Research

We track the AI landscape so we can pick the right tools for each job. This is the data we work from. Our AI search audit tools are free on GitHub.

Last updated 29 September 2026 · Funding figures reflect publicly disclosed or credibly reported data as of 27 September 2026.

We measure whole industries, and publish the results

We scan whole industries at our own cost and publish the numbers for anyone to check.

Method and corpus dates sit on every benchmark page. Aggregates only, never an individual business.

Comparison dashboard

Three views of the AI landscape. Toggle between valuation trajectories, total funding, and a quick-stats overview.

Valuation trajectories over time. Click any line or legend label to highlight. Google DeepMind & Meta AI shown at parent market cap. DeepSeek excluded (one external round so far, in mid 2026, at about $52B). xAI has been part of SpaceX since February 2026.

Total external equity raised, as of 27 September 2026. Google DeepMind, Meta AI, and Qwen excluded (funded by their parent companies). Debt and employee share sales are not counted.

OpenAI
$181.6B
Anthropic
~$130B
xAI
$37.1B
Kimi
$7.7B+
DeepSeek
~$7.4B
Mistral AI
~$6.6B
Perplexity
~$1.7B
OpenAI
Valuation$852B
Total Raised$181.6B
Key ModelGPT-6 Astra / Sol / Luna
Model AccessClosed
Revenue>$40B run-rate
Compute SpendNot disclosed
Anthropic
Valuation$965B
Total Raised~$130B
Key ModelClaude Fable 5.1 / Opus 5.5 / Sonnet 5.5
Model AccessClosed
Revenue>$65B run-rate
Compute Spend>$100B AWS deal (10 yrs)
Google DeepMind
Parent Cap~$4.2T
Key ModelGemini 3.1 Pro / 3.8 Flash
Model AccessClosed
Revenue~$99B Cloud run-rate
Compute Spend$195B to $205B CapEx (2026 guide)
Meta AI
Parent Cap~$1.9T
Key ModelMuse Spark 1.3 / Muse Glimmer
Model AccessMixed (Glimmer open weights)
RevenueIn $228B total rev. (TTM)
Compute Spend$130B to $145B CapEx (2026 guide)
xAI
Valuation$250B (at SpaceX deal)
Total Raised$37.1B
Key ModelGrok 4.7
Model AccessMostly closed
Revenue$3.2B (2025), loss-making
Compute SpendNot disclosed
DeepSeek
Valuation~$52B (round pending at ~$74B)
Total Raised~$7.4B (first round, 2026)
Key ModelDeepSeek V4-Pro / V4.1-Flash
Model AccessOpen Weights (MIT)
Revenue~$1B ARR (reported)
Compute Spend~$1.6B (2026, reported)
Qwen (Alibaba)
ParentAlibaba Group
Key ModelQwen3.8-Max
Model AccessOpen Weights (Apache 2.0)
Revenue~$28B Cloud run-rate
Compute Spend~$18B/yr CapEx (RMB 380B over 3 yrs)
Perplexity
Valuation$20B (talks above $30B)
Total Raised~$1.7B
Key ModelAgent API / Comet
Model AccessClosed
Revenue>$750M ARR
Compute SpendNot disclosed
Kimi (Moonshot AI)
Valuation$35B (in talks at ~$50B)
Total Raised$7.7B+
Key ModelKimi K3
Model AccessOpen Weights (Kimi K3 licence)
Revenue~$1B ARR (reported)
Compute SpendNot disclosed
Mistral AI
Valuation~$24B (€21B+)
Total Raised~$6.6B
Key ModelMistral Medium 3.5
Model AccessOpen Weights (Apache 2.0 / modified MIT)
Revenue~$1B ARR
Compute SpendNot disclosed

Figures as of 27 September 2026. Run-rate and ARR figures are company-stated or reported by Bloomberg, Reuters, the WSJ or The Information; "reported" marks a single-source figure. "Not disclosed" means no reliable public figure exists. Open Weights = model weights downloadable for use, but training data and code are not shared.

Cumulative funding raised

Total external equity raised over time. Click any line or legend label to highlight. Google DeepMind, Meta AI, and Qwen excluded (funded by their parent companies).

Company profiles

Click any card to expand. Use tabs to explore funding timelines and investor details.

Anthropic
Key model: Claude Fable 5.1 / Sonnet 5.5
Funding: ~$130B raised (PitchBook)
Model access: Closed

Safety-focused AI lab, and the most valuable private AI company at $965B after its May 2026 Series H ($65B, led by Altimeter, Dragoneer, Greenoaks, and Sequoia). It confidentially submitted a draft IPO registration to the SEC on 1 June 2026; no public prospectus had been filed by 29 September, and Reuters reported in September that it may raise up to $100B at around $2T. Founded by former OpenAI VP of Research Dario Amodei and President Daniela Amodei. Unique governance: a Long-Term Benefit Trust elects a growing share of the board. Amazon (about $13B invested) and Google (about $13B invested) are its largest strategic backers, and Google's stake carries no voting rights. Run-rate revenue passed $65B at the end of July 2026, up from $47B in May, driven heavily by Claude Code. Released Claude Fable 5 and Mythos 5 on 9 June 2026, Claude Fable 5.1 in September, and Claude Opus 5.5 on 22 September 2026.

