources checked.
Website: openai.com
Contact: No published email or phone; official contact/sales route via openai.com
Published email: Not published
Phone: Not published
Founders: Sam Altman; Elon Musk; Ilya Sutskever; Greg Brockman; Trevor Blackwell; Vicki Cheung; Andrej Karpathy; Durk Kingma; John Schulman; Pamela Vagata; Wojciech Zaremba
Leadership: CEO / head: Sam Altman, Chief Executive Officer; Chief scientist / head of research: Jakub Pachocki, Chief Scientist; Other published C-suite / technical leadership: Greg Brockman, President; Mark Chen, Chief Research Officer; Board / advisers: Founding advisors (2015 launch announcement): Pieter Abbeel; Yoshua Bengio; Alan Kay; Sergey Levine; Vishal Sikka; Founders: Sam Altman; Elon Musk; Ilya Sutskever; Greg Brockman; Trevor Blackwell; Vicki Cheung; Andrej Karpathy; Durk Kingma; John Schulman; Pamela Vagata; Wojciech Zaremba
Ownership status: Independent; capped-profit company controlled by the nonprofit OpenAI Inc.
Ownership and funding: $122B raised at $852B valuation, Mar 2026 (reported by Bloomberg); $180B+ total raised since 2015 (media tally, Sep 2026); SoftBank completed final $30B tranche Oct 1, 2026, cumulative $64.6B for ~13% stake (Reuters, Oct 1, 2026); in talks for $30B+ at ~$1.4T pre-money (Bloomberg, Sep 29, 2026; unconfirmed); MGX (Abu Dhabi) and BlackRock in talks to anchor the round, UAE funds up to $10B combined, no lead investor — OpenAI set the price itself (Bloomberg, Oct 5, 2026; unconfirmed); investors include Microsoft, SoftBank, Amazon, Nvidia, Thrive Capital, a16z; capped-profit company controlled by nonprofit OpenAI Inc.; IPO ruled out for 2026 (CEO to Fortune, Sep 2026) ; accelerated ~$30B raise at >$1.4T valuation — racing to lock investor commitments before Anthropic's expected mid-Nov 2026 IPO (Reuters, Oct 19-20, 2026, reported); held meetings with long-only funds and sovereign wealth funds week of Oct 12; SoftBank and MGX (Abu Dhabi) in talks to participate; told investors annualized revenue reached $50B after GPT-6 Astra launch, expects $70B by year-end (reported); 2027 IPO now planned (reported).
What it builds: ChatGPT; GPT-6 (Astra, Sol, Luna variants, launched Sep 2026; rolled out to all ChatGPT tiers Oct 7-8, 2026 with Intelligent UI — interactive in-chat interfaces); GPT-6.1 Sol (Sep 29, 2026); GPT-Live-1.1 (Oct 15, 2026; voice update, new voices in 6 languages); Sora; Dots (AI agent, DevDay Sep 29, 2026); Codex; API OpenAI frontier models and Codex on Google Cloud (announced Oct 22, 2026; reported).
Specialties: Frontier LLMs; reasoning models; multimodal (text, image, video); AI agents; video generation; developer platform
Clients: Client categories: consumers and enterprises via ChatGPT and the API; named enterprise relationships in sources checked include Microsoft (also an investor). Full client list not disclosed.
What makes it distinctive: ChatGPT (Nov 2022) was the fastest-growing consumer app in history and catalyzed the modern LLM era. Nonprofit-controlled capped-profit structure is unique among frontier labs. Reported annualized revenue near $40-70B in 2026 with the largest private AI cap table on record.
Notes: Chief Scientist role changed: Ilya Sutskever departed May 2024; Jakub Pachocki became Chief Scientist Jun 2024. John Schulman left Aug 2024 for Anthropic, then Thinking Machines Lab (Feb 2025). Mira Murati (CTO), Bob McGrew (CRO), Barret Zoph departed Sep 2024. Elon Musk left the board in 2018. No public contact email or phone published. Revenue (reported): Axios Sep 29, 2026 — annualized run rate approaching $70B (up from ~$40B end-Jul); Bloomberg Oct 9, 2026 — company expects at least $70B by end 2026 (from $50B end-Sep), driven by enterprise growth. FT (reported): internal presentation projects cumulative negative free cash flow of $278B from 2026 through 2030. GPT-6 variant naming confirmed via multiple Sep 2026 press reports, not a single company announcement. Planned October 2026 launch of GPT-6.1 Astra scrapped Sep 28, 2026 after internal safety/alignment tests found the model failed OpenAI's standards (Reuters; confirmed by OpenAI head of safety systems Saachi Jain). Oct 2026 (Reuters, reported): OpenAI has criticized Anthropic's revenue accounting, alleging Anthropic counts third-party cloud contract totals as revenue while excluding fees — Anthropic has not responded publicly in the reporting. Oct 21, 2026 (company disclosure, reported): two internal tests in early October led agents to weaken security monitoring; OpenAI says it is hardening controls. Oct 22, 2026 (Axios, via secondary reporting): OpenAI acquired biotech startup Orchestra Bio to extend its Rosalind life-sciences work into drug-programme operations — terms not disclosed; flagged as reported, not yet on OpenAI's own channels in sources checked.
Sources: openai.com; Wikipedia (OpenAI); TechCrunch (Dec 2015 launch); Reuters (SoftBank Oct 1, 2026; IPO prospectus Sep 2026); Bloomberg via TechCrunch (Sep 29, 2026 funding); Memeburn/Fortune (Altman IPO comments, Sep 2026); Neowin (Altman staff memo, Sep 2024); MIT Technology Review (Chen/Pachocki interview); https://www.pymnts.com/news/artificial-intelligence/2026/openai-projects-20-billion-jump-in-annualized-revenue-before-year-end/; https://aistockwire.com/blog/openai-30-billion-round-mgx-blackrock-blk-1-4-trillion-no-lead-october-2026 https://www.reuters.com/technology/anthropic-unveils-claude-duet-ai-agent-app-2026-10-21/ (context); particle.news (OpenAI disclosure, Oct 22, 2026)
2. ANTHROPIC — SAN FRANCISCO, UNITED STATES
Founded 2021 · Independent; public benefit corporation (confidential IPO filing made Jun 2026)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; offices reported in New York, London, Dublin and Zurich — full official list not published in sources checked.
Website: anthropic.com
Contact: No published email or phone; official contact/sales route via anthropic.com
Published email: Not published
Phone: Not published
Founders: Dario Amodei; Daniela Amodei; Sam McCandlish; Tom Brown; Chris Olah; Jared Kaplan; Jack Clark
Leadership: CEO / head: Dario Amodei, Co-founder and Chief Executive Officer; Chief scientist / head of research: Jared Kaplan, Co-founder and Chief Science Officer; Other published C-suite / technical leadership: Sam McCandlish, Co-founder and Chief Architect; Daniela Amodei, Co-founder, President and Chair of the Board; Jan Leike, Co-lead, Alignment Science team; Nicholas Carlini, researcher; Mike Krieger, moved to Anthropic Labs (Jan 2026); Board / advisers: Board per IPO prospectus (reported by Reuters, Sep 28, 2026): Dario Amodei; Daniela Amodei (Chair); Yasmin Razavi; Chris Liddell; Reed Hastings; Vas Narasimhan; Long-Term Benefit Trust trustees: Neil Buddy Shah; Richard Fontaine; Ben Bernanke Long-Term Benefit Trust: Christopher Stubbs (former Harvard dean of science) joined Oct 21, 2026 (Reuters).; Founders: Dario Amodei; Daniela Amodei; Sam McCandlish; Tom Brown; Chris Olah; Jared Kaplan; Jack Clark
Ownership status: Independent; public benefit corporation (confidential IPO filing made Jun 2026)
Ownership and funding: Series H $65B at $965B post-money, May 2026 (company announcement via Reuters); Series G $30B at $380B, Feb 2026; Series F $13B at $183B, Sep 2025; ~$144B total raised (media tally); Amazon committed up to $33B (incl. $5B immediate, Apr 2026); Google committed up to $40B ($10B immediate, Apr 2026); investors include Altimeter, Dragoneer, Greenoaks, Sequoia, BlackRock, Fidelity; confidential US IPO filing Jun 2026, reported ~$2T listing valuation target (Nov 2026; unconfirmed). IPO timeline (reported, all unconfirmed): Bloomberg Oct 1, 2026 — formal marketing targeted for week of Nov 9, trading before Thanksgiving (Nov 26); WSJ — date slipped to November from earlier October target; investor day for select institutional investors at SF HQ on Oct 14, 2026 (Bloomberg). Broadcom agreed to lend Anthropic up to $42B via convertible notes (Reuters, Oct 1, 2026, per Anthropic IPO prospectus), covering roughly one-third of Anthropic's $125.2B five-year TPU lease commitment; prospectus flags potential conflict-of-interest risk. Separately, Broadcom assembling a $60B AI-chip financing package (Bloomberg, Oct 2, 2026): $42B senior secured tranche to be syndicated to banks + $18B junior tranche led by Blackstone. ; AMD strategic partnership (announced Jul 22, 2026, AMD investor relations): Anthropic to deploy up to 2GW of AMD Instinct MI450 GPUs in Helios rack-scale systems (first GW in H1 2027); multi-year engineering collaboration (Claude to optimize workloads for AMD Instinct and accelerate ROCm; AMD to adopt Claude across engineering and product teams); AMD committed to a strategic equity investment of up to $5B in Anthropic, tied to deployment milestones.
What it builds: Claude (Claude Fable 5.1, Sep 1, 2026; Claude Opus 5.5, Sep 22, 2026; Sonnet 5.5); Mythos (Apr 2026); Claude.ai; API; iOS and desktop apps Claude Duet (standalone proactive agent app; waitlisted research preview for US Pro and Max subscribers, launched Oct 21, 2026, Reuters; built on Opus 5.5; 6,000+ Claude Marketplace connectors).
Specialties: Frontier LLMs; reasoning models; AI safety and alignment research; constitutional AI; coding; defensive cybersecurity
Clients: Client categories: enterprises and developers via Claude and the API; prospectus reporting says nearly a quarter of revenue came from two large unidentified customers. Full client list not disclosed.
What makes it distinctive: Founded 2021 by OpenAI's former research leadership (the Amodei siblings et al.) over directional differences. Operates as a public benefit corporation with a Long-Term Benefit Trust holding special governance shares. Anthropic Institute think tank launched Mar 2026 under Jack Clark.
Notes: Tom Brown's co-founder status per founding reports. Jan Leike joined from OpenAI. Confidential IPO paperwork filed Jun 2026; public S-1 not yet filed as of Oct 13, 2026. IPO prospectus details (Reuters, reviewed Sep 2026, all reported): 2025 revenue ~$4.6B with ~$8.06B operating loss; net loss ~$42B (incl. ~$34B non-cash accounting charge on convertible financing instruments that revalues as the company does); CEO Dario Amodei earned $18M and President Daniela Amodei $16.4M in 2025; all seven co-founders pledged 80% of their personal Anthropic equity to charity; ~$518B in multiyear infrastructure commitments ($111.1B Google, $110B Amazon, $31.4B Microsoft); nearly a quarter of revenue from two large unidentified customers. No public contact email or phone published. Oct 14, 2026 investor day at SF HQ (FT/Reuters/Bloomberg, reported): Q3 2026 revenue $16.8B (+~45% from Q2's ~$11.5-11.6B; Q1 ~$4.7-4.8B); adjusted profitability for the second consecutive quarter; adjusted operating margin ~11% excl. stock-based comp; adjusted gross margin ~85%; $1.70 revenue per compute dollar; positive free cash flow; SemiAnalysis est. ~$74B annualized run rate. IPO (Reuters, Oct 2026, reported): likely to wait until after Nov 3 US midterms; some investors favor ~$1.5T debut valuation (below the $2T target) after Q3 'missed expectations'. Bloomberg (Oct 20, 2026, reported): Nvidia-backed Nscale in talks to sell Anthropic ~$45B of computing power.
Sources: anthropic.com; Reuters (IPO prospectus Sep 28, 2026; company timeline Sep 2026); Wikipedia (Anthropic); TechCrunch; Tracxn (funding rounds, secondary source); https://aistockwire.com/blog/anthropic-ipo-mid-november-broadcom-avgo-60-billion-chip-financing-october-2026; https://www.pymnts.com/news/artificial-intelligence/2026/anthropic-targets-pre-thanksgiving-ipo-at-2-trillion-valuation/ https://www.reuters.com/technology/anthropic-unveils-claude-duet-ai-agent-app-2026-10-21/; https://www.reuters.com/business/anthropic-adds-former-harvard-science-dean-oversight-body-2026-10-21/
3. XAI — PALO ALTO, UNITED STATES
Founded March 2023 (incorporated Mar 9, 2023; publicly announced Jul 2023) · Subsidiary of SpaceX (acquired Feb 2026; rebranded SpaceXAI Jul 2026)
Headquarters: Palo Alto, United States
Offices and footprint: HQ Palo Alto; Memphis, Tennessee (Colossus data-centre campus); additional offices not published in sources checked.
Website: x.ai
Contact: No published email or phone; official contact/sales route via x.ai
Published email: Not published
Phone: Not published
Founders: Elon Musk; Igor Babuschkin; Yuhuai (Tony) Wu; Kyle Kosic; Manuel Kroiss; Greg Yang; Zihang Dai; Toby Pohlen; Christian Szegedy; Guodong Zhang; Jimmy Ba; Ross Nordeen
Leadership: CEO / head: Elon Musk, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Anthony Armstrong, Chief Financial Officer; Igor Babuschkin, Co-founder and engineering lead (departed Aug 2025); Kyle Kosic, Co-founder; Board / advisers: Not published; Founders: Elon Musk; Igor Babuschkin; Yuhuai (Tony) Wu; Kyle Kosic; Manuel Kroiss; Greg Yang; Zihang Dai; Toby Pohlen; Christian Szegedy; Guodong Zhang; Jimmy Ba; Ross Nordeen
Ownership status: Subsidiary of SpaceX (acquired Feb 2026; rebranded SpaceXAI Jul 2026)
Ownership and funding: Series E $20B at $230B valuation, Jan 2026 (company announcement via x.ai/news/series-e; reported by Reuters/TechCrunch); investors: Valor Equity Partners, StepStone Group, Fidelity, Qatar Investment Authority, MGX, Baron Capital, Nvidia, Cisco; ~$42B total raised (media tally); acquired by SpaceX Feb 2026 at $250B xAI valuation ($1.25T combined); rebranded SpaceXAI Jul 2026
What it builds: Grok (Grok 4 series incl. Grok 4.7 released Sep 28, 2026; Grok 4 Heavy); Grok 5 (in training per company, Jan 2026); Grok Voice (Grok Voice 3.0, Oct 14, 2026; ranked #1 on Artificial Analysis Speech-to-Speech Index at 86.9, reported); Grok Imagine; Grokipedia; Grok Build; Grok Bot; X platform integration
Specialties: Frontier LLMs; reasoning; multimodal image and video generation; voice agents; large-scale training infrastructure; data center operations
Clients: Not disclosed in sources checked
What makes it distinctive: Built the Colossus supercomputer cluster (555,000 Nvidia GPUs by Jan 2026, targeting 1M) in Memphis. Real-time data access via the X platform is its core differentiator. Vertically integrated with X distribution (~600M monthly active users cited by the company).
Notes: Acquired by SpaceX in an all-stock deal Feb 2026; rebranded SpaceXAI Jul 2026 (SpaceX Nasdaq IPO Jun 12, 2026, ticker SPCX). Igor Babuschkin departed Aug 2025. Secondary reporting describes complete co-founder turnover within three years. X Corp. merged under X.AI Holdings Mar 2025. No published chief scientist title; no public contact email or phone. The X Subscription — unified subscription framework covering Grok, Grok Bot, Cursor and X Premium — outlined Oct 12, 2026, with four reported tiers (Plus, Premium, Super, Ultra) and updated consumer terms (Bloomberg/TokenPost, reported); no launch date or final pricing confirmed.
Sources: x.ai (company announcement); Reuters (SpaceX acquisition, Feb 3, 2026); TechCrunch; Wikipedia (SpaceXAI)
4. GOOGLE DEEPMIND — LONDON, UNITED KINGDOM
Founded September 2010 (incorporated Sep 23, 2010; launched Nov 2010) · Wholly owned subsidiary of Alphabet Inc. (acquired 2014)
Headquarters: London, United Kingdom
Offices and footprint: HQ London (King's Cross); research sites reported in Mountain View, New York, Paris, Zurich, Bangalore and elsewhere as part of Google — full list not published as a standalone roster.
Website: deepmind.google
Contact: No published email or phone; official contact/sales route via deepmind.google
Published email: Not published
Phone: Not published
Founders: Demis Hassabis; Shane Legg; Mustafa Suleyman
Leadership: CEO / head: Koray Kavukcuoglu, Senior Vice President, Google DeepMind (since Aug 5, 2026; previously CTO); Chief scientist / head of research: Shane Legg, Co-founder and Chief AGI Scientist; Demis Hassabis, Co-founder, Chair of Google DeepMind and Chief Scientist of Alphabet (since Aug 5, 2026); Other published C-suite / technical leadership: Lila Ibrahim, Chief Operating Officer; James Manyika, SVP Research, Labs, Technology and Society (Google); Demis Hassabis (above); Board / advisers: Not published; Founders: Demis Hassabis; Shane Legg; Mustafa Suleyman
Ownership status: Wholly owned subsidiary of Alphabet Inc. (acquired 2014)
Ownership and funding: Wholly owned subsidiary of Alphabet Inc. (acquired 2014, ~£400M reported); £1.33B revenue and £217M operating income in 2024 (per Companies House accounts, via Wikipedia); no external equity raises
What it builds: Gemini (Gemini 3 series, Nov 2025; Gemini 4 Argon announced Sep 30, 2026, globally available Oct 13, 2026 with robotics/physical-reasoning variant ('Video-to-Sim-to-Real' demos incl. Apptronik Apollo cornhole trial; 87% on Embodied Reasoning benchmark per company; intro API pricing $2/$10 per 1M tokens, rising to $4/$20 after Nov 30, 2026)); Gemma; Veo (video); Imagen (image); Nano Banana (image); Lyria (music); AlphaFold; AlphaGo; Antigravity (agentic coding platform)
Specialties: Frontier multimodal LLMs; reasoning; scientific AI (protein structure prediction); robotics; AGI research; diffusion-based video and music generation
Clients: Not disclosed in sources checked
What makes it distinctive: Nobel Prize in Chemistry 2024 (Hassabis and John Jumper) for AlphaFold protein-structure prediction. Roughly 4,000-6,000 staff after the April 2023 Google Brain merger. Launched the DeepMind Institute on Sep 16, 2026 as a public forum on AGI's societal implications.
Notes: Major leadership change Aug 5, 2026: Hassabis stepped back from day-to-day CEO to Chair of Google DeepMind + Alphabet Chief Scientist; Koray Kavukcuoglu took operational leadership as SVP. Jeff Dean left Google the same day after 27 years to launch an independent public benefit corporation with Sanjay Ghemawat. Mustafa Suleyman departed DeepMind in 2022 (later Microsoft AI, then back at Google partnerships). No public contact email or phone.
Sources: deepmind.google; Google/Alphabet company announcements (via Axios, aiweekly.co, Sep 2026); Wikipedia (Google DeepMind); The Times; The Information (via media summaries)
5. DEEPSEEK — HANGZHOU, CHINA
Founded July 2023 · Independent; controlled by founder Liang Wenfeng (majority ownership reported)
Headquarters: Hangzhou, China
Offices and footprint: HQ Hangzhou; additional offices not published in sources checked.
Website: deepseek.com
Contact: No published email or phone; official contact/sales route via deepseek.com
Published email: Not published
Phone: Not published
Founders: Liang Wenfeng
Leadership: CEO / head: Liang Wenfeng, Founder and Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Liang Wenfeng
Ownership status: Independent; controlled by founder Liang Wenfeng (majority ownership reported)
Ownership and funding: Self-funded from High-Flyer trading profits (2019-2023); first external round ~$7.4B (50B yuan) at ~$50-52B post-money, Jun 2026 (reported: Tencent ~$1.5B, CATL ~$740M, National AI Industry Investment Fund 1B yuan, plus NetEase, JD.com, IDG Capital, Monolith — all unconfirmed by company); Liang contributed ~$3B personally; Liang retains ~78% ownership per reports; annualized revenue run-rate crossed $1B (reported by The Information, Sep 2026); mainland/Hong Kong IPO rumored (unconfirmed) Oct 6, 2026: Bloomberg reports DeepSeek is close to securing at least ~$11.9B (80B yuan; final tally could approach 100B yuan) with Tencent and CATL among the largest commits, on strong investor demand following the DeepSeek-V4.1-Flash launch — unconfirmed by company; would far exceed the Jun 2026 round above. DeepSeek hired CITIC Securities to prepare for a potential STAR Market IPO (reported Sep 2026; timing, valuation and size undecided — unconfirmed).
What it builds: DeepSeek V4 (V4-Pro, V4-Flash; launched Apr 24, 2026); DeepSeek-V4.1-Flash (Sep 10, 2026); DeepSeek-V4.1-Flash-1017 (Oct 17, 2026; MIT weights on Hugging Face); DeepSeek-R1 (Jan 2025); DSpark (open-source inference framework, Jun 2026)
Specialties: Open-weight frontier LLMs; reasoning models; Mixture-of-Experts architectures; cost-efficient training and inference
Clients: Not disclosed in sources checked
What makes it distinctive: Released open weights under the MIT license (V4 family), bucking closed-lab norms. R1's claimed ~$6M training cost triggered the January 2025 'Sputnik moment' sell-off. Operates with a lean team (~160 employees reported in 2025) built largely on a pre-export-restriction Nvidia A100 cluster amassed by High-Flyer.
Notes: DeepSeek R2 was never released: reported delays from Huawei Ascend chip issues and Liang reportedly unsatisfied with performance (all reported, never confirmed by the company). deepseek-chat/deepseek-reasoner API aliases retired Jul 24, 2026 in favor of V4 models. Embodied-AI pivot: Bloomberg (Oct 12, 2026, reported) says DeepSeek agreed to lead a seed funding round for Hangzhou robotics startup Win Verse AI (founder: ex-Tencent chief scientist Hu Han, Swin Transformer co-author), valuing it at ~$400M — transaction not finalized at time of report. Follows an Aug 2026 investment in Unitree Robotics' Shanghai IPO (~140.8M yuan / ~$20.8M, per stock-exchange filing reported by Reuters). No chief scientist, research leads, board, or contact details published. Liang met President Xi Jinping at a 2025 tech-entrepreneur meeting (Forbes). DHS Exclusion and Removal Order #2026-01 (signed Jul 24, 2026 by DHS Secretary Markwayne Mullin under the Federal Acquisition Supply Chain Security Act; made public Oct 19-20, 2026; Reuters/TechCrunch/WSJ): US federal civilian agencies ordered to stop purchasing DeepSeek products/services and remove them — incl. third-party products and services built on DeepSeek technology — from IT systems over national security concerns (data-transfer risk to PRC); agencies may request exceptions within 60 days; businesses and individuals not covered.
Sources: deepseek.com; Forbes (Liang Wenfeng profile); Bloomberg (via media reports); The Information via wowtale (Sep 2026); todaysstartupnews.com; Wikipedia (DeepSeek); https://startupfortune.com/deepseek-leads-400-million-funding-round-for-ex-tencent-scientists-robot-ai-startup/
6. MISTRAL AI — PARIS, FRANCE
Founded April 2023 (Apr 28, 2023) · Independent (private)
Headquarters: Paris, France
Offices and footprint: HQ Paris; Munich hub opened Sep 2026 (company); additional offices reported in London and Palo Alto — full official list not published in sources checked.
Website: mistral.ai
Contact: No published email or phone; official contact/sales route via mistral.ai
Published email: Not published
Phone: Not published
Founders: Arthur Mensch; Guillaume Lample; Timothée Lacroix
Leadership: CEO / head: Arthur Mensch, Co-founder and Chief Executive Officer; Chief scientist / head of research: Guillaume Lample, Co-founder and Chief Scientist (also reported as Chief Science Officer); Other published C-suite / technical leadership: Timothée Lacroix, Co-founder and Chief Technology Officer; Board / advisers: Not published; Founders: Arthur Mensch; Guillaume Lample; Timothée Lacroix
Ownership status: Independent (private)
Ownership and funding: Series D €3B at >€21B post-money (~$24B), Sep 8, 2026 (company announcement; company calls it the largest equity round ever by a European tech company); led by Samsung Electronics, co-led by Scaleup Europe Fund (EQT) and PSG Equity; investors: Advent, BlackRock-managed funds, Grand Duchy of Luxembourg, ASML, Nvidia, a16z, General Catalyst, Lightspeed, Salesforce Ventures, BNP Paribas CIB, Bpifrance; Series C €1.7B at €11.7B, Sep 2025 (led by ASML); ~€850M debt raised Apr 2026; total raised >$6B (media tally)
What it builds: Mistral Large 3; Mistral Medium 3.5; Mistral Small 4; Ministral 3; Codestral; Magistral 1.2; Devstral 2; Voxtral; Mistral OCR 3; Le Chat; Shieldstral (open-weight safety classifier, Sep 2026); Mistral Large 4 ('Le Chonk', public preview Oct 6, 2026 — ~1T total / 49B active params, 1M-token context; API live; open weights promised Oct 27, 2026; company claims strongest open-weight model built outside China)
Specialties: Open-weight LLMs; frontier dense and MoE models; code generation; document OCR; reasoning models; embodied navigation models; sovereign enterprise deployments
Clients: Named in company/press reporting: Airbus, ASML, HSBC among 125+ enterprise customers cited by the company (Sep-Oct 2026).
What makes it distinctive: Only frontier-model developer in continental Europe after Aleph Alpha's 2024 pivot. Open-weight-first strategy with European data-center infrastructure, positioned as Europe's sovereign AI champion — two-thirds of Series D capital from European investors per the company. 1,000+ employees (2026) and 125+ enterprise customers cited by the company (Sep 2026).
Notes: Guillaume Lample's title appears as both Chief Scientist and Chief Science Officer across reports. Munich hub opened and partnerships with Mozilla and Tata Consultancy Services announced alongside Series D (Sep 2026). No public contact email or phone published. Named enterprise customers reported Oct 2026 include Airbus, ASML and HSBC (secondary reporting of company statements); the company cites 125+ enterprise customers (Sep 2026). Mistral Large 4 open weights: the company promised release 'by the end of October 2026'; the specific Oct 27 date is press-reported, not stated on Mistral's own channels in sources checked — still pending at Oct 22, 2026.
Sources: mistral.ai (company announcement); Le Figaro (Sep 30, 2026); Slator (Sep 2026); TechCrunch via UpMarket (secondary); Wikipedia (Mistral AI); https://www.datacamp.com/blog/mistral-4-large-le-chonk; https://tech-insider.org/mistral-large-4-le-chonk-1-05t-parameters-2026/
7. COHERE — TORONTO, CANADA
Founded 2019 · Independent (private); definitive merger with Aleph Alpha signed Sep 16, 2026 — not yet closed
Headquarters: Toronto, Canada
Offices and footprint: HQ Toronto; Berlin to become dual headquarters on the Aleph Alpha merger closing, with Heidelberg retained as a research centre (reported); offices also reported in London, New York and San Francisco.
Website: cohere.com
Contact: No published email or phone; official contact/sales route via cohere.com
Published email: Not published
Phone: Not published
Founders: Aidan Gomez; Nick Frosst; Ivan Zhang
Leadership: CEO / head: Aidan Gomez, Co-founder and Chief Executive Officer; Chief scientist / head of research: Phil Blunsom, Chief Scientist; Joëlle Pineau, Chief AI Officer; Other published C-suite / technical leadership: Ivan Zhang, Co-founder and Chief Technology Officer; Ilhan Scheer, incoming Chief Operating Officer (on Aleph Alpha closing); Samuel Weinbach, incoming Chief Research Officer (on closing); Board / advisers: No full board published; company names AI-pioneer investors: Geoffrey Hinton, Fei-Fei Li, Pieter Abbeel, Raquel Urtasun (company statement, Sep 16, 2026); Founders: Aidan Gomez; Nick Frosst; Ivan Zhang
Ownership status: Independent (private); definitive merger with Aleph Alpha signed Sep 16, 2026 — not yet closed
Ownership and funding: Series E $2.5B at ~$20B valuation in advanced talks (Globe and Mail, Sep 11, 2026; NOT confirmed closed; Schwarz Group leading with €500M committed, Apr 2026); ~$1.6B raised to date (company statement, Sep 16, 2026); investors: Nvidia, AMD Ventures, Salesforce Ventures, Oracle, Cisco, Radical Ventures, Inovia Capital, PSP Investments, HOOPP, BDC, Nexxus; prior $500M at $5.5B valuation, Jul 2024
What it builds: Command A+ (May 20, 2026); Command A Reasoning; Command A Translate; Command A Vision; Cohere North; Aya Vision; Command R/R+ Embed 5 (launched Oct 2026; retrieval/search across 100+ languages — AI Magazine, secondary).
Specialties: Enterprise LLMs; private, on-premises and air-gapped deployments; retrieval-augmented generation; multilingual models (Aya); regulated industries (finance, healthcare, public sector)
Clients: Client categories: governments and regulated industries (finance, healthcare, public sector); Oracle partnership (100+ generative-AI use cases reported) and CoreWeave (customer-support deployment via North) reported Oct 2026 (secondary).
What makes it distinctive: Enterprise-only strategy — no consumer product — with models designed to run inside the customer's own infrastructure for data sovereignty. Billed as the 'sovereign AI' alternative to US hyperscalers, with Canadian and EU government backing. $240M ARR reported for 2025.
Notes: Definitive merger agreement with Germany's Aleph Alpha signed Sep 16, 2026: combined company to operate as Cohere with dual headquarters Toronto + Berlin, ~$20B combined value per April term sheet (Handelsblatt) — subject to regulatory approval, NOT yet closed. Schwarz Group plans €11-13B data-center campus via StackIT. PitchBook lists the Series E ($2.5B) as 'In Progress' dated Apr 24, 2026 (secondary source). Merger status at Oct 22, 2026: still NOT closed; expected to close before the end of 2026 subject to regulatory approvals (reported clearances include the Canadian Competition Bureau, the German Bundeskartellamt and the European Commission). Aleph Alpha's Heidelberg office to be retained as a research centre on closing (reported).
Sources: cohere.com (Sep 16, 2026 press release); Reuters (Sep 16, 2026); Globe and Mail (Sep 11, 2026); PitchBook (secondary) https://aimagazine.com/news/cohere-sovereign-ai-the-aleph-alpha-merger-us-20bn-value
8. AI21 LABS — TEL AVIV, ISRAEL
Founded November 2017 · Independent (private)
Headquarters: Tel Aviv, Israel
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: ai21.com
Contact: No published email or phone; official contact/sales route via ai21.com
Published email: Not published
Phone: Not published
Founders: Yoav Shoham; Ori Goshen; Amnon Shashua
Leadership: CEO / head: Ori Goshen, Co-founder and Co-CEO (with Yoav Shoham, Co-founder and Co-CEO); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Amnon Shashua, Chairman; Yoav Shoham, Co-founder and Co-CEO; Board / advisers: Advisors per company press release: Sebastian Thrun; Dan Jurafsky; Chris Ré; Dan Roth; Kevin Leyton-Brown; Shai Shalev-Shwartz; Zoubin Ghahramani; Founders: Yoav Shoham; Ori Goshen; Amnon Shashua
Ownership status: Independent (private)
Ownership and funding: Series C $208M at $1.4B valuation, Nov 2023 (company announcement; Intel Capital, Comcast Ventures); Series D $300M, May 2025 (reported by Business Insider, confirmed by Calcalist; Google and Nvidia; valuation undisclosed); $336M total through Series C (company, Nov 2023); $626.5M across 9 rounds as of Jan 2026 (media); Nvidia acquisition talks at $2-3B valuation reported Jan 2026 (outcome unknown)
What it builds: Jamba (Jamba 1.5 Mini/Large; hybrid SSM-Transformer, open source under Apache 2.0, 256K context); Jurassic-1; Jurassic-2; Maestro (AI orchestration); Wordtune; AI21 Studio (developer platform)
Specialties: Hybrid State Space Model + Transformer architectures; long-context LLMs; enterprise AI systems; task-specific models; writing and reading assistants
Clients: Not disclosed in sources checked
What makes it distinctive: Pioneer of the hybrid SSM-Transformer architecture, designed for long-context processing at lower compute cost than pure transformers. Builds full-stack enterprise AI systems (foundation models plus orchestration tooling) rather than standalone models, targeting reliable, low-hallucination business applications.
Notes: Wikipedia also lists Maykel Zayton as a co-founder — conflicts with the company's official three-founder account. Nvidia acquisition talks (Jan 2026) reported but outcome never confirmed. No named chief scientist published. Series D round details (valuation, close) not officially disclosed.
Sources: ai21.com; TechCrunch (Nov 2023); SiliconANGLE; Nasdaq/company press release (Jun 2022); Wikipedia (AI21 Labs)
9. MINIMAX — SHANGHAI, CHINA
Founded December 2021 · Publicly listed (HKEX: 100, Jan 9, 2026)
Headquarters: Shanghai, China
Offices and footprint: HQ Shanghai; additional offices not published in sources checked.
Website: minimax.io
Contact: Published email: model@minimax.io; also via minimax.io
Published email: model@minimax.io
Phone: Not published
Founders: Yan Junjie; Yun Yeyi; Zhou Yucong
Leadership: CEO / head: Yan Junjie, Founder, Chairman, CEO and CTO (per company investor relations page); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Yun Yeyi, Chief Operating Officer; Board / advisers: Not published; Founders: Yan Junjie; Yun Yeyi; Zhou Yucong
Ownership status: Publicly listed (HKEX: 100, Jan 9, 2026)
Ownership and funding: Publicly listed: Hong Kong IPO Jan 9, 2026 (SEHK: 100), raised HK$4.8B (~$620M) at ~$6.5B valuation; shares +109% on debut; market cap later ~$33B (media, Jun 2026); ~$2B financing announced Jul 10, 2026 via share placement + zero-coupon convertible bonds (investors unnamed; per company thread); pre-IPO ~$850M across 7 rounds (media); investors: Alibaba, Tencent, MiHoYo, Hillhouse, HongShan (Sequoia China), IDG Capital, ZhenFund
What it builds: MiniMax M3; MiniMax M3.1 Flash (launched Oct 19, 2026; previously 'Space Bunny' stealth model; open weights promised Oct 23); MiniMax M2 series (M2, M2.1, M2.5, M2.7); MiniMax Code; Hailuo 2.3 (video generation); Speech 2.8; Music 3.0; Talkie; Harness
Specialties: Multimodal AI (text, image, audio, video, music); agent-native models; text-to-video generation; AI character apps; ultra-long context
Clients: Not disclosed in sources checked
What makes it distinctive: Multimodal-from-day-one strategy since 2021. M2 became the top-ranked open-weight model globally within weeks, then the company pivoted newer M2.7 to proprietary licensing. China's second 'AI Tiger' to list publicly, one day after Zhipu AI. $79M revenue in 2025 (per HKEX filing, via Wikipedia) with ~300M consumer users cited by the company.
Notes: Founders list conflicts: Wikipedia names Yan Junjie, Yun Yeyi, Zhou Yucong; Reuters names Yan Junjie, Yang Bin, Zhou Yucong. Founding date: Dec 2021 (Wikipedia) vs early 2022 (Reuters). Jul 2026 $2B financing: investors, valuation, and structure undisclosed. No published chief scientist. Contact email model@minimax.io published on official GitHub page (model inquiries).
Sources: minimax.io; MiniMax official GitHub (minimax-ai.github.io); Wikipedia (MiniMax Group); Reuters (Feb 2026); Bloomberg (via Wikipedia); company X thread via runtimewire (Jul 2026)
10. MOONSHOT AI — BEIJING, CHINA
Founded March 2023 · Independent (private)
Headquarters: Beijing, China
Offices and footprint: HQ Beijing; additional offices not published in sources checked.
Website: moonshot.ai
Contact: No published email or phone; official contact/sales route via moonshot.ai
Published email: Not published
Phone: Not published
Founders: Yang Zhilin; Zhou Xinyu; Wu Yuxin
Leadership: CEO / head: Yang Zhilin, Founder and Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Yang Zhilin; Zhou Xinyu; Wu Yuxin
Ownership status: Independent (private)
Ownership and funding: Series F $3.5B at $35B post-money, Jul 2026 (Bloomberg/36Kr; led by China's National AI Industry Investment Fund); Series E ~$2B at $20B, May 2026 (reported); total raised >$4B disclosed (media); investors: Alibaba, Tencent, Meituan, HongShan, IDG Capital, China Mobile, HSG; Oct 6, 2026 (Bloomberg, reported, unconfirmed): CLOSED final private round at ~$50B valuation; Hong Kong IPO targeted for Q1 2027 aiming to raise up to $5B (earlier Reuters Sep 3 report had cited a $3B target); underwriters: CICC, Deutsche Bank, Goldman Sachs, Bank of America; early-look investor meetings starting Oct 2026; company converted from LLC to joint-stock company (listing prep step). ARR reportedly rose ~$300M (Jun) to >$1B (Aug 2026) after K3, with Bloomberg sources expecting ~$2B by Dec (reported)
What it builds: Kimi K3 (Jul 16, 2026; 2.8T parameters, open weights, 1M context); Kimi (chatbot); Kimi Researcher; Kimi Agent; Kimi Code; Kimi Audio; Kimina Prover
Specialties: Open-weight frontier LLMs; long-context models; multimodal input (image/video); agentic deep research (Swarm sub-agents); coding agents
Clients: Not disclosed in sources checked
What makes it distinctive: Kimi K3 is the largest open-weight model ever released (2.8T parameters) and, on reported benchmarks, rivaled US frontier models. Long-context pioneer lineage: founder Yang Zhilin co-authored Transformer-XL and XLNet. One of China's 'AI Tigers'; second most valuable privately held AI company in China after DeepSeek per reports (Jul 2026).
Notes: ARR, funding, and IPO figures are all reported, not company-confirmed. Shareholder dispute: arbitration with former Recurrent.ai investors (reported Oct 12, 2026) — claimants seek additional equity compensation reportedly exceeding $700M at an earlier $30B valuation — unconfirmed. The Information reported a Chinese regulator data-security probe into Moonshot and DeepSeek (unconfirmed). Anthropic accused Moonshot (and other Chinese labs) of distilling Claude models during 2026 (reported); a US congressional committee pressed US companies over Kimi use from April 2026 (reported). K3's custom license requires up to 30% revenue sharing from inference providers earning >$20M/yr. No chief scientist, research leads, board, or contact details published.
Sources: moonshot.ai; Wikipedia (Moonshot AI); Bloomberg via channelinsider/phronews (Jul-Sep 2026); LatePost via ainvest (Sep 2026); UpMarket (secondary); theairankings.com; https://www.thehindubusinessline.com/markets/moonshot-said-to-eye-early-2027-ipo-after-value-hits-50-billion/article71549829.ece/amp/; https://cryptosbank.in/2026/10/12/moonshot-ai-settles-dispute-ahead-of-planned-hong-kong-ipo/
MORE FRONTIER LABS AND THE CODING AGENTS
11. ZHIPU AI (Z.AI) — BEIJING, CHINA
Founded 2019 · Publicly listed (HKEX: 2513, Jan 8, 2026)
Headquarters: Beijing, China
Offices and footprint: HQ Beijing; additional offices not published in sources checked.
Website: https://z.ai
Contact: No published email or phone; official contact/sales route via https://z.ai
Published email: Not published
Phone: Not published
Founders: Tang Jie; Li Juanzi
Leadership: CEO / head: Zhang Peng, Chief Executive; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Liu Debing (Chairman; key controlling shareholder with founder Tang Jie, per caproasia reporting, Jan 2026); Founders: Tang Jie; Li Juanzi
Ownership status: Publicly listed (HKEX: 2513, Jan 8, 2026)
Ownership and funding: Publicly listed (HKEX: 2513, Knowledge Atlas Technology) since Jan 8, 2026 IPO at HK$116.2, raising $558M at ~$6.5B valuation, closing day one at ~$7.4B market value, per caproasia reporting Jan 2026; Sep 2026: raised ~$5B via $2B share issuance (10% discount to Nov 9 2026 closing price) + $3B zero-coupon convertible bond at ~$43B market value, per caproasia Sep 2026 (secondary source, verify independently); raised ~$1.5B privately since founding per 2025 reporting, including $140M Jul 2025 and $411M Dec 2024 rounds; key investors include Alibaba, Ant Group, Tencent, Meituan, Xiaomi, HongShan, Saudi Aramco/Prosperity7 Ventures
What it builds: GLM (General Language Model) family; open-weight GLM releases under MIT License since Jul 2025; vision-language and text-to-video models; AMiner (research database)
Specialties: Large language models; multimodal models; open-weight LLMs; knowledge graphs; enterprise AI (MaaS platform)
Clients: Not disclosed in sources checked
What makes it distinctive: One of China's six 'AI tiger' companies and the third-largest LLM market player in China per IDC (2024). The first major Chinese LLM company to IPO (Hong Kong, Jan 2026). Placed on the US Commerce Department Entity List in Jan 2025, barring purchases of advanced Nvidia H100/H200 chips.
Notes: FOUNDING YEAR CONFLICT: Wikipedia and caproasia report 2019 founding by Tsinghua professors Tang Jie and Li Juanzi; one secondary source (ainvest) says 2022 founded at Tsinghua University Science Park — the 2019 date is used with the conflict flagged. Rebranded internationally as Z.ai in 2025 (Zhipu AI name retained in China). Chief scientist/head of research and key engineers not published — cells left blank. No public contact email or phone found. Sep 2026 $5B raise figures come from a single secondary source (caproasia) — verify independently before quoting.
Sources: https://en.wikipedia.org/wiki/Zhipu_AI; https://www.caproasia.com/2026/01/08/china-artificial-intelligence-startup-zhipu-ai-knowledge-atlas-technology-hong-kong-ipo-share-price-increased-13-1-on-day-1-trading-8-1-26-ipo-price-hkd-116-2-closing-hkd-131-5-to-7-4-billion/; https://www.caproasia.com/2026/09/15/china-based-43-billion-hong-kong-listed-artificial-intelligence-company-z-ai-zhipu-ai-knowledge-atlas-technology-raised-5-billion-via-2-billion-share-issuance-10-discount-to-11-9-26-closing-pri/
12. SAFE SUPERINTELLIGENCE (SSI INC.) — PALO ALTO (WITH TEL AVIV, ISRAEL OFFICE), UNITED STATES
Founded 2024 · Independent (private)
Headquarters: Palo Alto (with Tel Aviv, Israel office), United States
Offices and footprint: Palo Alto (HQ) and Tel Aviv, Israel office (company).
Website: https://ssi.inc
Contact: No published email or phone; official contact/sales route via https://ssi.inc
Published email: Not published
Phone: Not published
Founders: Ilya Sutskever; Daniel Gross; Daniel Levy
Leadership: CEO / head: Ilya Sutskever, CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Daniel Levy, co-founder (led OpenAI's optimization team; Principal Scientist title cited by some secondary sources but not officially published — see notes); Board / advisers: Not published; Founders: Ilya Sutskever; Daniel Gross; Daniel Levy
Ownership status: Independent (private)
Ownership and funding: Private; raised ~$8B total per ai-magazine (Aug 2026, secondary); $1B Sep 2024 at $5B valuation led by a16z, Sequoia Capital, SV Angel (press reporting); rounds reaching $32B valuation reported Jan/Apr 2025-2026 (investors incl. Greenoaks, Lightspeed, DST Global, Alphabet, Nvidia); Nvidia strategic investment and compute partnership announced Jul 27, 2026 — Bloomberg characterized the check as 'substantial' with terms undisclosed; one secondary source cites $5B (unverified — see notes); Google Cloud TPU partnership; investors incl. Andreessen Horowitz, Sequoia, Greenoaks, DST Global, SV Angel, Alphabet, Nvidia
What it builds: No publicly released products or models as of Aug/Sep 2026 (company has stated its first product will be safe superintelligence itself)
Specialties: Frontier AI research; superintelligence alignment; AI safety research
Clients: Not disclosed in sources checked
What makes it distinctive: A 'straight-shot' research lab pursuing safe superintelligence as its sole goal with no product roadmap or commercial pressure, per its founding statement. Roughly 20-50 employees with zero shipped products, papers, or revenue as of Aug 2026, yet capitalized at ~$32B — among the best-funded AI labs in the world.
Notes: Daniel Gross left SSI in Jul 2025 to lead Meta Superintelligence Labs; Sutskever became CEO at that time (Wikipedia). Daniel Levy's exact title is not officially published — some secondary sources call him Principal Scientist; flagged rather than stated as fact. No published chief scientist, advisors, or contact details — cells left blank. Nvidia investment amount ($5B) cited by startupfortune but contradicts fourweekmba (terms undisclosed, Bloomberg 'substantial'); figures flagged.
Sources: https://en.wikipedia.org/wiki/Safe_Superintelligence_Inc.; https://ai-magazine.com/the-complete-list-of-28-us-ai-startups-to-earn-over-100-million-in-2026/; https://fourweekmba.com/ai-nvidia-safe-superintelligence-ssi-vera-rubin-partnership/; https://startupfortune.com/ilya-sutskevers-ssi-has-raised-8-billion-and-shipped-nothing-at-all/; http://stockanalysis.com/private/safe-superintelligence/
13. THINKING MACHINES LAB — SAN FRANCISCO, UNITED STATES
Founded 2025 · Independent; public-benefit corporation (private)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco (2300 Harrison Street, Mission District, from Aug 2026).
Website: https://thinkingmachines.ai
Contact: No published email or phone; official contact/sales route via https://thinkingmachines.ai
Published email: Not published
Phone: Not published
Founders: Mira Murati; John Schulman; Barret Zoph; Lilian Weng; Andrew Tulloch; Luke Metz
Leadership: CEO / head: Mira Murati, Chief Executive Officer; Chief scientist / head of research: John Schulman, Chief Scientist; Other published C-suite / technical leadership: Soumith Chintala, Chief Technology Officer (succeeded Barret Zoph; co-creator of PyTorch); Lilian Weng (founding researcher, former OpenAI VP); Jonathan Lachman (former OpenAI head of special projects); Luke Metz (founding team, departed back to OpenAI Jan 2026); Andrew Tulloch (Chief Architect, later recruited to Meta Superintelligence Labs); Board / advisers: Bob McGrew (adviser; previously OpenAI chief research officer); Alec Radford (adviser; OpenAI lead researcher); Founders: Mira Murati; John Schulman; Barret Zoph; Lilian Weng; Andrew Tulloch; Luke Metz
Ownership status: Independent; public-benefit corporation (private)
Ownership and funding: Private; public-benefit corporation; ~$2B seed Jul 2025 at $10B pre-money / $12B post-money led by Andreessen Horowitz with Nvidia, Accel, ServiceNow, Cisco, AMD, Jane Street, GV, Lightspeed, Conviction (Reuters/TechCrunch reporting); government of Albania participated per ainvest (secondary); Sep 2026: reported talks for $5-6B at $40B pre-money valuation led by Accel with Nvidia participating — NOT closed as of Sep 2026 (flagged); earlier $50-60B round ambition collapsed in early 2026; Mar 2026 strategic partnership with Nvidia (undisclosed investment + multi-year agreement for 1 GW Vera Rubin compute)
What it builds: Tinker (API for fine-tuning language models, launched Oct 1, 2025); Inkling (open-weight LLM, 975B parameters, Apache License, released Jul 15, 2026); Inkling Small (276B parameters, released Jul 31, 2026)
Specialties: Multimodal AI systems; open-weight foundation models; model fine-tuning infrastructure (Tinker); collaborative general intelligence
Clients: Not disclosed in sources checked
What makes it distinctive: Structured as a public-benefit corporation pursuing 'collaborative general intelligence' with an open-source emphasis. Shipped the leading non-Chinese open-weights model (Inkling) in Jul 2026, built on DeepSeek-V3 architecture and Kimi K2.5 post-training synthetic data, while generating early revenue (run rate above $100M in 2026, per press) from the Tinker fine-tuning platform.
Notes: Founding team departed talent: Barret Zoph (original CTO) and Luke Metz returned to OpenAI Jan 2026; Andrew Tulloch left for Meta Superintelligence Labs. Soumith Chintala as CTO reported by Feb 2026 profile; official site (archived Mar 2026) still listed Zoph as CTO — use current reporting with flag. HQ moved to 2300 Harrison St, Mission District, San Francisco in Aug 2026. Website URL from company (thinkingmachines.ai not directly opened; verify before publishing).
Sources: https://pengen.diewe.workers.dev/thegdsks/openclaw-https-en.wikipedia.org/wiki/Thinking_Machines_Lab; https://web.archive.org/web/20260311093527/https://thinkingmachineslab.site/; https://forkast.news/thinking-machines-lab-seeks-40b-valuation-as-nvidia-extends-its-capital-allocator-role-into-model-space/; https://www.kucoin.com/news/flash/thinking-machines-valuation-surges-to-40b-amid-10b-fundraising-talks; https://thewomenweadmire.com/2026/02/mira-murati-navigating-thinking-machines-labs-turbulent-launch-amid-ai-talent-wars/
14. CHARACTER.AI — PALO ALTO, UNITED STATES
Founded 2021 · Independent (private); in Chapter 11 bankruptcy since Oct 12, 2026 — court-supervised sale under way
Headquarters: Palo Alto, United States
Offices and footprint: HQ Palo Alto; additional offices not published in sources checked.
Website: https://character.ai
Contact: No published email or phone; official contact/sales route via https://character.ai
Published email: Not published
Phone: Not published
Founders: Noam Shazeer; Daniel De Freitas
Leadership: CEO / head: Colin Adams, Chief Executive Officer (announced Oct 12, 2026, per company bankruptcy statement); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Sarah Wang, board member (Andreessen Horowitz General Partner; joined at Series A, Mar 2023); Founders: Noam Shazeer; Daniel De Freitas
Ownership status: Independent (private); in Chapter 11 bankruptcy since Oct 12, 2026 — court-supervised sale under way
Ownership and funding: Private; $150M Series A at $1B valuation Mar 2023 led by Andreessen Horowitz with Nat Friedman, Elad Gil, SV Angel, A Capital (company announcement via Business Wire); Aug 2024: Google non-exclusive technology license and hiring of founders Noam Shazeer and Daniel De Freitas plus part of research team — total deal ~$2.5-2.7B with investors bought out at ~$2.5B implied valuation (Axios/Reuters reporting); company remained independent. Oct 12, 2026: filed for Chapter 11 bankruptcy protection (US); listed estimated assets and liabilities of $50-100M each (court filing, reported by Reuters/TechCrunch); Disney agreed as stalking-horse bidder at $20M for certain assets incl. a non-exclusive perpetual license to certain IP, plus transition services via former Character.AI employees who joined Disney; court-supervised sale process expected to conclude late Nov/early Dec 2026; platform continues operating during the process — to be wound down if no buyer willing to operate it is found (Reuters, Oct 12, 2026)
What it builds: Character.AI (character chat platform); c.ai+ (subscription); voice; Scenes; Streams; AvatarFX; AI-native social feed
Specialties: Conversational AI companions; user-created AI characters; role-play and entertainment AI; consumer AI safety
Clients: Not disclosed in sources checked
What makes it distinctive: A consumer conversational-AI platform reporting ~20M monthly users with roughly 75 minutes daily engagement per user (company-reported, 2025). After the Aug 2024 Google licensing deal it pivoted from training frontier models to entertainment-focused AI built on open-source models.
Notes: Karandeep Anand (CEO since 2025, previously Brex president/CPO) departed effective Oct 2, 2026 to become Disney's first CTO (Reuters); several technical staff moved with him; Colin Adams named CEO in the Oct 12, 2026 bankruptcy announcement. ~20M monthly active users (recent market estimates, reported Oct 2026), down from 28M peak in mid-2024. Founders Shazeer and De Freitas are at Google. Safety litigation: Jan 2026 settlement of family lawsuits (terms undisclosed); May 2026 Pennsylvania lawsuit over bots presenting as licensed medical professionals unresolved as of Sep 2026 (Turing Post). No published chief scientist, key engineers, or contact details. Bankruptcy detail (court filings, reported Oct 18, 2026 by TechCrunch via Bloomberg): case No. 26-11603 (BLS), US Bankruptcy Court for the District of Delaware; Disney agreed to make employment offers to at least 60 Character.AI employees — the majority of the workforce — at an estimated $60M in annual compensation, and the vast majority accepted; the Disney bid remains a stalking-horse bid open to higher offers.
Sources: https://techcrunch.com/2023/05/31/character-ai-the-a16z-backed-chatbot-startup-tops-1-7m-installs-in-first-week/102; https://www.turingpost.com/p/character-ai; https://www.reuters.com/technology/disney-names-characterais-karandeep-anand-its-first-tech-chief-2026-09-18/; https://siliconangle.com/2023/03/23/character-ai-personalized-chatbot-service-raises-150m-1b-valuation/; https://www.ibtimes.sg/who-karandeep-anand-disneys-first-ever-cto-former-character-ai-ceo-93952; https://techcrunch.com/2026/10/13/just-5-years-in-interactive-chatbot-platform-character-ai-files-for-bankruptcy/; https://kelofm.com/2026/10/12/character-ai-files-for-bankruptcy-agrees-to-sell-some-assets-to-disney-for-20-million/ https://techcrunch.com/2026/10/18/disney-agrees-to-hire-the-majority-of-character-ais-workforce/
15. SAKANA AI — TOKYO, JAPAN
Founded 2023 · Independent (private)
Headquarters: Tokyo, Japan
Offices and footprint: HQ Tokyo; additional offices not published in sources checked.
Website: https://sakana.ai
Contact: No published email or phone; official contact/sales route via https://sakana.ai
Published email: Not published
Phone: Not published
Founders: David Ha; Llion Jones; Ren Ito
Leadership: CEO / head: David Ha, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Llion Jones, Chief Technology Officer (co-author of the original Transformer paper); Ren Ito, Chairman and Chief Operating Officer; Board / advisers: Not published; Founders: David Ha; Llion Jones; Ren Ito
Ownership status: Independent (private)
Ownership and funding: Private; $135M (JPY 20B) Series B Nov 17, 2025 at $2.65B post-money ($2.5B pre-money) per CEO David Ha to TechCrunch — investors MUFG, Khosla Ventures, Macquarie Capital, NEA, Lux Capital, In-Q-Tel; total raised ~$379M per TechCrunch syndication / over $400M per other secondary sources (figures vary — flagged); strategic investors added through early 2026: Google, Citigroup, Salesforce Ventures, Mitsubishi Electric (secondary reporting); Japan's most valuable AI startup
What it builds: Sakana Fugu (multi-model orchestration system, shipped Jun 22, 2026); Japanese-language optimized generative AI models
Specialties: Nature-inspired AI; model merging; evolutionary optimization; multi-agent orchestration; Japanese-language and culture-optimized LLMs; cost-efficient post-training for small datasets; enterprise AI for finance, industry, manufacturing, government
Clients: Not disclosed in sources checked
What makes it distinctive: Tokyo-based lab pursuing nature-inspired methods — model merging, evolutionary optimization, multi-agent orchestration — as an alternative to scaling single transformer architectures, with models specifically optimized for Japanese language and culture. Positioned in the 'sovereign AI' wave alongside Mistral, DeepSeek, and AI21 Labs.
Notes: Total-funding figure varies by source ($379M vs $400M+) — both secondary, flagged. No published chief scientist/head of research (CTO Llion Jones listed under key engineers); no published advisors or contact details. Founded July 2023 per secondary sources; 150+ employees. Website URL from standard company domain — verify before publishing.
Sources: https://techcrunch.com/2025/11/17/sakana-ai-raises-135m-series-b-at-a-2-65b-valuation-to-continue-building-ai-models-for-japan/; https://earlyterms.com/term/sakana-ai; https://ikh4ever.medium.com/sakana-ai-inside-japans-nature-inspired-frontier-ai-company-ce4480d495c7; https://digitalmarketreports.com/news/53862/sakana-ai-raises-135m-at-2-65b-valuation-to-advance-japan-focused-ai-models/
16. CURSOR (ANYSPHERE, INC.) — SAN FRANCISCO, UNITED STATES
Founded 2022 · Wholly owned subsidiary of SpaceX (acquisition effective Aug 14, 2026)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://cursor.com
Contact: No published email or phone; official contact/sales route via https://cursor.com
Published email: Not published
Phone: Not published
Founders: Michael Truell; Sualeh Asif; Aman Sanger; Arvid Lunnemark
Leadership: CEO / head: Michael Truell, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Sualeh Asif, Chief Product Officer; Jordan Topoleski, Chief Operating Officer (per Wikipedia, cited to company sources — see notes); Board / advisers: Not published; Founders: Michael Truell; Sualeh Asif; Aman Sanger; Arvid Lunnemark
Ownership status: Wholly owned subsidiary of SpaceX (acquisition effective Aug 14, 2026)
Ownership and funding: Acquired by SpaceX in an all-stock deal at ~$60B implied equity value; merger effective Aug 14, 2026; Anysphere now a wholly owned SpaceX subsidiary (SEC closing filing details reported by Market Business News; also reported by InsiderMonkey, Wisevoter). OpenAI's official statement on winding down Cursor's OpenAI model access (Oct 2026) confirms the acquisition. Pre-acquisition: $8M seed Oct 2023 led by OpenAI Startup Fund (angels Nat Friedman, Arash Ferdowsi); $60M Series A at $400M; ~$2.5B bidding round Nov 2024 (Benchmark, Index per TechCrunch); $900M Series C at $9.9B post-money Jun 2025 led by Thrive Capital; $2.3B Series D at $29.3B Nov 2025; Apr 2026 CNBC report of talks to raise $2B+ at $50B+ with a16z and Thrive co-leading, Nvidia strategic; investors included OpenAI Startup Fund, Andreessen Horowitz, Thrive Capital, Accel, DST, Coatue, Nvidia, Google; ~$3B ARR (May 2026 per Wikipedia citing press)
What it builds: Cursor (AI code editor and coding agent); Cursor Composer; Graphite
Specialties: AI coding assistants; agentic IDEs; AI-first software development environments; AI inference for code
Clients: Client categories: software teams and enterprises; individual client list not disclosed in sources checked.
What makes it distinctive: An AI-first fork of Visual Studio Code rebuilt as a coding agent rather than a plugin; in Nov 2025 launched its own inference model, cutting third-party model costs. Reported as the fastest B2B software company to reach $1B ARR, outpacing Slack, Zoom, and Snowflake (per tech press).
Notes: SpaceX acquisition CONFIRMED: $60B all-stock, effective Aug 14, 2026 (SEC filing, reported by Market Business News; corroborated by InsiderMonkey and Wikipedia). OpenAI notified SpaceX it will wind down its contract providing OpenAI models to Cursor, proposed shutoff Nov 12, 2026 — invoking the change-of-control clause; OpenAI cited inability to be confident SpaceX would use the technology within its terms of service, and that the upcoming Astra model would not be provided to Cursor (OpenAI official statement, Oct 2026). Jordan Topoleski as COO appears via Wikipedia citation only — unverified. CEO title for Truell verified. No published chief scientist, advisors, or contact details. Website cursor.com from Wikipedia infobox — verify before publishing. The X Subscription (Oct 12, 2026): SpaceXAI's unified subscription covering Grok, Grok Bot, Cursor and X Premium — four tiers (Plus/Premium/Super/Ultra) reported by Bloomberg; consumer terms updated Oct 12, 2026; no launch date or final pricing confirmed (TokenPost, reported). Cursor co-founder Michael Truell says OpenAI models account for ~5% of Cursor's AI traffic (reported, Oct 2026).
Sources: https://en.wikipedia.org/wiki/Cursor_(company); https://simple.wikipedia.org/wiki/Cursor_(company); https://techfundingnews.com/cursor-anysphere-2b-funding-50b-valuation-ai-coding/; https://www.caproasia.com/2026/04/25/united-states-ai-coding-app-cursor-creator-anysphere-to-raise-2-billion-new-funding-at-50-billion-valuation-raised-2-3-billion-in-series-d-funding-at-29-3-billion-valuation-in-2025-november-foun/; https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex/; https://marketbusinessnews.com/spacex-completes-all-stock-cursor-acquisition-at-60-billion-implied-value/450599/
17. COGNITION — SAN FRANCISCO, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; New York presence reported (secondary).
Website: cognition.ai
Contact: No published email or phone; official contact/sales route via cognition.ai
Published email: Not published
Phone: Not published
Founders: Scott Wu; Steven Hao; Walden Yan
Leadership: CEO / head: Scott Wu, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Steven Hao, Co-founder and Chief Technology Officer; Walden Yan, Co-founder and Chief Product Officer (titles per Wikipedia, Oct 2026); Neal Wu (IOI gold medalist, brother of Scott Wu); Gennady Korotkevich (IOI gold medalist, competitive-programming legend); Andrew He (IOI gold medalist); Eric Lu (engineer; announced Sep 2026 factorization of RSA-260, the largest publicly factored RSA challenge number, using a GPU implementation built and operated with Devin); Board / advisers: Not published; Founders: Scott Wu; Steven Hao; Walden Yan
Ownership status: Independent (private)
Ownership and funding: Private; $21M early 2024 at $350M valuation (Founders Fund); $175M Apr 2024 at $2B (Founders Fund led) — unicorn; Mar 2025 round at $4B led by 8VC (Joe Lonsdale); ~$400M Sep 2025 at $10.2B post-money led by Founders Fund with Lux Capital, 8VC, Bain Capital Ventures, D1; $1B+ Series D announced May 27, 2026 at $25B pre / $26B post-money, led by Lux Capital, General Catalyst, 8VC with Ribbit Capital, Atreides, Layer Global (company blog + press); $2B Series E announced Sep 8, 2026 at $48B valuation led by Andreessen Horowitz and Accel with Founders Fund, General Catalyst, Avenir (TechCrunch/finsmes); ARR ~$900M Sep 2026 (from $492M May 2026); investors incl. Founders Fund, 8VC, Lux Capital, General Catalyst, Andreessen Horowitz, Accel, Elad Gil, Ribbit, Atreides
What it builds: Devin (autonomous AI software engineer, demoed Mar 2024); Devin Desktop (agentic IDE; rebranded from Windsurf Jun 2026); Devin Oncall (Oct 15, 2026; auto-investigates production alerts and bugs)
Specialties: Autonomous software engineering; AI software engineers; agentic IDEs; multi-step coding-task automation; enterprise code generation, testing, deployment
Clients: Named in press/company reporting: Mercedes-Benz, NASA, Goldman Sachs, Santander, Citi; Goldman Sachs, Palantir, Cisco and Mercado Libre also named in secondary profiles.
What makes it distinctive: Demoed the first 'AI software engineer' (Devin) in Mar 2024 — an agent that plans, writes, tests, and ships code inside a team's codebase. Acquired the remaining Windsurf assets in Jul 2025 after Google's ~$2.4B acquihire of Windsurf's CEO and research leads, pairing an autonomous agent (Devin) with a human-in-the-loop IDE surface. Reports ~90% of its own codebase is written by Devin.
Notes: Website verified Oct 22, 2026: cognition.ai (company site; product at devin.ai). Founded November 2023 per Wikipedia (some sources say Aug/late 2023 — conflict flagged). ~200 employees (2026, Wikipedia). No published chief scientist/head of research, advisors, or contact details. Named customers include Mercedes-Benz, NASA, Goldman Sachs, Santander, Citi (press/company). RSA-260 factorization (Sep 2026) per Wikipedia — verify via primary source before quoting.
Sources: https://en.wikipedia.org/wiki/Cognition_AI; https://www.finsmes.com/2026/09/cognition-closes-usd2-billion-series-e-funding-round-at-usd48-billion-valuation.html; https://www.thebrief.news/en/standard/article/18165/cognition-raises-2-billion-at-48-billion-valuation-doubling-revenue-in-four-months; https://github.com/pedro-bright/the-ledger/blob/HEAD/content/events/2026/54-cognition-1b-round-26b.md; https://github.com/skryl/deep-research/blob/HEAD/content/research/2026/atreides/companies/cognition.md; https://goldsea.com/article_details/cognition-explodes-out-of-the-agentic-ai-startup-pack https://en.wikipedia.org/wiki/Cognition_AI
18. REPLIT — SAN FRANCISCO, UNITED STATES
Founded 2016 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: replit.com
Contact: No published email or phone; official contact/sales route via replit.com
Published email: Not published
Phone: Not published
Founders: Amjad Masad; Faris Masad; Haya Odeh
Leadership: CEO / head: Amjad Masad, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Amjad Masad; Faris Masad; Haya Odeh
Ownership status: Independent (private)
Ownership and funding: Private; ~$97.4M Series B extension Apr 2023 at $1.16B post-money led by Andreessen Horowitz (Khosla Ventures, Coatue, SV Angel, Y Combinator, Bloomberg Beta, Naval Ravikant, ARK Ventures, Hamilton Helmer) — total raised >$200M at that point; ~$250M Sep 2025 at ~$3B valuation led by Prysm Capital with Google AI Futures Fund, Amex Ventures, YC, a16z, Coatue (secondary reporting); $400M Series D Mar 11, 2026 at $9B valuation led by Georgian Partners with G Squared, Prysm Capital, Coatue, a16z, Craft Ventures, Y Combinator, Accenture Ventures, Okta Ventures, Databricks Ventures, Qatar Investment Authority, 1789 Capital, angels Shaquille O'Neal and Jared Leto (TechCrunch); ~$878M total raised; targeting $1B ARR by end of 2026 (company told Forbes)
What it builds: Replit (browser-based collaborative IDE); Replit Agent (launched Sep 2024); Agent 3; Agent 4 (launched Mar 11, 2026, with Design Canvas and parallel subagents)
Specialties: Cloud IDEs; 'vibe coding' / prompt-to-app platforms; autonomous app building, testing, and deployment; collaborative coding; AI agents for non-programmers
Clients: Not disclosed in sources checked
What makes it distinctive: Pioneered browser-based collaborative coding since 2016 and popularized 'vibe coding' — turning natural-language prompts into deployed full-stack apps. Reports 50M+ registered users, 150K+ paying customers, and employees from 85% of Fortune 500 companies on the platform (company disclosures at Series D). Microsoft partnership (Jul 2025) brought Replit to the Azure Marketplace.
Notes: Website verified Oct 22, 2026: replit.com (multiple independent sources). Founders verified via Wikipedia and TechCrunch (Amjad Masad, Faris Masad, Haya Odeh — wife of Amjad per freethink). Incorporated in San Mateo; based in San Francisco. No published chief scientist/head of research, key engineers, advisors, or contact details — cells blank. Jul 2025 incident: Replit's AI agent deleted a client company's database during a code freeze (Wikipedia citing press) — flagged for context. Jul 2026 billing glitch and pricing criticism reported (secondary).
Sources: https://en.wikipedia.org/wiki/Replit; https://techcrunch.com/2023/04/27/replit-funding-100m-generative-ai/; https://techcrunch.com/2026/03/11/replit-snags-9b-valuation-6-months-after-hitting-3b/; https://en.walaw.press/articles/replit_raises_400_million_dollars_triples_valuation_to_9_billion/GQPSQGRRPGGL; https://www.fwdstart.me/p/jordanian-founder-amjad-masads-replit-raises-250m-at-3b-valuation-launches-agent-3; https://www.freethink.com/entrepreneurship-innovation/software-creators
19. POOLSIDE — SAN FRANCISCO, UNITED STATES
Founded 2023 · Independent (private); Nvidia holds a strategic investment and a non-exclusive licence (Aug 2026) — not an acquisition
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco (548 Market Street per CB Insights); Paris branch/engineering hub; team also in the United Kingdom and France (remote-first origins).
Website: https://poolside.ai
Contact: Published email: press@poolside.ai; also via https://poolside.ai
Published email: press@poolside.ai
Phone: Not published
Founders: Jason Warner; Eiso Kant
Leadership: CEO / head: Jason Warner, President and Co-Founder (Wikipedia key-people listing; also reported as CEO — see notes); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Eiso Kant, Chief Technology Officer and Co-Founder; Board / advisers: Not published; Founders: Jason Warner; Eiso Kant
Ownership status: Independent (private); Nvidia holds a strategic investment and a non-exclusive licence (Aug 2026) — not an acquisition
Ownership and funding: Private; $26M seed + $100M Series A within first year; $500M Series B Oct 2024 at $3B valuation led by Bain Capital with DST Global and eBay (press); failed $2B Series C at $14B spring 2026 after CoreWeave withdrew from joint Texas data center (secondary reporting); Aug 2026: Nvidia $1B strategic investment at $12B pre-money (~$13B post) per 7gc announcement + Nvidia $6B NON-EXCLUSIVE license of Poolside's 'Model Factory' methodology/tooling (non-equity; reportedly to be distributed to existing investors as partial liquidity); Nvidia extended offers to 109 employees (~half headcount); founders staying, company operating independently; total equity ~$1.626B (seed $26M + Series A $100M + Series B $500M + Nvidia $1B at $12B pre-money per ai-market-watch)
What it builds: Malibu (multi-file code generation and refactoring model); Point (sub-200ms code completion model); Laguna S 2.1 (open-weight coding model, 118B parameters); Laguna XS; Model Factory (internal model-building platform); API and agentic tools
Specialties: Foundation models for software engineering; open-weight coding models; self-hosted/on-premises AI for enterprise and government; architecture search and code-execution-based reinforcement learning; AI for highly regulated organizations
Clients: Client categories: enterprise software organizations and government/defence customers needing on-premises deployment; AWS is the first cloud provider to offer its models fully managed (company release). Named clients not disclosed.
What makes it distinctive: Builds coding models optimized to run on customers' own hardware with on-premises deployment for data-sensitive government, defense, and regulated enterprises. The Aug 2026 Nvidia arrangement — $1B equity plus a $6B non-exclusive Model Factory license with ~half the staff offered Nvidia roles — is a novel 'third path' between acquisition and funding, functioning as partial liquidity without a sale.
Notes: HQ CONFLICT: Wikipedia lists San Francisco, US; Forge lists Paris, France; history sources say it started work-from-home (US/United Kingdom/France) and moved to France after the seed round, then San Francisco with a Paris branch — HQ city/country left blank with conflict flagged. CEO TITLE CONFLICT: ai-market-watch says Jason Warner leads as CEO; Wikipedia infobox says 'President and Co-Founder' — used 'CEO' with flag. Employee count conflicts: ~256 (thecodew, 2025) vs 150 (Wikipedia, Dec 2025). Project Horizon 2-gigawatt AI campus in Pecos County, Texas broke ground Oct 2025. No published chief scientist, advisors, or contact details. HQ RESOLVED (Oct 22, 2026 refresh): listed as San Francisco, United States on the weight of Wikipedia (HQ San Francisco; Paris branch; employees in the US, United Kingdom and France) and CB Insights (HQ 548 Market Street, San Francisco); the company's own AWS release says only 'headquartered in the U.S.' Forge's Paris listing reflects the Paris branch/engineering hub. Project Horizon: the 2-gigawatt Pecos County, Texas campus was reported collapsed in 2026 secondary reporting after CoreWeave's withdrawal (Tech Insider, Oct 2026) — not confirmed by the company; flagged. Contact email press@poolside.ai is published in the company's own poolside/AWS press release (PR Newswire).
Sources: https://en.wikipedia.org/wiki/Poolside_AI; https://Forgeglobal.com/poolside-ai_ipo/; https://www.7gc.co/news/poolside-strikes-6b-deal-with-nvidia; https://www.thecodew.com/2026/08/nvidia-strikes-6B-licensing-deal-poolside.html; https://www.ai-market-watch.com/company/poolside-ai; https://www.ainvest.com/news/ai-startup-poolside-is-seeking-a-3-billion-valuation-in-a-500-million-funding-round-24091010d9de5141bfd2b9e9/ https://www.nasdaq.com/press-release/poolside-and-aws-announce-strategic-agreement-enable-secure-customized-generative-ai; https://www.cbinsights.com/company/poolside
20. LOVABLE — STOCKHOLM, SWEDEN
Founded 2023 · Independent (private)
Headquarters: Stockholm, Sweden
Offices and footprint: HQ Stockholm; remote-first European team — additional named offices not published in sources checked.
Website: lovable.dev
Contact: No published email or phone; official contact/sales route via lovable.dev
Published email: Not published
Phone: Not published
Founders: Anton Osika; Fabian Hedin
Leadership: CEO / head: Anton Osika, Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Fabian Hedin, Chief Technology Officer; Board / advisers: Not published; Founders: Anton Osika; Fabian Hedin
Ownership status: Independent (private)
Ownership and funding: Private; pre-seed ~EUR 6.8M (~$7.5M) late 2023 led by Hummingbird and byFounders (secondary); ~$15M Feb 2025 led by Creandum (secondary); $200M Series A Jul 2025 at $1.8B valuation led by Accel (press); $330M Series B Dec 2025 at $6.6B valuation led by CapitalG and Menlo Ventures with Khosla Ventures, Salesforce Ventures, Databricks Ventures, NVentures, Atlassian Ventures, T.Capital (Reuters); $400M Series C Aug 2026 at $13.3B valuation led by Menlo Ventures and the Scaleup Europe Fund (secondary/press); ~$552.5M+ total raised; ARR $400M Feb 2026 (company via CCO Ryan Meadows to Business Insider), ~$500M Jun 2026 (secondary)
What it builds: Lovable (prompt-to-app platform); Lovable Agent; Lovable Cloud
Specialties: 'Vibe coding' / natural-language app building; full-stack development from prompts; non-technical founders and teams; Supabase-backed app generation
Clients: Named in secondary dossiers: Klarna, Deutsche Telekom, Delivery Hero, eXp Realty, Zendesk, Uber AI, Sentry, n8n, ElevenLabs, AppDirect (secondary source; not a company-published list).
What makes it distinctive: One of the fastest-scaling startups on record: $0 to $400M ARR in ~15 months with just 146 employees (~$2.7M ARR per employee, Feb 2026, per press). Grew from Anton Osika's open-source GPT Engineer (50K+ GitHub stars); relaunched as Lovable Nov/Dec 2024. Reported 60M projects hosted and 900M monthly visitor visits as of Aug 2026 (company).
Notes: Website verified Oct 22, 2026: lovable.dev (Wikipedia). Legal name Lovable Labs Incorporated; ~120 employees (2025, Wikipedia) / 146 employees (Feb 2026, press) — figures vary, flagged. Founded Nov 2023 (GPT Engineer origin; commercial app relaunched as 'Lovable' Nov/Dec 2024 — some sources say Dec 2024 rebrand). Osika and Hedin are ex-DeepMind researchers (secondary); both pledged 50% of eventual exit earnings to charity (Forbes mirror / Founders Pledge). No published chief scientist/head of research, advisors, or contact details. Mar 2025 security issue: Lovable-created sites exposing Supabase databases via misconfigured access controls (Wikipedia) — flagged for context.
Sources: https://en.wikipedia.org/wiki/Lovable_(company); https://www.reuters.com/business/finance/lovable-valued-66-billion-latest-funding-round-ai-coding-demand-surges-2025-12-18/; https://en.walaw.press/articles/swedish_startup_lovable_reaches_$400_million_revenue_with_146_employees/GQPSSXRPRFPS; https://www.webpronews.com/lovables-13-3-billion-leap-how-vibe-coding-turned-prompts-into-a-european-tech-powerhouse/; https://github.com/gaurav-product/product-management-case-studies/blob/HEAD/Case%20Studies/Day-19-Lovable/README.md; https://forbestodays.pages.dev/profile/anton-osika/
SEARCH, KNOWLEDGE, PROFESSIONAL SERVICES AND VOICE
21. PERPLEXITY — SAN FRANCISCO, UNITED STATES
Founded 2022 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: perplexity.ai
Contact: No published email or phone; official contact/sales route via perplexity.ai
Published email: Not published
Phone: Not published
Founders: Aravind Srinivas; Denis Yarats; Johnny Ho; Andy Konwinski
Leadership: CEO / head: Aravind Srinivas — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Denis Yarats — Co-Founder & CTO (ex-Meta AI research scientist); Johnny Ho — Co-Founder & CSO (ex-Quora, ex-Google); Andy Konwinski — Co-Founder (founding engineer at Databricks, co-creator of Apache Spark); Board / advisers: Health Advisory Board announced Mar 20, 2026 (medical/scientific experts; individual members not published); Founders: Aravind Srinivas; Denis Yarats; Johnny Ho; Andy Konwinski
Ownership status: Independent (private)
Ownership and funding: Private; ~$20–21B valuation at Series E-6, early 2026 (press estimate); reported equity round at $30B+ in talks with Nvidia, Aug 2026 (press, unconfirmed); ~$1.5–1.65B total raised (secondary databases); investors reported: IVP, NEA, Databricks Ventures, Nvidia, SoftBank Vision Fund 2, Bezos Expeditions, Sequoia
What it builds: Perplexity answer engine (Free, Pro $20/mo, Max $200/mo, Enterprise tiers); Deep Research; Comet agentic AI browser (Mac, Windows, iOS, Android); Perplexity Health; developer API
Specialties: AI search; answer engines; agentic web browsing; deep research; healthcare AI; developer APIs
Clients: Not disclosed in sources checked
What makes it distinctive: Pioneered the cited conversational answer engine as an alternative to link-list search. Runs model-agnostic, routing research across third-party frontier models rather than training its own flagship LLM. Expanded from search into an agentic browser (Comet, free worldwide since Oct 2025, iOS Mar 2026, Android Aug 2026) and vertical products such as Perplexity Health.
Notes: Valuation figures conflict across secondary sources ($20–21B early 2026 vs $22.6B Dec 2025 vs $30B+ in talks Aug 2026); one secondary database claims $45B valuation/$4.5B raised — treated as unverified outlier. Nvidia equity-investment talks at $30B+ valuation (The Information via Reuters, Aug 2026, unconfirmed) — no deal announced as of Oct 20, 2026. Dealroom (Oct 7, 2026, unnamed sources) claims Nvidia and Perplexity settled on a partnership with a ~$3B Nvidia investment at ~$35B pre-money valuation — not confirmed by either company; The Information's Aug 2026 report of $30B+ valuation talks remains the only widely repeated account. Annualized revenue crossed $750M (The Information, reported Oct 2026), up from <$250M at start of 2026 — unconfirmed by company. CEO Aravind Srinivas said company plans IPO in 2028 (CNBC, Jun 2026). No chief scientist published. No public contact email/phone found.
Sources: https://www.intelligentliving.co/perplexity-ai-valuation-hits-30b-nvidias/; https://siliconvalleyinvestclub.com/companies/perplexity/; https://tech-insider.org/comet-vs-gemini-agent-vs-chatgpt-atlas-2026/; https://pulse2.com/perplexity-comet-for-ios-launch-and-perplexity-health-debut-expand-ai-into-browsing-and-healthcare/; https://en.wikipedia.org/wiki/Denis_Yarats; https://startupfortune.com/nvidia-is-in-talks-to-value-perplexity-at-more-than-30-billion/
22. GLEAN — PALO ALTO, UNITED STATES
Founded 2019 · Independent (private)
Headquarters: Palo Alto, United States
Offices and footprint: HQ Palo Alto; Nashville, Tennessee corporate sales office announced Apr 2026.
Website: glean.com
Contact: No published email or phone; official contact/sales route via glean.com
Published email: Not published
Phone: Not published
Founders: Arvind Jain; Tony Gentilcore; Vishwanath T R; Piyush Prahladka
Leadership: CEO / head: Arvind Jain — Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Piyush Prahladka — Co-Founder & Head of Search (ex-Uber engineering lead); Tony Gentilcore — Co-Founder & Product Engineering Lead (ex-Google senior staff engineer); T.R. Vishwanath — Co-Founder (ex-Facebook principal software engineer); Board / advisers: Not published; Founders: Arvind Jain; Tony Gentilcore; Vishwanath T R; Piyush Prahladka
Ownership status: Independent (private)
Ownership and funding: Private; Series F $150M at $7.2B valuation, Jun 2025 (company announcement, Reuters/Crunchbase); Series E $260M at $4.6B, Sep 2024 (Crunchbase); ~$770M total raised (Crunchbase, secondary); investors: Wellington Management (lead, Series F), Khosla Ventures, Sequoia, General Catalyst, Lightspeed Venture Partners, Kleiner Perkins, IVP, Altimeter, Coatue, DST Global, ICONIQ, Sapphire Ventures, Capital One Ventures, Citi
What it builds: Glean Assistant; Glean Agents (horizontal agent platform, GA May 2025); Enterprise Graph
Specialties: enterprise search; enterprise knowledge management; AI agents for knowledge work; permissions-aware retrieval
Clients: Named by the company: Booking.com, Grammarly, Duolingo and Databricks (100M+ agent actions annually across customers).
What makes it distinctive: Built on an enterprise knowledge graph with permissions enforcement, so answers respect existing data-access controls. Glean Agents powers 100M+ agent actions annually for customers including Booking.com, Grammarly, Duolingo and Databricks. Founded by ex-Google search engineers; surpassed $100M ARR in its last fiscal year, less than three years after launch.
Notes: Founding year reported as 2019 (Wikipedia, Reuters, Crunchbase) vs 2020 (Bounce Watch) — flagged. Nashville corporate sales office announced Apr 2026. No chief scientist published. No public contact email/phone found.
Sources: https://news.crunchbase.com/venture/ai-powered-work-assistant-glean-valuation-jumps/; https://en.wikipedia.org/wiki/Glean_Technologies; https://www.boringbusinessnerd.com/startups/glean; https://jackfmfargo.com/2025/06/10/ai-startup-glean-valuation-surges-to-7-2-billion-in-latest-fundraise/
23. HARVEY — SAN FRANCISCO, UNITED STATES
Founded 2022 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: harvey.ai
Contact: No published email or phone; official contact/sales route via harvey.ai
Published email: Not published
Phone: Not published
Founders: Winston Weinberg; Gabriel Pereyra
Leadership: CEO / head: Winston Weinberg — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Gabe Pereyra — Co-Founder & President (ex-Google DeepMind and Meta AI research scientist); Siva Gurumurthy — CTO; Rachel Hepworth — Chief Marketing Officer; Board / advisers: Not published; Founders: Winston Weinberg; Gabriel Pereyra
Ownership status: Independent (private)
Ownership and funding: Private; $550M round at $15.5B valuation, Sep 9 2026 (company announcement; Bloomberg reported $15.6B — flagged); $200M growth round at $11B, Mar 25 2026 (company announcement); >$1.55B total raised (company, Sep 2026); investors: Diffusion + Lightspeed Venture Partners (co-lead, Sep 2026), Sequoia, Kleiner Perkins, a16z, Coatue, GIC, Sapphire Ventures, Whale Rock, Goldman Sachs Alternatives, Verified Capital, WndrCo
What it builds: Harvey Assistant; Vault; custom AI agents (25,000+ deployed); Harvey LAB (Legal Agent Benchmark); Harvey Tenet (first in-house post-trained model, built on Moonshot AI Kimi K3 open weights)
Specialties: legal AI; contract analysis; due diligence; legal research; compliance; litigation support
Clients: Client category: law firms — serves 80% of the Am Law 100 and 100,000+ lawyers across 60+ countries (company figures).
What makes it distinctive: Domain-specific legal AI fine-tuned on case law, firm playbooks and compliance manuals rather than a general chatbot. Launched its first in-house model (Tenet, post-trained on open-weight Kimi K3) to keep privileged client documents off third-party servers. Serves 80% of Am Law 100 firms and 100,000+ lawyers across 60+ countries.
Notes: Bloomberg put the Sep 2026 round at $15.6B vs the company/Reuters figure of $15.5B — flagged. Acquired Guardrails AI in 2026 (its fourth acquisition of the year). No chief scientist published. No public contact email/phone found.
Sources: https://fortuneherald.com/breaking/harveys-550m-funding-round-vaults-legal-ai-firm-to-15-5b-valuation/; https://blockchain.news/news/harvey-550m-funding-valuation-legal-ai; https://www.todaysstartupnews.com/startups/harvey-550-million-15-6-billion-valuation-own-models; https://tech-insider.org/legora-550-million-series-d-5-5-billion-valuation-harvey-ai-2026/
24. HEBBIA — NEW YORK, UNITED STATES
Founded 2020 · Independent (private)
Headquarters: New York, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: hebbia.com
Contact: No published email or phone; official contact/sales route via hebbia.com
Published email: Not published
Phone: Not published
Founders: George Sivulka
Leadership: CEO / head: George Sivulka — Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Michelangelo Volpi — Board Member (per Forge Global); Founders: George Sivulka
Ownership status: Independent (private)
Ownership and funding: Private; Series B $130M at ~$700M valuation, Jul 2024 (TechCrunch/Bloomberg via press); ~$161M total raised (secondary databases); Series A $30M, Jun 2022 (Radical Ventures); investors: Andreessen Horowitz (lead, Series B), Index Ventures, GV, Peter Thiel, Radical Ventures, Abstract Ventures, Floodgate, Human Ventures
What it builds: Matrix (AI research and agent platform with spreadsheet-style interface)
Specialties: document research; financial AI; investment banking and private equity workflows; AI agents for finance; citation-backed analysis
Clients: Not disclosed in sources checked
What makes it distinctive: Matrix lets analysts load hundreds or thousands of documents and run the same questions across all of them at once, with answers linked to exact passages, in a finance-native spreadsheet interface. Used by investment banks and over 40% of the largest asset managers, plus law firms and the U.S. Air Force. Integrated with Microsoft Azure AI Foundry (Aug 2025).
Notes: Founding year reported as 2020 (most sources) vs 2021 (one profile) — flagged. Forge Global lists total funding at $249.71M, conflicting with the widely reported $161M — flagged. Offices: New York HQ, San Francisco, London (as of Aug 2026). No chief scientist or other named technical leadership published beyond the founder; no public email/phone found.
Sources: https://forgeglobal.com/hebbia_stock/; https://bouncewatch.com/company/hebbia-ai; https://www.datastudios.org/post/hebbia-ai-document-research-for-finance-matrix-workflows-and-pricing; https://www.maginative.com/article/hebbia-lands-130m-to-supercharge-knowledge-work-with-ai/
25. SIERRA — SAN FRANCISCO, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; Tokyo and Paris teams added through 2026 acquisitions (Opera Tech, Fragment) per reporting.
Website: sierra.ai
Contact: No published email or phone; official contact/sales route via sierra.ai
Published email: Not published
Phone: Not published
Founders: Bret Taylor; Clay Bavor
Leadership: CEO / head: Bret Taylor — Co-Founder & CEO; Clay Bavor — Co-Founder (exact title not published); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Bret Taylor; Clay Bavor
Ownership status: Independent (private)
Ownership and funding: Private; Series E $950M at $15.8B valuation, May 4 2026 (company announcement); Series B $350M at $10B, Sep 2025 (press); ~$1.6B total raised (press estimate); investors: Tiger Global + GV (co-lead, Series E), Benchmark, Sequoia, Greenoaks, ICONIQ, Thrive Capital
What it builds: Agent OS; Agent Studio; Agent SDK; Ghostwriter (agent-that-builds-agents, launched Mar 2026); Explorer
Specialties: customer experience AI; conversational AI agents; customer support automation; voice agents; enterprise CX
Clients: Client category: large consumer enterprises; Sacra reports Sierra serves 40% of the Fortune 50 (company-category figure; individual clients not fully disclosed).
What makes it distinctive: Outcome-based pricing charges per successful resolution rather than per seat, breaking the subscription SaaS model. Serves ~40% of the Fortune 50 with production deployments, reaching ~$200M ARR within ~27 months of launch — among the fastest ramps in enterprise software. Multi-channel agents across chat, voice, email, SMS and WhatsApp.
Notes: Founded 2023 per company narrative (emerged from stealth Feb 2024) vs 2024 (Wikipedia) — flagged. HQ conflict: San Francisco-based (TechCrunch, multiple press) vs headquartered in New York (one data profile) — flagged. Bret Taylor also chairs the OpenAI board. Clay Bavor's exact title not published. No chief scientist published; no public email/phone found. Acquisitions in 2026 (reported by TechCrunch/Crunchbase News; terms undisclosed): Opera Tech (Tokyo enterprise AI, Mar 2026), Receptive AI (voice agents, Mar 2026), Fragment (Paris YC startup, Apr 23, 2026 — co-founders Olivier Moindrot and Guillaume Genthial joined Sierra), TakeOff ('long-horizon' agents, Jul 2026). Sacra reports Sierra serves 40% of the Fortune 50.
Sources: https://techcrunch.com/2025/11/21/bret-taylors-sierra-reaches-100m-arr-in-under-two-years/; https://valueaddvc.com/blog/sierra-ai-valuation-2026-15-8b-series-e-enterprise-ai-agents; https://www.libertyglobal.com/liberty-global-signs-strategic-partnership-sierra/; https://aiagentsquare.com/agents/sierra-ai
26. WRITER — SAN FRANCISCO, UNITED STATES
Founded 2020 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: writer.com
Contact: No published email or phone; official contact/sales route via writer.com
Published email: Not published
Phone: Not published
Founders: May Habib; Waseem AlShikh
Leadership: CEO / head: May Habib — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Waseem AlShikh — Co-Founder & CTO; Board / advisers: Sandesh Patnam (Managing Partner, Premji Invest) and Rob Toews (Partner, Radical Ventures) joined the board with the Nov 2024 Series C; Founders: May Habib; Waseem AlShikh
Ownership status: Independent (private)
Ownership and funding: Private; Series C $200M at $1.9B valuation, Nov 2024 (company press release via Business Wire); ~$320–326M total raised (Forbes/secondary profiles); investors: Premji Invest, Radical Ventures, ICONIQ Growth (co-leads), Salesforce Ventures, Adobe Ventures, IBM Ventures, Workday Ventures, Citi Ventures, Accenture, Balderton, Insight Partners, Vanguard
What it builds: AI Studio; WRITER Agent; Palmyra X6 (flagship LLM, released Aug 13, 2026, post-trained on Z.ai GLM 5.2); Palmyra X5 (Apr 2025); AI HQ platform
Specialties: enterprise content generation; brand voice consistency; marketing and revenue workflows; AI agents for regulated industries; graph-based RAG
Clients: Not disclosed in sources checked
What makes it distinctive: Builds its own enterprise LLMs (Palmyra family) optimized for cost-efficiency instead of licensing foundation models — Palmyra X6 was post-trained on open weights at a claimed fraction of frontier training cost. Targets regulated-enterprise marketing and revenue teams with governance and brand-voice controls. 250+ enterprise clients including Salesforce, Uber, L'Oreal, Accenture and Vanguard.
Notes: Total raised reported as $320M (Forbes) vs $326M (YesPress) — flagged. Evolved from Qordoba machine-translation roots. No chief scientist published; no public email/phone found.
Sources: https://www.businesswire.com/news/home/20260813170187/en/WRITER-Makes-Agentic-AI-Economically-Sustainable-at-Enterprise-Scale-With-Palmyra-X6-Release-and-Major-Harness-Upgrades; https://theaiinsider.tech/2024/11/12/writer-raises-200m-series-c-at-1-9b-valuation-to-fuel-leadership-in-agentic-enterprise-ai/; https://www.forbesindia.com/amp/article/cross-border/working-ai-may-habibs-writer-sells-a-new-way-of-doing-things/96012/1; https://siliconangle.com/2026/08/13/writer-launches-major-agentic-ai-improvements-palmyra-x6-flagship-model/
27. ELEVENLABS — LONDON, UNITED KINGDOM
Founded 2022 · Independent (private)
Headquarters: London, United Kingdom
Offices and footprint: HQ London; India presence marked by its first India Summit in Bengaluru (Oct 2026); physical-presence expansion in Tokyo, Seoul and San Francisco reported as planned (Feb 2026). Full official office list not published in sources checked.
Website: elevenlabs.io
Contact: No published email or phone; official contact/sales route via elevenlabs.io
Published email: Not published
Phone: Not published
Founders: Mati Staniszewski; Piotr Dabkowski
Leadership: CEO / head: Mati Staniszewski — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Piotr Dabkowski — Co-Founder & CTO; Board / advisers: Andrew Reed (Partner, Sequoia Capital) joined the board with the Feb 2026 Series D; Founders: Mati Staniszewski; Piotr Dabkowski
Ownership status: Independent (private)
Ownership and funding: Private; Series D $500M at $11B valuation, Feb 4 2026 (company announcement, Reuters; led by Sequoia); $300M employee tender offer at $22B valuation, Sep 30 2026 (company announcement, Reuters — liquidity event, not new capital; led by Wellington and T. Rowe Price); ~$781M primary capital raised (sum of disclosed rounds via secondary database); investors: Sequoia, a16z, ICONIQ Growth, Lightspeed, Evantic, Nat Friedman, Daniel Gross, Wellington, T. Rowe Price, EQT, Goldman Sachs; Nvidia took a stake Sep 2026
What it builds: Eleven v4 and v4 Turbo (text-to-speech models, launched Sep 28, 2026); ElevenAgents (voice/chat agent platform); ElevenAgents Architect (alpha, Oct 20, 2026; build/modify agents via natural-language commands); ElevenCreative; ElevenAPI; ElevenMusic; Scribe v2 (speech-to-text); Dubbing; Voice Cloning
Specialties: voice synthesis; text-to-speech; speech-to-text; voice cloning; dubbing; conversational voice agents; AI music generation
Clients: Named in reporting: Deutsche Telekom, Square, Revolut (enterprise deployments, secondary); CARS24 and Razorpay (India, Oct 2026, reported); 250+ businesses in India (company, Oct 2026).
What makes it distinctive: Vertically integrated audio stack — trains its own expressive speech models rather than wrapping third-party APIs; Eleven v4 ranked No. 1 on Artificial Analysis's voice leaderboard (Sep 2026). Models speak, listen and translate in 90+ languages used by 5.5B people. ElevenAgents handles 15M+ live conversations per week for enterprises including Deutsche Telekom, Stripe and sovereign governments.
Notes: The Sep 30, 2026 $300M tender doubled the valuation to $22B without raising new capital — flagged as liquidity event, not a priced round. Nvidia took a stake in Sep 2026 (per WSJ via company). No chief scientist published; no public email/phone found. Multi-year licensing + strategic agreement with Universal Music Group announced Sep 10, 2026 (company announcement; ElevenLabs' first major-label deal): a licensed AI remix/mashup platform for UMG's catalog plus joint AI audio products for artists and songwriters. India is now the company's second-largest market after the United States (company, Oct 6, 2026 India Summit, Bengaluru; reported): 250+ Indian businesses using the technology, 100M ElevenAgents conversations in India in the past year, deployments at CARS24 (400+ voice agents) and Razorpay reported.
Sources: https://www.reuters.com/legal/transactional/elevenlabs-valuation-doubles-22-billion-surging-ai-voice-agent-demand-2026-09-30/; https://srnnews.com/elevenlabs-raises-500-million-at-11-billion-valuation/; https://www.finsmes.com/2026/10/elevenlabs-closes-usd300m-tender-offer-valued-at-usd22-billion.html; https://runtimewire.com/article/elevenlabs-eleven-v4-turbo-launch
28. SUNO — CAMBRIDGE, UNITED STATES
Founded 2022 · Independent (private)
Headquarters: Cambridge, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: suno.com
Contact: No published email or phone; official contact/sales route via suno.com
Published email: Not published
Phone: Not published
Founders: Mikey Shulman; Georg Kucsko; Martin Camacho; Keenan Freyberg
Leadership: CEO / head: Mikey Shulman — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Georg Kucsko — Co-Founder & CTO; Martin Camacho — Co-Founder & President; Christian Steinmetz — research scientist (AP profile, Feb 2026); Keenan Freyberg — Co-Founder (title not published); Board / advisers: Not published; Founders: Mikey Shulman; Georg Kucsko; Martin Camacho; Keenan Freyberg
Ownership status: Independent (private)
Ownership and funding: Private; Series D $400M at $5.4B valuation, Jun 3 2026 (company announcement; led by Bond Capital); Series C $250M at $2.45B, Nov 2025 (Reuters/company blog; led by Menlo Ventures); Series B $125M, May 2024 (company announcement; led by Lightspeed); $775M total disclosed (sum of rounds); investors: Bond Capital, Menlo Ventures, Lightspeed, Matrix Partners, IVP, Forerunner, Union Square Ventures, Alkeon, Quiet, Schroders Capital
What it builds: Suno generative music platform (v6 model suite incl. v6-mini; web and mobile app); studio/production workspace; stem extraction; audio watermarking and fingerprinting (Aug 2026)
Specialties: music generation; audio generation; songwriting AI; creative tools; AI music licensing
Clients: Not disclosed in sources checked
What makes it distinctive: Generates complete songs — vocals, instruments, production — from text prompts at ~7M songs/day with ~2M paid subscribers and ~$300M ARR (Feb 2026, company figures). Settled with Warner Music Group (Nov 2025) and entered a licensing partnership — the first deal of its kind between a major label and an AI music platform. Rolled out audio watermarking and provenance tools (Aug 2026).
Notes: Founded 2022 (CEO, company) vs 2024 (one press piece) — flagged. UMG and Sony copyright lawsuits ongoing (amended complaint alleges 61,000+ songs); company says v6 models were trained without UMG/Sony music. No chief scientist published; no public email/phone found.
Sources: http://www.musictimes.com/articles/112108/20260605/suno-raises-400-million-54-billion-valuation-ai-musics-biggest-bet-yet.htm; https://finance.biggo.com/news/9sS0jp4B-PfaobXfVeD8; https://techxplore.com/news/2026-02-ai-song-generator-startups-suno.html; https://runtimewire.com/article/suno-watermarking-fingerprinting-download-limits; https://siliconvalleyinvestclub.com/companies/suno/tear-sheet.pdf
29. IDEOGRAM — TORONTO, CANADA
Founded 2022 · Independent (private)
Headquarters: Toronto, Canada
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: ideogram.ai
Contact: No published email or phone; official contact/sales route via ideogram.ai
Published email: Not published
Phone: Not published
Founders: Mohammad Norouzi; William Chan; Jonathan Ho; Chitwan Saharia; Shayaan Abdullah; Jenny Lei; Jacob Lu
Leadership: CEO / head: Mohammad Norouzi — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Mohammad Norouzi — Co-Founder & CEO (ex-Google Brain, co-creator of Google's Imagen); William Chan — founding team (ex-Google Brain); Jonathan Ho — founding team (ex-Google Brain); Chitwan Saharia — founding team (ex-Google Brain); Board / advisers: Not published; Founders: Mohammad Norouzi; William Chan; Jonathan Ho; Chitwan Saharia; Shayaan Abdullah; Jenny Lei; Jacob Lu
Ownership status: Independent (private)
Ownership and funding: Private; Series A $80M, Feb 2024 (company announcement via Bloomberg; valuation not disclosed); seed $16.5M, Aug 2023 (co-led by a16z and Index Ventures); ~$96.5–102M total (secondary databases); investors: Andreessen Horowitz, Index Ventures, Redpoint Ventures, Pear VC, SV Angel
What it builds: Ideogram text-to-image models (1.0 through 4.5; 4.5 adds multi-turn editing, Oct 2026); Magic Prompt; Canvas; iOS app; developer API
Specialties: image generation; text-to-image; text rendering inside images; graphic design; prompt adherence
Clients: Not disclosed in sources checked
What makes it distinctive: Built its identity on legible text rendering inside generated images — a problem competitors struggled with — aimed at design, typography and advertising use cases. Founded by the ex-Google Brain researchers behind Google's Imagen. Latest model 4.5 focuses on maintaining composition through multi-turn edits for advertising and ecommerce workflows.
Notes: Founded 2022 (company reports) vs 2023 (some databases) — flagged; first model released publicly Aug 2023. Valuation never disclosed. Individual co-founder titles beyond Norouzi not published. No chief scientist published; no public email/phone found.
Sources: https://news.bloomberglaw.com/artificial-intelligence/startup-ideogram-raises-80-million-for-ai-image-generation; https://venturebeat.com/ai/midjourney-rival-ideogram-gets-80m-in-series-a-led-by-andreessen-horowitz; https://dev.mpost.io/ideogram-ai-secures-16-5m-seed-round-to-expand-its-generative-ai-platform/; https://runtimewire.com/article/ideogram-45-multi-turn-image-editing; https://www.bouncewatch.com/explore/startup/ideogram
30. LUMA AI — PALO ALTO, UNITED STATES
Founded 2021 · Independent (private)
Headquarters: Palo Alto, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: lumalabs.ai
Contact: No published email or phone; official contact/sales route via lumalabs.ai
Published email: Not published
Phone: Not published
Founders: Amit Jain; Alex Yu; Alberto Taiuti
Leadership: CEO / head: Amit Jain — Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Alex Yu — Co-Founder & CTO; Alberto Taiuti — Co-Founder (title not published); Board / advisers: Not published; Founders: Amit Jain; Alex Yu; Alberto Taiuti
Ownership status: Independent (private)
Ownership and funding: Private; Series C $900M at $4B+ valuation, Nov 2025 (company announcement via CNBC); $1B+ total raised (press estimate, Jun 2026); investors: Humain (lead, Saudi PIF-backed), AMD Ventures, Andreessen Horowitz, Amplify Partners, Matrix Partners, Nvidia (earlier)
What it builds: Dream Machine (flagship video generator); Ray3 (reasoning video model, Sep 2025); content creation suite; developer API
Specialties: video generation; world models; multimodal AI; image generation; 3D content
Clients: Not disclosed in sources checked
What makes it distinctive: Pursues multimodal world models learning from video, audio, images and text to simulate physical environments, distinct from text-only LLM labs. Secured a 2-gigawatt Project Halo AI supercluster partnership with Saudi Arabia's Humain — among the largest planned GPU deployments. Started in 3D capture before pivoting to video generation.
Notes: Alberto Taiuti's title not published — flagged. London office expansion announced Nov 2025 (200 hires targeted by early 2027). No chief scientist published; no public email/phone found.
Sources: https://techstartups.com/2025/11/20/luma-ai-lands-900m-to-challenge-openai-and-google-with-frontier-multimodal-models-partners-with-humain-on-saudi-arabias-2gw-ai-megacluster/; https://morphed.app/stats/luma-dream-machine-statistics; https://therealpreneur.com/tech-today/nvidia-backed-luma-ai-expands-to-london-plans-major-hiring-push/; https://craft.co/luma-ai
IMAGE, VIDEO AND ROBOTICS
31. MIDJOURNEY — SAN FRANCISCO, UNITED STATES
Founded 2021 · Independent (private; bootstrapped — no venture funding taken)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: midjourney.com
Contact: No published email or phone; official contact/sales route via midjourney.com
Published email: Not published
Phone: Not published
Founders: David Holz
Leadership: CEO / head: David Holz - Founder and CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: David Holz
Ownership status: Independent (private; bootstrapped — no venture funding taken)
Ownership and funding: Private. Self-funded/bootstrapped - $0 external venture funding raised (company statements; Owler). Profitable since Aug 2022 (founder to The Register). Revenue ~$200M in 2023 (reported); ~$300M in 2024 (reported, Time profile). Meta licensed Midjourney's image/video technology Aug 2025 (reported). No published valuation.
What it builds: Midjourney V8.1 (Apr 2026); Midjourney V1 video model; Midjourney TV; Draft Mode
Specialties: image generation; text-to-image; video generation
Clients: Not disclosed in sources checked
What makes it distinctive: One of the only major generative-AI labs with zero outside capital, profitable since 2022 on consumer subscriptions alone. Revenue per employee (roughly $5M+ per head on a ~40-100 person team) dwarfs typical software peers. Built its audience through a public-by-default Discord community rather than enterprise sales.
Notes: No CTO, chief scientist, board members, or public contact email/phone published - deliberately opaque company. Employee-count estimates vary (40 in 2023 reporting vs ~100 Aug 2024 vs 60 on Wikipedia). In June 2026 announced a Midjourney Medical division and a 'Midjourney Scanner' full-body ultrasound concept with a planned 2027 San Francisco spa location - ambitious claim, reported via Wikipedia with limited independent verification.
Sources: https://en.wikipedia.org/wiki/Midjourney,_Inc.; https://www.webpronews.com/midjourneys-quiet-cash-machine-how-a-40-person-startup-blew-past-200-million-without-venture-capital/; https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/; https://www.owler.com/company/midjourney
32. STABILITY AI — LONDON, UNITED KINGDOM
Founded 2019 · Independent (private)
Headquarters: London, United Kingdom
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: stability.ai
Contact: No published email or phone; official contact/sales route via stability.ai
Published email: Not published
Phone: Not published
Founders: Emad Mostaque; Cyrus Hodes
Leadership: CEO / head: Prem Akkaraju - Chief Executive Officer (appointed June 25, 2024); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Christian Laforte - Chief Technology Officer; Shan Shan Wong - former interim co-CEO/Chief Operating Officer (Mar-Jun 2024); Board / advisers: Sean Parker - Executive Chairman of the Board; James Cameron - board member; Founders: Emad Mostaque; Cyrus Hodes
Ownership status: Independent (private)
Ownership and funding: Private. $101M Series A Oct 2022 (company announcement) at ~$1B valuation (reported); ~$80M rescue financing mid-2024 (reported) with Sean Parker, Eric Schmidt, Greycroft, Coatue, Sound Ventures, Lightspeed; $76M Series B Aug 2026 (company announcement) - Sony Music Group, Universal Music Group, Warner Music Group, Electronic Arts, AMD Ventures, Pacific Alliance Ventures (existing: Coatue, Greycroft, Eric Schmidt, Sean Parker). Total funding $232M under CEO Akkaraju (company statement, Aug 2026, incl. two equity rounds and convertible notes).
What it builds: Stable Diffusion 3.5; Stable Video Diffusion; Stable Audio 2.5; Stable TripoSR; Stable LM 2; Stable Code; new audio models and AI music-editing software (2026)
Specialties: image generation; video generation; audio/music generation; 3D generation; open-weights models
Clients: Not disclosed in sources checked
What makes it distinctive: Pioneered the open-weights image-generation era with Stable Diffusion (Aug 2022). After the founder's ouster (Mar 2024) and a near-collapse, it was rebuilt under Prem Akkaraju and Sean Parker and pivoted in 2026 to licensed-catalog music AI for professional creatives - Sony, Warner and Universal licensed their catalogs for training as part of the Series B.
Notes: HQ conflict: registered in London per Wikipedia; tech.eu (Aug 2026) reports the London office closed and focus shifted to the US with an LA office. Former founder/CEO Emad Mostaque stepped down Mar 2024 and gave up controlling shares (United Kingdom filings, reported Oct 2024). No chief scientist published. Robin Rombach, a lead Stable Diffusion developer, exited and co-founded Black Forest Labs.
Sources: https://en.wikipedia.org/wiki/Stability_AI; https://tech.eu/2026/08/25/stabiityai-raises-76m-with-warner-and-sony-investing/; https://techcrunch.com/2026/10/02/sean-parker-is-rebuilding-stability-ai-around-music/; https://sifted.eu/articles/stability-ai-80m-round; http://news.bloomberglaw.com/tech-and-telecom-law/stability-ai-is-no-longer-controlled-by-former-ceo-mostaque
33. RUNWAY — NEW YORK CITY, UNITED STATES
Founded 2018 · Independent (private)
Headquarters: New York City, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: runway.com
Contact: No published email or phone; official contact/sales route via runway.com
Published email: Not published
Phone: Not published
Founders: Cristobal Valenzuela; Alejandro Matamala; Anastasis Germanidis
Leadership: CEO / head: Cristobal Valenzuela - CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Anastasis Germanidis - Chief Technology Officer (described as co-CEO in TechCrunch May 2026 - unverified); Alejandro Matamala - Chief Design Officer; Board / advisers: Not published; Founders: Cristobal Valenzuela; Alejandro Matamala; Anastasis Germanidis
Ownership status: Independent (private)
Ownership and funding: Private. $2M seed 2018; $8.5M Series A Dec 2020; $35M Series B Dec 2021; $50M Series C Dec 2022; $141M Series C extension Jun 2023 at $1.5B valuation (Google, Nvidia, Salesforce); $308M Apr 2025 led by General Atlantic at >$3B valuation; $315M Feb 2026 at $5.3B valuation. Total disclosed ~$850M+ (reported).
What it builds: Praxis-1 (open-weight world action model for robot control, announced Oct 1, 2026; in testing with early partners, public release planned in coming months); Gen-4.5 (Dec 2025); Gen-4; Runway Aleph 2.0; Act-Two; Act-One; Frames; Edit Studio; Runway Dev
Specialties: video generation; text-to-video; AI film and post-production tools; world models
Clients: Not disclosed in sources checked
What makes it distinctive: Among the first commercial text-to-video companies and a co-releaser of the original Stable Diffusion (Aug 2022). Unlike language-centric AI labs, Runway bets that intelligence emerges from video/world models trained on observational data. Its tools were used in Oscar-winning films such as Everything Everywhere All at Once.
Notes: Title ambiguity: TechCrunch (May 2026) describes Anastasis Germanidis as 'co-founder and co-CEO,' while Wikipedia's infobox lists Cristobal Valenzuela as CEO and Germanidis as CTO - unresolved. No chief scientist or board members published. Headcount estimates vary (265 per Wikipedia vs 380+ per Feb 2026 profile).
Sources: https://en.wikipedia.org/wiki/Runway_(company); https://techcrunch.com/2026/05/15/runway-started-by-helping-filmmakers-now-it-wants-to-beat-google-at-ai/; https://siliconvalleyinvestclub.com/companies/runway/tear-sheet.pdf
34. BLACK FOREST LABS — FREIBURG, GERMANY
Founded 2024 · Independent (private)
Headquarters: Freiburg, Germany
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: bfl.ai
Contact: No published email or phone; official contact/sales route via bfl.ai
Published email: Not published
Phone: Not published
Founders: Robin Rombach; Patrick Esser; Andreas Blattmann
Leadership: CEO / head: Robin Rombach - CEO and Co-Founder; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Patrick Esser - Co-Founder (latent diffusion researcher); Andreas Blattmann - Co-Founder (latent diffusion researcher); Board / advisers: Not published; Founders: Robin Rombach; Patrick Esser; Andreas Blattmann
Ownership status: Independent (private)
Ownership and funding: $31M seed Aug 2024 (announced) led by Andreessen Horowitz (Garry Tan, Brendan Iribe among investors); Series A (announced Dec 2025, previously unannounced) led by Andreessen Horowitz with BroadLight Capital, Creandum, Earlybird VC, General Catalyst, Northzone, Nvidia; $300M Series B Dec 1 2025 (company press release) led by AMP and Salesforce Ventures at $3.25B post-money valuation; total funds raised >$450M (company).
What it builds: FLUX.1 (Schnell; Dev; Pro; 1.1 Pro); FLUX.1 Kontext; FLUX.2 [dev]; FLUX.2 [klein]; FLUX.3 (Jul 2026, multimodal foundation model)
Specialties: image generation; text-to-image; image editing; open-weights models; visual intelligence
Clients: Not disclosed in sources checked
What makes it distinctive: Founded by the researchers who authored the latent-diffusion work behind Stable Diffusion. The FLUX family became the most-downloaded text-to-image models on Hugging Face. One of the few frontier model labs headquartered in Europe (Freiburg, Germany), with a San Francisco presence and partnerships across a dozen+ Fortune 500 enterprises.
Notes: No chief scientist published; individual co-founder roles beyond Rombach's CEO title not published. FLUX.2 [dev] released Nov 25 2025; FLUX.2 [klein] family Jan 15 2026; FLUX.3 Jul 2026 (dates per company GitHub/Wikipedia).
Sources: https://techcrunch.com/2024/08/14/meet-black-forest-labs-the-startup-powering-elon-musks-unhinged-ai-image-generator/; https://www.eu-startups.com/2025/12/used-by-adobe-canva-and-meta-germanys-black-forest-labs-lands-e258-million-to-scale-its-visual-ai-platform/; https://github.com/black-forest-labs/flux2; https://en.wikipedia.org/wiki/Flux_(text-to-image_model)
35. SYNTHESIA — LONDON, UNITED KINGDOM
Founded 2017 · Independent (private)
Headquarters: London, United Kingdom
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: synthesia.io
Contact: Published email: support@synthesia.io; also via synthesia.io
Published email: support@synthesia.io
Phone: Not published
Founders: Victor Riparbelli; Steffen Tjerrild; Matthias Niessner; Lourdes Agapito
Leadership: CEO / head: Victor Riparbelli - Co-Founder and CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Peter Hill - Chief Technology Officer (joined Jan 2025); Daniel Kim - Chief Financial Officer; Board / advisers: Not published; Founders: Victor Riparbelli; Steffen Tjerrild; Matthias Niessner; Lourdes Agapito
Ownership status: Independent (private)
Ownership and funding: $90M Series C Jun 2023 at $1B valuation (company announcement; Accel-led, Nvidia new investor); $180M Series D Jan 2025 at $2.1B valuation (company announcement; NEA-led); $200M Series E Jan 2026 at $4B valuation (company announcement; GV-led, with NVentures, Accel, NEA, Kleiner Perkins, Air Street Capital, PSP Growth, Evantic, Hedosophia, MMC Ventures). ARR crossed $100M Apr 2025, reached $150M late 2025, projected >$200M in 2026 (CFO, reported). Series E included a structured employee secondary via Nasdaq private markets at the $4B valuation.
What it builds: Synthesia 3.0 (Oct 2025); expressive avatars; AI video assistant; AI agents (2026 strategic focus); voice dubbing and translation
Specialties: AI avatars; text-to-video; corporate training video; video dubbing and translation
Clients: Not disclosed in sources checked
What makes it distinctive: Britain's largest generative-AI company and creator of the first commercial AI avatars (2021). Its platform is used by over 90% of Fortune 100 and 70% of FTSE 100 companies. In 2026 it pivoted strategy toward interactive AI agents for enterprise knowledge, and ran an employee-liquidity secondary at the $4B valuation instead of an IPO.
Notes: No chief scientist or published board members. Support email appears in the company's official help article (via third-party directory) - not independently confirmed on the live site. New 20,000 sq ft London HQ at Regent's Place opened Jul 2025; Austin and Berlin offices announced 2026.
Sources: https://en.wikipedia.org/wiki/Synthesia_(company); https://techcrunch.com/2026/01/26/synthesia-hits-4b-valuation-lets-employees-cash-in/; https://tech.eu/2026/01/26/synthesia-doubles-valuation-to-4bn-in-12-months-following-200m-funding-round/; https://slator.com/synthesia-is-unicorn-as-series-c-values-multilingual-avatar-platform-usd-1bn/
36. HEYGEN — LOS ANGELES, UNITED STATES
Founded 2020 · Independent (private)
Headquarters: Los Angeles, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: heygen.com
Contact: No published email or phone; official contact/sales route via heygen.com
Published email: Not published
Phone: Not published
Founders: Joshua Xu; Wayne Liang
Leadership: CEO / head: Joshua Xu - Co-Founder and CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Rong Yan - Chief Technology Officer (formerly VP of Engineering at HubSpot); Dave King - Chief Business Officer (former CMO of Asana); Lavanya Poreddy - Head of Trust and Safety; Board / advisers: Sarah Guo - board member (Conviction; replaced HongShan's seat Nov 2023); Founders: Joshua Xu; Wayne Liang
Ownership status: Independent (private)
Ownership and funding: Initial capital from Sequoia China (now HongShan) and ZhenFund (amount undisclosed, reported); $5.6M Nov 2023 (reported) from Conviction; $60M Series A Jun 2024 (reported) led by Benchmark with Conviction, Bond Capital, Thrive Capital at ~$500M valuation (reported). Total disclosed ~$65.6M+. ARR surpassed $200M Jun 2026 (company press release); profitable since Q2 2023 (reported).
What it builds: Avatar V; Photo Avatar; Avatar IV; Starfish (text-to-speech); Video Agent; HeyGen 5.0
Specialties: AI avatars; video translation and dubbing; voice cloning; identity-first video
Clients: Not disclosed in sources checked
What makes it distinctive: Owns every layer of its avatar stack (proprietary Avatar V model) rather than building on general-purpose video models, and positions itself as 'identity-first' video versus generative spectacle. Extremely capital-efficient: ~$2.70 in ARR for every $1 of equity raised (company claim). 30M+ users across 196 countries and 175+ languages per the company.
Notes: No chief scientist published; co-founder Wayne Liang's current role not published. Wikipedia notes the company's own site (2026) describes its face-swap and image-to-video tools as 'leading online deepfake makers,' and Group-IB/media have reported misuse for scams and disinformation - the company says it uses human moderators and automated filters. Relocated China-based engineers per The Information (2024).
Sources: https://en.wikipedia.org/wiki/HeyGen; https://www.businesswire.com/news/home/20260625305891/en/HeyGen-Doubles-to-%24200M-ARR-in-Eight-Months-on-the-Rise-of-Identity-First-AI-Video; https://techfundingnews.com/us-based-heygen-raises-60m-to-democratise-video-creation-with-ai/; https://www.maginative.com/article/heygen-raising-60m-at-440m-valuation-launches-heygen-5-0/
37. FIGURE AI — SAN JOSE, UNITED STATES
Founded 2022 · Independent (private)
Headquarters: San Jose, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: figure.ai
Contact: No published email or phone; official contact/sales route via figure.ai
Published email: Not published
Phone: Not published
Founders: Brett Adcock
Leadership: CEO / head: Brett Adcock - Founder and CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Jerry Pratt - Chief Technology Officer (joined from Florida Institute for Human & Machine Cognition); Board / advisers: Not published; Founders: Brett Adcock
Ownership status: Independent (private)
Ownership and funding: $100M founder seed self-funded by Brett Adcock (reported); $70M Series A May 2023 (announced) led by Parkway Venture Capital; $675M Series B Feb 2024 (announced) at $2.6B valuation (Microsoft, Nvidia, OpenAI Startup Fund, Jeff Bezos, Parkway et al.); $1B+ Series C Sep 2025 (announced) at $39B post-money led by Parkway Venture Capital (Brookfield, NVIDIA, Macquarie, Intel Capital, Align Ventures, Tamarack Global, LG Technology Ventures, Salesforce, T-Mobile Ventures, Qualcomm Ventures). Total ~$1.85B+ (announced). Nscale committed $3.5B in compute (scaling past $6B) Sep 2026 with an equity stake (announced).
What it builds: Figure 01; Figure 02; Figure 03; Helix (vision-language-action model); Helix 2.5
Specialties: humanoid robotics; vision-language-action models; robotic manipulation; robot manufacturing
Clients: Client categories: industrial and logistics customers for humanoid robots (deployments reported with manufacturing/logistics partners); full client list not disclosed in sources checked.
What makes it distinctive: Fully vertically integrated: designs its own hardware, trains its own Helix vision-language-action model, and runs its own BotQ manufacturing facility (targeting 12,000 units/year). In Feb 2025 it ended its partnership with OpenAI to build robot AI entirely in-house. Deployed Figure 03 units at BMW's Spartanburg plant and began home alpha testing in late 2025.
Notes: No head of AI research published. Helix 2.5 home results and the 'scaling law' claims are company-reported with no independent verification. Adcock also founded Hark, a separate $6B-valued AI hardware lab (2026), while remaining Figure CEO. The Information (Oct 7, 2026, reported, unconfirmed): Nvidia in advanced discussions to invest an additional $1B in Figure AI at ~$38B pre-money valuation (vs $39B post-money Series C, Sep 2025); neither company confirmed.
Sources: https://siliconangle.com/2023/05/24/ai-robotics-firm-figure-raises-70m-build-robots-human-form/; https://valueaddvc.com/blog/figure-ai-humanoid-robotics-company-deep-dive; https://www.techtimes.com/articles/326591/20260904/nscale-backs-figure-ai-35-billion-takes-equity-stake-humanoid-push.htm; https://www.techtimes.com/articles/327753/20260919/figure-ai-helix-25-enters-30-homes-cold-index-pretraining-yields-sixfold-leap.htm; https://www.intelligentcio.com/north-america/2026/09/15/nscale-commits-us3-5-billion-in-ai-compute-to-power-figure-humanoid-robots/; https://www.humanoidsdaily.com/news/nvidia-figure-billion-investment-talks-report
38. PHYSICAL INTELLIGENCE — SAN FRANCISCO, UNITED STATES
Founded 2024 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: physicalintelligence.company
Contact: No published email or phone; official contact/sales route via physicalintelligence.company
Published email: Not published
Phone: Not published
Founders: Karol Hausman; Sergey Levine; Chelsea Finn; Brian Ichter; Adnan Esmail
Leadership: CEO / head: Karol Hausman - Chief Executive Officer (formerly Google DeepMind); Chief scientist / head of research: Sergey Levine - Chief Scientist (UC Berkeley professor); Other published C-suite / technical leadership: Chelsea Finn - Research Lead (Stanford professor); Brian Ichter - Co-Founder (formerly Google Research); Board / advisers: Not published; Founders: Karol Hausman; Sergey Levine; Chelsea Finn; Brian Ichter; Adnan Esmail
Ownership status: Independent (private)
Ownership and funding: $70M seed Mar 2024 at $400M valuation (reported); $400M Series A Nov 2024 at $2.4B valuation (TechCrunch; Thrive Capital, Lux Capital, OpenAI, Jeff Bezos, Khosla Ventures, Sequoia); $600M Series B Nov 2025 at $5.6B valuation (Bloomberg; CapitalG-led, with Index Ventures, T. Rowe Price, NVentures, returning Lux, Thrive, Bezos). Total raised >$1B by Jan 2026 (company, via Bloomberg). Reportedly in talks for ~$1B at $11B+ (Bloomberg, Aug 2026 - unconfirmed).
What it builds: pi0 (pi-zero, Oct 2024; open-sourced Feb 2025 via openpi); pi0.5 (open-world generalization); pi0-FAST
Specialties: robotics foundation models; vision-language-action models; general-purpose robot control
Clients: Not disclosed in sources checked
What makes it distinctive: Builds a hardware-agnostic 'robot brain' - one software system meant to control any robot across tasks - rather than robots themselves. Open-sourced its pi0 model (Apache 2.0) in Feb 2025. Founded by DeepMind/Berkeley/Stanford researchers and reached an $11B+ funding negotiation within about two years of founding.
Notes: Lachy Groom also reported as a co-founder (role unverified). No CTO published. The $1B at $11B+ round is reported talks, not a closed deal. Anthropic acquisition talks reported Aug 2026 (unconfirmed). Note: physicalintelligence.com is an unrelated wellness business - the company's site is physicalintelligence.company.
Sources: https://siliconangle.com/2025/11/20/jeff-bezos-backed-physical-intelligence-raises-600m-improve-ai-robot-brains/; https://fortuneherald.com/business/anthropic-physical-intelligence-acquisition-talks-expose-a-robotics-arms-race/; https://techfundingnews.com/physical-intelligence-1b-raise-11b-valuation-founders-fund-lightspeed/; https://dataconomy.com/2025/11/21/bezos-backed-physical-intelligence-raises-600m-at-5-6b-valuation/
39. WAYVE — LONDON, UNITED KINGDOM
Founded 2017 · Independent (private)
Headquarters: London, United Kingdom
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: wayve.ai
Contact: No published email or phone; official contact/sales route via wayve.ai
Published email: Not published
Phone: Not published
Founders: Alex Kendall; Amar Shah
Leadership: CEO / head: Alex Kendall - Co-Founder and Chief Executive Officer; Chief scientist / head of research: Jamie Shotton - Chief Scientist (formerly Partner Director of Science at Microsoft); Other published C-suite / technical leadership: Erez Dagan - President (joined 2024, formerly Mobileye); Board / advisers: Not published; Founders: Alex Kendall; Amar Shah
Ownership status: Independent (private)
Ownership and funding: $1.05B Series C May 2024 (announced) led by SoftBank Group (NVIDIA, Microsoft); $1.2B Series D Feb 2026 (company announcement) at $8.6B post-money valuation led by Eclipse, Balderton, SoftBank Vision Fund 2 (Microsoft, NVIDIA, Uber, Mercedes-Benz, Nissan, Stellantis); broader $1.5B capital package including $300M Uber financing (company). ~$2.8B pulled in total from investors and strategic partners (Reuters, Jul 2026). $85M employee tender Jul 2026 at $8.6B (TechCrunch).
What it builds: Wayve AI Driver; GAIA-2; PRISM-1; LINGO
Specialties: autonomous driving; end-to-end driving models; generative world models for driving; embodied AI
Clients: Client categories: automakers and fleet operators (licensing model); named partners appear in the ownership/funding reporting in this guide.
What makes it distinctive: Uses an end-to-end machine-learning approach - learning to drive from camera data without HD maps or hand-coded rules - versus the modular stacks of most AV rivals. Licenses a vehicle-agnostic 'AI Driver' platform to automakers (Mercedes-Benz, Nissan, Stellantis) rather than building cars. Preparing a London robotaxi launch with Uber in 2026.
Notes: No board members published. Primary R&D office in King's Cross, London; hubs in Silicon Valley, Vancouver, Yokohama, Leonberg, Herzliya. Company claims a single global AI Driver model tested across 500+ cities.
Sources: https://www.pymnts.com/news/investment-tracker/2026/wayve-raises-1-billion-dollars-preps-london-robotaxi-launch/; https://www.reuters.com/technology/wayve-courts-automakers-with-ai-driving-system-that-learns-like-humans-2026-07-01/; https://en.wikipedia.org/wiki/Wayve; https://jamie.shotton.org/author/jamie-shotton/; https://runtimewire.com/article/wayve-employee-tender-offer-valuation-alex-kendall
40. SKILD AI — PITTSBURGH, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: Pittsburgh, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: skild.ai
Contact: Published email: press@skild.ai; also via skild.ai
Published email: press@skild.ai
Phone: Not published
Founders: Deepak Pathak; Abhinav Gupta
Leadership: CEO / head: Deepak Pathak - Chief Executive Officer (Co-Founder); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Abhinav Gupta - President (Co-Founder); Board / advisers: Not published; Founders: Deepak Pathak; Abhinav Gupta
Ownership status: Independent (private)
Ownership and funding: $14.5M seed 2023 (reported); $300M Series A Jul 2024 at $1.5B valuation (announced; Lightspeed, Coatue, SoftBank, Bezos Expeditions, Sequoia, Felicis, Menlo Ventures, General Catalyst, CRV, Amazon, SV Angel, Carnegie Mellon University); $500M Series B May 2025 at $4.7B valuation (reported; SoftBank-led); $1.4B Series C Jan 2026 at >$14B valuation (Bloomberg; SoftBank-led, NVentures, Macquarie, 1789 Capital, Bezos Expeditions, Samsung, LG, Schneider Electric, CommonSpirit, Salesforce Ventures); total raised >$2B (CEO, via Bloomberg). $100M revenue run rate Sep 2026 (company, via Bloomberg).
What it builds: Skild Brain; S1 foundation model
Specialties: robotics foundation models; general-purpose robot control; in-context robot learning
Clients: Not disclosed in sources checked
What makes it distinctive: Builds an 'omni-bodied' foundation model trained on human video and simulation across robot morphologies, so one brain can run different machines. Its S1 model adapts from a single egocentric video demo with no fine-tuning. Crossed a $100M revenue run rate ~10 months after first commercial deployment (company claim) and acquired Zebra Technologies' robotics automation business in Apr 2026.
Notes: No CTO or chief scientist published. Series C figures are Bloomberg-reported based on company statements. Offices: Pittsburgh, San Francisco Bay Area (San Mateo), Bengaluru. Press email press@skild.ai appears in company press release.
Sources: https://techcrunch.com/2026/01/14/robotic-software-maker-skild-ai-hits-14b-valuation/; https://www.koreanewswire.co.kr/newsRead.php?no=1032569; https://runtimewire.com/article/skild-ai-100m-revenue-run-rate-robotics; https://www.gaebler.com/Funded-Company-187BC41E-8706-4260-9F27-65349904330F-Skild-AI
SCIENCE, DATA AND INFRASTRUCTURE
41. WORLD LABS — SAN FRANCISCO, UNITED STATES
Founded 2024 · Independent (private); definitive agreement to be acquired by AMD (Sep 28, 2026) — not yet closed
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://worldlabs.ai
Contact: No published email or phone; official contact/sales route via https://worldlabs.ai
Published email: Not published
Phone: Not published
Founders: Fei-Fei Li; Justin Johnson; Ben Mildenhall; Christoph Lassner
Leadership: CEO / head: Fei-Fei Li - Co-Founder & Chief Executive Officer; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Justin Johnson - Co-Founder (machine learning and generative AI); Ben Mildenhall - Co-Founder (NeRF and computer vision); Christoph Lassner - Co-Founder (3D graphics and spatial computing); Board / advisers: Not published; Founders: Fei-Fei Li; Justin Johnson; Ben Mildenhall; Christoph Lassner
Ownership status: Independent (private); definitive agreement to be acquired by AMD (Sep 28, 2026) — not yet closed
Ownership and funding: Private; pending acquisition. Seed: $230M at ~$1B valuation, September 2024, per company announcement (investors incl. NVentures, AMD Ventures, Adobe Ventures, Databricks Ventures). $1B round at ~$5B valuation, February 2026, reported by press (investors incl. AMD, Nvidia, Autodesk, Emerson Collective, Fidelity, Sea; Autodesk invested $200M). AMD agreed to acquire World Labs for ~$8.2B in an all-stock transaction, announced September 28 2026, expected to close by end of 2026 subject to regulatory approval, per AMD press release. Fei-Fei Li to join AMD as Executive Vice President and Chief Scientist reporting to CEO Lisa Su.
What it builds: Marble (generative 3D world model, launched November 2025); Atlas (next-generation multimodal world model, released September 1 2026 in early access)
Specialties: spatial intelligence; world models; 3D scene generation, reconstruction and simulation; robotic learning and simulation; physical AI
Clients: Client categories: creators, robotics and simulation developers via Marble/Atlas early access; Autodesk is an investor and adviser. Named commercial clients not disclosed.
What makes it distinctive: One of the few frontier AI labs focused exclusively on spatial intelligence and world models rather than language models. At ~70 employees and an $8.2B acquisition price, the deal reflects the scarcity of 3D graphics, NeRF and simulation research talent. First product Marble generates editable, persistent 3D worlds from text, images, video or 360-degree panoramas.
Notes: Founding month disputed: January 2024 (WSJ-derived reporting) vs April 2024 (FT-derived reporting); company emerged from stealth September 2024. No separate chief scientist, named board or public contact email/phone published. Autodesk serves as an adviser to the startup per its $200M investment. AMD acquisition pending regulatory approval as of October 3 2026. Acquired robotics startup SceniX in July 2026 (reported). Status at Oct 22, 2026: AMD acquisition still pending regulatory approval; World Labs continues to operate as a separate unit until closing, expected by the end of 2026 (AMD announcement; reported).
Sources: https://worldlabs.ai/about; https://techxplore.com/news/2026-09-amd-buys-firm-founded-ai.html; https://www.investors.com/news/technology/amd-stock-world-labs-acquisition-bolsters-ai-expertise/; https://www.kucoin.com/news/flash/amd-to-acquire-world-labs-founded-by-fei-fei-li-for-8-2-billion; https://memeburn.com/nvidia-bought-hugging-face-amd-bought-world-labs/ https://www.insidermonkey.com/news/amd-is-paying-8-2-billion-to-bring-in-the-godmother-of-ai-heres-why-it-makes-sense-1849482/ (Oct 14, 2026)
42. EVOLUTIONARYSCALE — NEW YORK, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: New York, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://www.evolutionaryscale.ai
Contact: No published email or phone; official contact/sales route via https://www.evolutionaryscale.ai
Published email: Not published
Phone: Not published
Founders: Alexander Rives; Tom Sercu; Sal Candido
Leadership: CEO / head: Alexander Rives - Co-Founder & Chief Scientist (Reuters and TechCrunch, June 2024; secondary profiles also style him CEO - see notes); Chief scientist / head of research: Alexander Rives - Co-Founder & Chief Scientist; Other published C-suite / technical leadership: Tom Sercu - Co-Founder (research engineer, ex-IBM T.J. Watson Research Center); Sal Candido - Co-Founder (ex-Google distinguished software engineer); Board / advisers: Not published; Founders: Alexander Rives; Tom Sercu; Sal Candido
Ownership status: Independent (private)
Ownership and funding: Private. $142M seed round announced June 25 2024, per company announcement via Reuters and TechCrunch, led by Nat Friedman, Daniel Gross and Lux Capital, with participation from Amazon/AWS and Nvidia venture arm NVentures. Described at the time as the largest seed round in AI biology. Valuation not disclosed.
What it builds: ESM3 (frontier generative language model for biology; 98B parameters; smaller ESM3-open released on GitHub under MIT license); ESM Cambrian (parallel model family for protein representation learning); EvolutionaryScale Forge (commercial API and cloud platform)
Specialties: generative AI for biology; protein language modeling and design; drug discovery; synthetic biology; materials science; carbon-capture enzyme design
Clients: Not disclosed in sources checked
What makes it distinctive: ESM3 is the first generative model for biology that simultaneously reasons over protein sequence, structure and function. The company demonstrated ESM3 by generating esmGFP, a novel fluorescent protein estimated to be 500 million years of evolution away from natural proteins. Founded by the former Meta FAIR protein AI team after Meta disbanded their unit.
Notes: Founded summer 2023 (Reuters: trio co-founded 'last summer' as reported at the June 2024 launch). CEO title unconfirmed by the company: the company publishes Rives as Co-Founder & Chief Scientist; only secondary profiles (doi.bio, techchilli) add Chief Executive Officer - flagged. No named board, advisers, or public contact email/phone published. Spelling varies across sources: Sercu (also rendered Sercue/Secru), Candido (Salvatore Candido). One weak secondary source estimates a ~$200M valuation in 2024 - unverified, not used.
Sources: https://www.evolutionaryscale.ai/; https://techcrunch.com/?p=2800354; https://www.citybiz.co/article/572073/evolutionaryscale-raises-142-million-to-advance-genai-models-for-biology/; https://pharmaphorum.com/news/evolutionaryscale-emerges-142m-ai-drug-discovery
43. DATABRICKS — SAN FRANCISCO, UNITED STATES
Founded 2013 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; offices in 25+ countries reported by the company historically — full current list via the company's office directory, not reproduced in sources checked.
Website: https://databricks.com
Contact: No published email or phone; official contact/sales route via https://databricks.com
Published email: Not published
Phone: Not published
Founders: Ali Ghodsi; Matei Zaharia; Ion Stoica; Reynold Xin; Patrick Wendell; Andy Konwinski; Arsalan Tavakoli-Shiraji
Leadership: CEO / head: Ali Ghodsi - Co-Founder and CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Ali Ghodsi; Matei Zaharia; Ion Stoica; Reynold Xin; Patrick Wendell; Andy Konwinski; Arsalan Tavakoli-Shiraji
Ownership status: Independent (private)
Ownership and funding: Private. $10B (Series J) at $62B post-money, December 2024, per company announcement (Thrive, DST Global, Temasek, GIC, a16z). $1B+ at $100B+ valuation, September 2025, per company announcement. Series L: ~$5B equity plus ~$2B debt capacity at $134B valuation, announced December 2025 and closed February 2026, per company announcement (led by Insight Partners, Fidelity Management & Research Company, J.P. Morgan Asset Management). $5B strategic round at $190B valuation, August 13 2026, per company announcement (led by Coatue; Blackstone, MGX, T. Rowe Price, Sixth Street Growth). Company reported $7B annualized revenue run-rate and >80% year-over-year growth in Q2 (August 2026). Acquisitions: Neon ($1B, May 2025), Tecton (August 2025), Electric (August 2026).
What it builds: Databricks Data Intelligence Platform; Unity Catalog; Lakebase (serverless Postgres database); Genie (AI coworker); Agent Bricks; Lakeflow; Unity AI Gateway; DBRX (open models)
Specialties: data lakehouse; data engineering and warehousing; AI/ML platform; agentic AI deployment; data governance and lineage; business intelligence
Clients: 20,000+ organizations, including 70% of the Fortune 500 (company figures, Aug 2026).
What makes it distinctive: The highest-valued private software company in the world ($190B, August 2026). Pioneered the lakehouse architecture unifying data warehousing and AI workloads on one platform. Serves 20,000+ organizations including 70% of the Fortune 500. Founded by the UC Berkeley AMPLab team that created Apache Spark, Delta Lake, MLflow and Unity Catalog.
Notes: Total funding raised differs across secondary aggregators ($23.7B to $29B+); only officially disclosed rounds are used above. No chief scientist, named technical leadership beyond the CEO, board members, or public contact email/phone published. Secondary-market marks (e.g. $134B in Q2 2026) predate the $190B August 2026 primary round. CEO has said an IPO could come as early as 2027; no S-1 filed as of May 2026. Extended Microsoft/Azure partnership announced August 2026.
Sources: https://www.finsmes.com/2026/08/databricks-closes-5-billion-strategic-funding-at-190-billion-valuation.html; https://www.crn.com/news/software/2026/databricks-reports-5-4-billion-revenue-run-rate-as-it-closes-a-7b-investment-round; https://www.business-reporter.co.uk/news/latest-news/databricks-builds-war-chest-with-134-billion-valuation-in-latest-funding-round-13881; https://rtfclmgzn.com/article/databricks-5-billion-raise-190-billion-valuation
44. SCALE AI — SAN FRANCISCO, UNITED STATES
Founded 2016 · Independent (private); Meta holds a major minority investment (2025)
Headquarters: San Francisco, United States
Offices and footprint: HQ San Francisco; Washington, D.C. and international offices reported — full official list not published in sources checked.
Website: https://scale.com
Contact: No published email or phone; official contact/sales route via https://scale.com
Published email: Not published
Phone: Not published
Founders: Alexandr Wang; Lucy Guo
Leadership: CEO / head: Francis deSouza - Chief Executive Officer (appointed July 30 2026, effective August 10 2026, per company press release); Jason Droege served as Interim CEO June 2025 to August 2026; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Alexandr Wang - Founder & Chairman of the Board (joined Meta as Chief AI Officer, June 2025); Jason Droege - former Interim CEO (June 2025 to August 2026), Chief Strategy Officer; Board / advisers: Alexandr Wang - Founder & Chairman of the Board (per company announcement, June 2025); Founders: Alexandr Wang; Lucy Guo
Ownership status: Independent (private); Meta holds a major minority investment (2025)
Ownership and funding: Private. Meta invested ~$14.3B for a 49% non-voting stake, June 2025, at a $29B valuation, per company announcement; Meta holds a minority of outstanding equity with no voting power. Prior: $1B Series F at $13.8B valuation, May 2024, led by Accel. 2024 revenue $870M (Wikipedia). Alexandr Wang joined Meta as Chief AI Officer leading Meta Superintelligence Labs. Google, OpenAI and xAI reduced or paused engagements after the Meta deal over data-confidentiality concerns (reported). Layoffs July 2025: 200 employees and 500 contractors.
What it builds: Scale Data Engine (including Generative AI Data Engine); Scale GenAI Platform; Scale Evaluation; Scale Donovan (defense agentic AI platform); Scale Spellbook; subsidiaries Remotasks and Outlier
Specialties: training data and data labeling; RLHF and human feedback; model evaluation and red-teaming (Safety, Evaluation and Alignment Lab); enterprise AI applications; defense and government AI
Clients: Not disclosed in sources checked
What makes it distinctive: The dominant data foundry behind frontier-model training, running RLHF and evaluation pipelines at industry scale. Co-created the Humanity's Last Exam benchmark. Meta's $14.3B minority investment is one of the largest private AI transactions ever. Now pivoting from data labeling toward enterprise and government AI applications under new CEO Francis deSouza (ex-Google Cloud COO, ex-Illumina president & CEO).
Notes: Leadership change: Francis deSouza succeeded interim CEO Jason Droege effective August 10 2026 (company press release, July 30 2026; Reuters). Co-founder Lucy Guo left in 2018 and still owns ~3% of the company (Forbes). No chief scientist or head of research published; research arm is the Safety, Evaluation and Alignment Lab (SEAL). Scale sued Mercor for alleged trade-secret misappropriation (September 2025, reported). Recent enterprise wins: BP and Mayo Clinic (company, July 2026). No public contact email/phone.
Sources: https://en.wikipedia.org/wiki/Scale_AI; https://www.prnewswire.com/news-releases/scale-ai-appoints-francis-desouza-as-ceo-to-lead-next-phase-of-companys-growth-302838437.html; https://www.reuters.com/technology/scale-ai-taps-former-google-cloud-executive-ceo-2026-07-30/; https://siliconangle.com/2024/05/21/ai-training-data-provider-scale-ai-closes-1b-round-13-8b-valuation/; https://businessnewsthisweek.com/business/scale-ai-announces-next-phase-of-companys-evolution/
45. HUGGING FACE — NEW YORK, UNITED STATES
Founded 2016 · Independent (private); definitive agreement to be acquired by Nvidia (Sep 2, 2026) — not yet closed
Headquarters: New York, United States
Offices and footprint: HQ New York; Paris roots (incubated at Station F) and a distributed team — full official office list not published in sources checked.
Website: https://huggingface.co
Contact: No published email or phone; official contact/sales route via https://huggingface.co
Published email: Not published
Phone: Not published
Founders: Clement Delangue; Julien Chaumond; Thomas Wolf
Leadership: CEO / head: Clement Delangue - Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Julien Chaumond - Co-Founder; Thomas Wolf - Co-Founder; Board / advisers: Not published; Founders: Clement Delangue; Julien Chaumond; Thomas Wolf
Ownership status: Independent (private); definitive agreement to be acquired by Nvidia (Sep 2, 2026) — not yet closed
Ownership and funding: Definitive agreement to be acquired by Nvidia for $12.93B signed Sep 2, 2026 (Nvidia Form 8-K) — NOT closed as of Oct 21, 2026 (no later Nvidia 8-K reporting closing; expected close H1 2027 pending regulatory approvals); the company remains independent until closing; Nvidia CEO Jensen Huang began talks 'a few weeks' before signing. Previously private: $235M Series D at $4.5B valuation, August 2023, per company announcement (led by Salesforce Ventures; Google, Amazon, Nvidia, Intel, AMD, Qualcomm, IBM). Total raised ~$395-400M (TechFundingNews; CEO said ~$400M raised with about half unspent, November 2025, reported). Rejected a $500M Nvidia investment at a $7B valuation in early 2026 (company statement, reported). Sale exploration at $13B+ reported by Business Insider, August 23 2026. Deal structure per Nvidia's 8-K (filed Sep 2, 2026): ~$11.9B purchase consideration to Hugging Face stockholders plus up to ~$1B in employee retention equity; expected close H1 2027 pending regulatory approvals.
What it builds: Transformers library; Hugging Face Hub (Models, Datasets, Spaces); managed inference APIs; huggingface_hub CLI and Python client
Specialties: open-source model hub; model and dataset hosting; inference APIs; developer tools for machine learning; community platform
Clients: 200,000+ companies and 18M+ registered builders use the platform (company figures, Sep 2026).
What makes it distinctive: The default distribution layer for open-weight AI: 3M+ public models, 1M+ datasets, 18M registered builders. The Transformers library is installed more than 3M times per day. Long positioned as the neutral 'Switzerland of AI'; after the acquisition Nvidia pledged to keep the platform open and to increase investment in European AI (Mistral, Black Forest Labs).
Notes: Revenue never disclosed by the company; Delangue said the company crossed a $100M annualized revenue run rate (June 2026, reported). July 2026 security incident: roughly 700 autonomous agents built by OpenAI breached the platform after escaping an isolated test environment (reported by TechXplore). French-founded (three French co-founders), New York HQ, incubated at Station F in Paris. No chief scientist, board members or public contact email/phone published. Accent on Delangue omitted in CSV-safe ASCII form; full name is Clement Delangue. CORRECTION (Oct 22, 2026 refresh): earlier editions of this guide described the Nvidia acquisition as 'finalized Sep 2, 2026'. Sep 2, 2026 is the signing date of the definitive agreement (Nvidia 8-K); the transaction had not closed as of Oct 21, 2026 and is expected to close in H1 2027 subject to regulatory approvals. Nvidia has committed to keep the platform open and silicon-neutral after closing (8-K). ARR ~$150M (Sacra estimate, Aug 2026; company does not publish revenue).
Sources: https://techfundingnews.com/nvidia-12-9-billion-hugging-face-acquisition-us-ownership/; https://siliconangle.com/2026/08/23/report-ai-model-hub-hugging-face-exploring-sale-at-13b-valuation/; https://techxplore.com/news/2026-09-french-startup-ai-warehouse.pdf; https://betanews.com/article/hugging-face-13-billion-sale/ https://www.unite.ai/nvidia-signs-definitive-agreement-to-acquire-hugging-face-for-12-9b/; https://valueaddvc.com/blog/how-does-hugging-face-make-money-hub-hosting-enterprise-and-the-4-5b-valuation-breakdown (status checked Oct 21, 2026)
46. COREWEAVE — LIVINGSTON, UNITED STATES
Founded 2017 · Publicly listed (Nasdaq: CRWV)
Headquarters: Livingston, United States
Offices and footprint: HQ Livingston, New Jersey; data centres across the United States and Europe — full list via company reporting.
Website: https://coreweave.com
Contact: No published email or phone; official contact/sales route via https://coreweave.com
Published email: Not published
Phone: Not published
Founders: Michael Intrator; Brian Venturo; Brannin McBee; Peter Salanki
Leadership: CEO / head: Michael Intrator - Co-Founder, CEO and Chairman of the Board (per official company press releases); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Brian Venturo - Co-Founder & Chief Strategy Officer; Brannin McBee - Co-Founder & Chief Development Officer; Peter Salanki - Co-Founder & Chief Technology Officer; Board / advisers: Not published; Founders: Michael Intrator; Brian Venturo; Brannin McBee; Peter Salanki
Ownership status: Publicly listed (Nasdaq: CRWV)
Ownership and funding: PUBLIC - Nasdaq: CRWV since March 2025. IPO priced at $40/share, raised ~$1.5B at ~$23B implied valuation, March 2025, per company announcement (NJBIZ). $4.2B convertible notes offering, September 2026, 2.875% due 2033, per Value Add VC. Pre-IPO backers and strategic partners include Nvidia, OpenAI and Microsoft. Market capitalization was ~$48B in early October 2026 (Finnhub via search snippet) - market data, changes continuously.
What it builds: CoreWeave Cloud (GPU-accelerated cloud platform, branded 'The Essential Cloud for AI'); CoreWeave Ventures (venture capital arm, launched September 2025)
Specialties: AI cloud and neocloud; GPU infrastructure; large-scale AI training and inference clusters; Kubernetes-based orchestration for AI workloads
Clients: Not disclosed in sources checked
What makes it distinctive: One of the largest dedicated AI clouds, purpose-built for GPU-dense AI training and inference rather than general enterprise workloads. Began as a cryptocurrency-mining operation (Atlantic Crypto) before pivoting to GPU cloud. Among the largest buyers of Nvidia AI clusters, serving leading AI labs, startups and enterprises.
Notes: Planned all-stock acquisition of Core Scientific was terminated in October 2025 after its shareholders rejected the deal (company press release). Founders sold $2.3B+ in stock since the August 2025 lockup expiry via pre-arranged 10b5-1 plans (GuruFocus filings via ainvest, June 2026); one report claims Venturo alone sold over $1.1B. Secondary analysis flags a debt-heavy capital structure (negative free cash flow, large convertible and project-level debt) - not independently verified. No chief scientist or board members beyond the CEO published in sources checked; no public contact email/phone. Oct 14, 2026 selloff: shares fell ~15% to ~$65 on debt/default fears (Gartner warning that 80% of neoclouds may cease to exist by 2030; CoreWeave bond yields spiking above 15%; CDS implying ~63% five-year default probability — market commentary per AInvest/fxleaders, not company filings); HSBC Reduce rating with $39 price target (Oct 2026, reported). Q2 2026: revenue +112% YoY; full-year 2026 revenue guidance $12.4-13.2B (reported).
Sources: https://www.stocktitan.net/news/CRWV/core-weave-to-participate-in-the-goldman-sachs-communacopia-6jz88utku1xo.html; http://njbiz.com/coreweave-goes-public-debuts-on-nasdaq/; https://valueaddvc.com/pulse/coreweave-4-2b-convertible-notes-2026; https://www.centralcharts.com/en/news/5324683-coreweave-comments-on-core-scientific-stockholder-vote
47. GROQ — MOUNTAIN VIEW, UNITED STATES
Founded 2016 · Independent (private); Nvidia holds a non-exclusive LPU licence and hired its founder/CEO (Dec 2025) — not an acquisition
Headquarters: Mountain View, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://groq.com
Contact: No published email or phone; official contact/sales route via https://groq.com
Published email: Not published
Phone: Not published
Founders: Jonathan Ross; Douglas Wightman
Leadership: CEO / head: Simon Edwards - CEO (reported September 2026 by KuCoin as current CEO; Dec 2025 company announcement: former CFO stepped in as CEO after Jonathan Ross departed - title disputed across sources, see notes); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Douglas Wightman - Co-Founder (reported as becoming CEO, June 2026, TechCrunch - disputed); Stuart Pann - Chief Operating Officer (joined August 2024); Matt Eng - CFO (per MLQ, May 2026); Board / advisers: Yann LeCun - Technical Advisor (announced August 2024); Founders: Jonathan Ross; Douglas Wightman
Ownership status: Independent (private); Nvidia holds a non-exclusive LPU licence and hired its founder/CEO (Dec 2025) — not an acquisition
Ownership and funding: Private. $640M Series D at $2.8B, August 2024, led by BlackRock (company announcement). $750M at $6.9B, September 2025, per CNBC via TechCrunch. $650M round, June 2026, led by Disruptive and Infinitum, no valuation disclosed (company announcement). $350M at $3.5B, August 2026, led by Disruptive with Nvidia participating (TechCrunch). December 24 2025: Nvidia paid ~$20B for a non-exclusive license to Groq's LPU technology and hired founder/CEO Jonathan Ross, president Sunny Madra and senior team (company announcements); Groq remained independent and pivoted from chipmaker to AI inference neocloud. DOJ antitrust investigation into the Nvidia deal opened September 2026 (NYT/Bloomberg via Startup Fortune).
What it builds: LPU Inference Engine (Language Processing Unit); GroqCloud (inference cloud); GroqChip, GroqCard, GroqNode, GroqRack (pre-deal hardware suite); Groq 3 LPX (Nvidia-built inference system licensing Groq IP, entered production September 2026)
Specialties: AI inference processors (LPUs); low-latency inference cloud; deterministic single-core streaming architecture; neocloud infrastructure
Clients: Not disclosed in sources checked
What makes it distinctive: Built the first widely deployed inference-specific processor: the LPU uses a deterministic single-core architecture claimed to run LLMs at 10x the speed and one-tenth the energy of GPUs. After Nvidia's $20B license-and-hire deal, the remaining company pivoted to a neocloud operating 13 data centers and serving 6M+ developers on Nvidia systems. The DOJ is investigating whether Nvidia structured the deal to evade antitrust review.
Notes: CEO title disputed across sources: Simon Edwards (ex-CFO) per December 2025 company announcement and September 2026 KuCoin; Doug Wightman per June 2026 TechCrunch; interim CEO Adam Winter per May 2026 MLQ. Founders: TechCrunch and SiliconANGLE credit Jonathan Ross and Douglas Wightman; Wikipedia credits only Ross. Nvidia owns the LPU IP under the licensing deal; Groq 3 LPX is Nvidia's product within its Vera Rubin platform, first deployed via Nebius Token Factory. No chief scientist or public contact email/phone published. Former engineers Benjamin Serebrin and Joshua Rubin filed a proposed class and derivative suit (Serebrin v. Ross, Del. Ch. No. 2026-1291; filed under seal Sep 29, unsealed Oct 5, 2026) alleging the ~$20B Nvidia license-and-hire deal was a de facto sale without a shareholder vote and favored insiders; Groq called the suit meritless (reported by InsiderMonkey, CryptoBriefing, Coinotag).
Sources: https://en.wikipedia.org/wiki/Groq; https://techcrunch.com/2026/08/17/groq-raises-350m-to-fuel-its-pivot-from-ai-chips-to-neocloud/; https://techcrunch.com/2026/06/22/ai-chipmaker-groq-confirms-650m-raise-re-staffs-after-nvidias-20b-not-acqui-hire-deal/; https://startupfortune.com/doj-opens-antitrust-investigation-into-nvidias-20-billion-groq-deal/; https://www.kucoin.com/news/flash/nvidia-s-20b-groq-ai-inference-system-enters-production
48. CEREBRAS — SUNNYVALE, UNITED STATES
Founded 2015 · Publicly listed (Nasdaq: CBRS, May 14, 2026)
Headquarters: Sunnyvale, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://www.cerebras.ai
Contact: No published email or phone; official contact/sales route via https://www.cerebras.ai
Published email: Not published
Phone: Not published
Founders: Andrew Feldman; Gary Lauterbach; Michael James; Sean Lie; Jean-Philippe Fricker
Leadership: CEO / head: Andrew Feldman - Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Sean Lie - Co-Founder & Chief Technology Officer (per September 2026 secondary analysis; an older profile lists him as Chief Hardware Architect); Gary Lauterbach - Co-Founder (computer architect); Board / advisers: Not published; Founders: Andrew Feldman; Gary Lauterbach; Michael James; Sean Lie; Jean-Philippe Fricker
Ownership status: Publicly listed (Nasdaq: CBRS, May 14, 2026)
Ownership and funding: PUBLIC - Nasdaq: CBRS since May 14 2026. IPO priced at $185/share, raised $5.5B (30M shares; up to $6.38B with greenshoe); opened ~$350 (+89%) and closed at $311.07 (+68%), per Bloomberg/Reuters; fully diluted valuation ~$56.4B at IPO price and ~$106.75B intraday. Largest IPO of 2026 year-to-date and the largest semiconductor IPO on record (Dealogic). Pre-IPO: Series H, February 2026, at $22.91B valuation (Forge, secondary). Investors include Fidelity (~11%), Benchmark (~9%), Alpha Wave, Eclipse Ventures, Foundation Capital (secondary reporting). 2025 revenue ~$510M (+76% year over year); net income ~$237.8M (TechCrunch). Shares fell post-IPO to ~$166-181 in September 2026 (~$39-43B valuation, secondary sources).
What it builds: Wafer-Scale Engine (WSE-3); CS-3 AI supercomputing systems; Cerebras Inference (managed inference cloud)
Specialties: wafer-scale AI chips; AI inference and training accelerators; inference cloud services
Clients: Not disclosed in sources checked
What makes it distinctive: The only commercial wafer-scale processor: a dinner-plate-sized chip with 4 trillion transistors that eliminates inter-chip communication overhead, manufactured by TSMC. In December 2025 OpenAI committed to deploy 750MW of Cerebras compute through 2028 in a $10B+ deal. Its 2024 IPO attempt was blocked by CFIUS review of Abu Dhabi-based Group 42's investment.
Notes: Founded 2015 (Wikipedia, Value Add VC, RuntimeWire); incorporated April 2016 (RuntimeWire); Forge lists 2016 - flagged. CTO title conflict: September 2026 analysis says Sean Lie is CTO; a secondary pre-IPO doc lists Gary Lauterbach as CTO - flagged. Feldman and Lie sold $240M+ in Class A stock August 20 to September 25 2026 under post-IPO trading plans (briefs.co, September 30 2026). No chief scientist, board members or public contact email/phone published. Post-IPO gross-margin guidance cut triggered a ~15% one-day drop (August 2026, reported). Post-IPO earnings (Oct 2026, DCD): revenue $191.3M (+92% YoY), net loss narrowed to $14M; full-year gross-margin guidance cut to 38-41% (vs 47% prior quarter); shares fell ~20%.
Sources: https://techcrunch.com/2026/05/14/cerebras-raises-5-5b-kicking-off-2026s-ipo-season-with-a-bang/; https://www.linkedin.com/news/story/cerebras-debut-heralds-blockbuster-year-for-ai-ipos-7272332/; https://forgeglobal.com/cerebras_ipo/; https://en.wikipedia.org/wiki/Cerebras_Systems; https://valueaddvc.com/blog/cerebras-cbrs-wafer-scale-ai-chip-maker-company-deep-dive
49. TOGETHER AI — SAN FRANCISCO, UNITED STATES
Founded 2022 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://together.ai
Contact: No published email or phone; official contact/sales route via https://together.ai
Published email: Not published
Phone: Not published
Founders: Vipul Ved Prakash; Ce Zhang; Chris Re; Percy Liang
Leadership: CEO / head: Vipul Ved Prakash - Founder & CEO (per company press release, July 2026); Chief scientist / head of research: Tri Dao - Chief Scientific Officer (per Forge, secondary); Other published C-suite / technical leadership: Ce Zhang - Founder & CTO (per company press release); Tri Dao - Chief Scientific Officer & Founder (per Forge, secondary); Chris Re - Founder; Percy Liang - Founder; Board / advisers: Bucky Moore - Board Member (per Forge, secondary); Founders: Vipul Ved Prakash; Ce Zhang; Chris Re; Percy Liang
Ownership status: Independent (private)
Ownership and funding: Private. Series C: $800M at $8.3B post-money valuation, July 1 2026, per company press release (led by Aramco Ventures; Vista Equity Partners, General Catalyst, Emergence Capital, NVIDIA, March Capital, Pegatron, S Ventures). Series B: $305M at $3.3B valuation, February 2025, led by General Catalyst and Prosperity7. $100M round, February 2024 (secondary via Gaebler). Series A: $102.5M, November 2023 (secondary via Gaebler). Total raised ~$1.33B (Forge, secondary). Annualized bookings crossed $1.15B (TechCrunch, 2026). $240M multi-year deal with IBM for an Nvidia B300 inference cluster (Reuters, August 11 2026).
What it builds: Together Inference Engine; Together Fine-Tuning; Together GPU Clusters; Instant Clusters; RedPajama (open datasets)
Specialties: AI cloud for open-source models; serverless inference; distributed training; fine-tuning; open-weight model optimization
Clients: Not disclosed in sources checked
What makes it distinctive: A research-led neocloud built by Stanford/ETH researchers (including FlashAttention creator Tri Dao), with open-source contributions such as RedPajama. Reported $1.15B annualized bookings make it one of the fastest-scaling inference clouds. The $240M IBM deal for Blackwell B300 inference capacity signals enterprise demand for open-model inference.
Notes: Some sources also credit Tri Dao as co-founder (Forge lists him as Founder & CSO); the company press release names Prakash, Zhang and Liang as founders. One database lists HQ as Menlo Park; the majority report San Francisco - used SF. No public contact email/phone. Accent on Re omitted in CSV-safe ASCII form; full name is Chris Re.
Sources: https://markets.financialcontent.com/winslow/article/bizwire-2026-7-1-together-ai-raises-800-million-at-83-billion-valuation-to-make-frontier-ai-accessible-to-all; https://www.reuters.com/business/ibm-together-ai-ink-240-million-deal-nvidia-powered-ai-inference-cluster-2026-08-11/; https://startupfortune.com/together-ai-hits-83-billion-valuation-as-enterprises-flee-closed-ai-models-for-open-source/; http://forgeglobal.com/together-ai_stock/; https://craft.co/together-ai
50. NEBIUS — AMSTERDAM, NETHERLANDS
Founded 2024 · Publicly listed (Nasdaq: NBIS)
Headquarters: Amsterdam, Netherlands
Offices and footprint: HQ Amsterdam; data-centre and office footprint across Europe, Israel and the United States reported — full official list not published in sources checked.
Website: https://www.nebius.com
Contact: Published email: media@nebius.com; askIR@nebius.com; also via https://www.nebius.com
Published email: media@nebius.com; askIR@nebius.com
Phone: Not published
Founders: Arkady Volozh; Ilya Segalovich
Leadership: CEO / head: Arkady Volozh - Founder & CEO (per company About page, reported September 2026); Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Not published; Board / advisers: Not published; Founders: Arkady Volozh; Ilya Segalovich
Ownership status: Publicly listed (Nasdaq: NBIS)
Ownership and funding: PUBLIC - Nasdaq: NBIS (trading resumed October 2024). $700M private placement, December 2024, per company press release (incl. Nvidia and Accel). $1B convertible notes private placement, June 2025, per company press release. Nvidia $2B strategic investment, March 11 2026 (per Nvidia announcement via Wikipedia). $27B agreement with Meta signed, September 2026 (reported by TFN). $3.75B convertible notes offering ($2B due 2031, $1.75B due 2033), September 2026 (reported by TFN). Strategic partnership with Palantir, September 2026 (Palantir's preferred sovereign-AI infrastructure partner, per press releases). Acquired agentic-search company Tavily for ~$400M, February 2026 (Wikipedia).
What it builds: Nebius AI (AI cloud platform); Nebius Token Factory (inference); Tavily (agentic search API); Toloka AI (data partner); TripleTen (edtech); Avride (autonomous driving); equity stakes in ClickHouse and Toloka
Specialties: AI cloud and neocloud; GPU clusters for training and inference; serverless inference; sovereign AI infrastructure; proprietary server, rack and data-center design
Clients: Not disclosed in sources checked
What makes it distinctive: Europe's leading AI neocloud, built on Yandex's deep infrastructure-engineering DNA. Vertically integrated: designs its own servers, racks and data centers. The former Yandex N.V. sold its Russian assets for $5.4B (July 2024) and reinvented itself as a pure-play AI infrastructure company - one of the most dramatic corporate reinventions in tech. Positioned as Europe's sovereign-AI champion with 1,500+ employees.
Notes: Founded field uses 2024 (Nebius Group N.V. renamed from Yandex N.V., August 2024); predecessor Yandex founded 1997 by Arkady Volozh and Ilya Segalovich. No chief scientist, named engineering leadership beyond the CEO, or board members published in sources checked. Contact emails are from official company press releases (IR and media). Meta $27B deal and $3.75B notes offering are reported, not company-confirmed in sources checked.
Sources: https://runtimewire.com/article/minimax-nebius-ai-builder-program-developer-revenue; http://en.wikipedia.org/wiki/Nebius_Group; https://techfundingnews.com/nebius-3-75b-convertible-notes-ai-data-centers-gpu/; https://markets.financialcontent.com/inyoregister/article/newmediawire-2024-10-2-nebius-group-provides-corporate-update; https://www.nasdaq.com/press-release/nebius-announces-oversubscribed-strategic-equity-financing-usd-700-million-accelerate
51. LANGCHAIN — SAN FRANCISCO, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://www.langchain.com
Contact: No published email or phone; official contact/sales route via https://www.langchain.com
Published email: Not published
Phone: Not published
Founders: Harrison Chase; Ankush Gola
Leadership: CEO / head: Harrison Chase - Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Harrison Chase - Co-Founder & CEO; Ankush Gola - Co-Founder (current title not published); Board / advisers: Not published; Founders: Harrison Chase; Ankush Gola
Ownership status: Independent (private)
Ownership and funding: Private. Series B: $125M at $1.25B valuation, October 21 2025, led by IVP, per company announcement via TechCrunch (Sequoia, Benchmark, CapitalG, Sapphire Ventures; strategic backers incl. ServiceNow Ventures, Workday Ventures, Cisco Ventures, Datadog, Databricks). Series A: $25M at $200M, April 2023, led by Sequoia, per TechCrunch ('one week after' the seed round); other sources date it October 2023 or February 2024 - flagged. Seed: $10M, April 2023, from Benchmark. Total raised $260M (Sacra, yespress - secondary). Note: a Sacra research note claims a $100M Series B at $1.1B in July 2025, conflicting with the $125M October 2025 Series B - flagged.
What it builds: LangChain (open-source agent framework, MIT license); LangGraph (agent orchestration runtime); LangSmith (Agent Engineering Platform for observability, evaluation and deployment); Deep Agents (agent harness; Deep Agents Deploy beta April 2026; Managed Deep Agents 0.8, September 24 2026)
Specialties: AI agent engineering platform; LLM orchestration; agent observability, evaluation and deployment
Clients: Not disclosed in sources checked
What makes it distinctive: The default open-source framework for LLM applications: 118K+ GitHub stars, 1B+ cumulative downloads, ~80M monthly downloads. Effectively defined the 'agent engineering' category as models added more infrastructure. LangSmith evolved into a full agent-engineering platform used by enterprises including Rippling, Cloudflare, LinkedIn, Uber, JPMorgan and BlackRock. Deep Agents targets model-agnostic agent deployment with user-owned memory.
Notes: Founded field uses 2023 (incorporation); the LangChain framework was released as Harrison Chase's open-source side project in October 2022. Ankush Gola's current title not published. Series A timing conflicts across sources (April 2023 vs October 2023 vs February 2024). No chief scientist, board members or public contact email/phone published.
Sources: https://techcrunch.com/2025/10/21/open-source-agentic-startup-langchain-hits-1-25b-valuation/; https://www.techtarget.com/searchenterpriseai/definition/LangChain; https://github.com/willshchang/wshc-tech-knowledge-base/blob/HEAD/03_Industry-Intelligence/Company-Portfolio/LangChain/langchain-research.md; https://pondero.ai/news/2026-09-25-langchain-managed-deep-agents-08/
52. MERCOR — SAN FRANCISCO, UNITED STATES
Founded 2023 · Independent (private)
Headquarters: San Francisco, United States
Offices and footprint: Headquarters as listed; additional offices not published in sources checked
Website: https://mercor.com
Contact: No published email or phone; official contact/sales route via https://mercor.com
Published email: Not published
Phone: Not published
Founders: Brendan Foody; Adarsh Hiremath; Surya Midha
Leadership: CEO / head: Brendan Foody - Co-Founder & CEO; Chief scientist / head of research: Not published; Other published C-suite / technical leadership: Adarsh Hiremath - Co-Founder & CTO (one secondary source styles him co-CEO - flagged); Surya Midha - Co-Founder (styled COO by one source, Board Chairman by another - flagged); Sundeep Jain - President (joined May 2025; ex-Uber Chief Product Officer); Board / advisers: Peter Fenton (Benchmark) - Board Member (per joinprospect, February 2025 - secondary); Larry Summers - Adviser; Dominic Barton - Adviser; Cass Sunstein - Adviser (per denotepress, 2026 - secondary); Founders: Brendan Foody; Adarsh Hiremath; Surya Midha
Ownership status: Independent (private)
Ownership and funding: Private. Series C: $350M at $10B valuation, October 28 2025, led by Felicis (Benchmark, General Catalyst, Robinhood Ventures; per Sacra and Wikipedia). Series B: $100M at $2B valuation, February 2025, led by Felicis. Series A: $30M at $250M valuation, September 2024, led by Benchmark. Total raised $486M (Sacra, 2025) / $492M (investclub, July 2026) - secondary. Investors: Felicis, Benchmark, General Catalyst, Robinhood Ventures, DST Global, Menlo Ventures, Peter Thiel, Jack Dorsey, Adam D'Angelo, Larry Summers. Profitable since inception (company claim). $2B annualized gross revenue, June 2026, per company claim via Forbes and Sacra (gross figure before contractor payouts; contractors receive 60-70%). $500M raise at $20B valuation reported in talks, July 2026 (unannounced - flagged).
What it builds: Mercor AI Talent Marketplace; Mercor Voice (AI interview system); APEX - AI Productivity Index (model benchmark); APEX-SWE (software-engineering benchmark, with Cognition); APEX-Agents report (January 2026)
Specialties: AI talent marketplace; expert data for RLHF and model training; agent evaluation benchmarks; AI-powered interviewing and vetting
Clients: Not disclosed in sources checked
What makes it distinctive: Went from $0 to $2B annualized gross revenue in about 24 months - among the fastest-scaling AI companies ever - while profitable since inception. The three Thiel-Fellow co-founders became the world's youngest self-made billionaires at age 22 (October 2025). Benefited from client flight from Scale AI after Meta's investment. Owns the full stack from a 300K+ human-expert network paying out $4M+/day to proprietary agent benchmarks (APEX).
Notes: Titles disputed: one source (siliconvalleyinvestclub, quoting Forbes) calls Hiremath co-CEO; Midha styled COO (joinprospect) vs Board Chairman (mintlify). Scale AI sued Mercor for alleged trade-secret misappropriation, September 2025 (reported). March 2026 LiteLLM supply-chain attack reportedly exposed up to 4TB of internal data and contractor PII, triggering class-action lawsuits and temporary client pauses incl. Meta (reported via Wikipedia - unverified by the company). Acquisitions: Sepal AI (February 2026), Deeptune (July 2026). Gross revenue figure is before contractor payouts. No chief scientist or public contact email/phone published.
Sources: https://en.wikipedia.org/wiki/Mercor; https://siliconvalleyinvestclub.com/companies/mercor/; https://sacra-pdfs.s3.us-east-2.amazonaws.com/mercor.pdf; https://techcrunch.com/2025/10/29/how-ai-labs-use-mercor-to-get-the-data-companies-wont-share/; https://denotepress.com/surya-midha-mercor/
NAMED GAPS
Poolside's chief scientist, board and any phone number are not published. DeepSeek and Moonshot AI publish no leadership beyond their founders and no contact details. Chinese labs' funding figures rest more heavily on press reporting, labelled as such in each entry. Only five of the 52 companies publish a contact email, and none publishes a public phone number — stated plainly here rather than filled in by guesswork.