May 2021Series A$124Mn/a
Apr 2022Series B$580Mn/a
Feb 2023Google$300Mn/a
May 2023Series C$450M$4.1B
Aug 2023SK Telecom$100Mn/a
Sep 2023Amazon I$1.25Bn/a
Oct 2023Google (up to)$2Bn/a
Dec 2023Series D (reported)$750M$18.4B
Mar 2024Amazon II$2.75Bn/a
Nov 2024Amazon III$4Bn/a
Jan 2025Google$1Bn/a
Mar 2025Series E$3.5B$61.5B
Sep 2025Series F$13B$183B
Nov 2025Microsoft + Nvidia (up to)$15B~$350B
Feb 2026Series G$30B$380B
Apr 2026Amazon (+ up to $20B more)$5Bn/a
Apr 2026Google (+ up to $30B more)$10B$350B
May 2026Series H$65B$965B

Rows overlap. Part of the Microsoft and Nvidia commitments sits inside Series G, and the April 2026 Amazon and Google money is part of the $15B of earlier commitments counted in Series H, so the rows add to more than the ~$130B total.

Lead Investors

Altimeter (co-led Series H) Dragoneer (co-led Series H) Greenoaks (co-led Series H) Sequoia (co-led Series H) GIC (led Series G) Coatue (led Series G) D1 Capital ICONIQ Capital Group Spark Capital

Strategic

Amazon (~$13B invested, up to $20B more) Google (~$13B invested, up to $30B more) Microsoft (up to $5B) Nvidia (up to $10B) SK Telecom Salesforce Ventures

Other

Menlo Ventures Lightspeed Fidelity D. E. Shaw MGX Temasek Goldman Sachs BlackRock Founders Fund Bessemer Insight Partners

Amazon and Google are the largest strategic investors. Court filings in March 2025 put Google's stake at about 14%, with no voting rights; later stakes are not public. No single investor holds a controlling interest.

OpenAI
Key model: GPT-6 Astra / Sol / Luna
Total raised: $181.6B (latest round $122B, Mar 2026)
Model access: Closed

Recapitalised in October 2025: the for-profit arm became a public benefit corporation, with the nonprofit OpenAI Foundation holding about 26% and Microsoft about 27%. It raised $122B at $852B post-money in March 2026, anchored by Amazon ($50B committed, $15B of it at close), Nvidia ($30B), and SoftBank ($30B). Anthropic overtook it at $965B in May 2026. A roughly $7B employee share sale in August 2026 held the $852B valuation. Board reshaped after the Nov 2023 crisis. Run-rate revenue passed $40B in August 2026 (Bloomberg), about double the end of 2025, and the company remains heavily loss-making on leaked figures. It announced a confidential IPO filing on 8 June 2026, but on 12 September Sam Altman said OpenAI will not go public in 2026. GPT-6 Astra, its current flagship, began rolling out on 3 September 2026, and the faster, cheaper GPT-6 Sol and Luna followed later in September.

2019Microsoft I$1Bn/a
2021Microsoft II$2Bn/a
Jan 2023Microsoft III$10B$29B
Oct 2024Funding round (Thrive-led)$6.6B$157B
Mar 2025SoftBank-led round$40B$300B
Oct 2025Employee share sale$6.6B$500B
Mar 2026Funding round$122B$852B
Aug 2026Employee share sale~$7B$852B

The $181.6B total counts primary equity only: Microsoft's $13B, the October 2024 round, the March 2025 SoftBank-led round, and the March 2026 round. Employee share sales move existing shares and are not counted. The March 2026 round was reported in February at $110B before closing at $122B, so it appears once.

Lead Investors

Amazon ($50B committed) Nvidia ($30B) SoftBank ($30B in 2026, led the 2025 round) Microsoft (~$13B, ~27% as of Oct 2025)

Other

a16z D. E. Shaw MGX TPG T. Rowe Price Thrive Capital Khosla Ventures
$852B Post-money
Employees & Earlier Investors~40.3% (~$343B)
Microsoft~23.1% (~$197B)
OpenAI Foundation~22.3% (~$190B)
March 2026 Round Investors~14.3% (~$122B)

Illustrative estimate. Built from OpenAI's October 2025 split (Foundation ~26%, Microsoft ~27%, others ~47%) diluted pro rata by the $122B March 2026 round. Not an official disclosure. See full cap table ↓

Google DeepMind
Key model: Gemini 3.1 Pro / 3.8 Flash
Parent market cap: ~$4.2T
Model access: Closed

One of the world's largest AI labs. Google Brain and DeepMind merged in 2023. Parent Alphabet was worth about $4.2T in late September 2026, among the largest public companies globally. Its compute budget is the largest in this group: Alphabet spent $91.4B on CapEx in 2025 and guides to $195B to $205B for 2026. Gemini runs across Google's products. Gemini 3.1 Pro (February 2026) is still the top Pro model, and Gemini 3.8 Flash shipped on 2 September 2026. Alphabet revenue was about $446B over the twelve months to June 2026. Google Cloud grew 82% to $24.8B in Q2 2026, about $99B a year at that pace, with a $514B backlog.

2010DeepMind foundedn/an/a
Jan 2014Acquired by Google~$500M (reported)n/a
Apr 2023Merged with Google Brainn/an/a
2024Alphabet CapEx$52.5Bn/a
2025Alphabet CapEx$91.4Bn/a
2026Alphabet CapEx (guidance)$195B to $205Bn/a

Parent Company

Alphabet Inc. (GOOGL): 100% ownership

Pre-Acquisition Investors

Founders Fund Horizons Ventures Elon Musk (angel)

Fully owned subsidiary of Alphabet since the 2014 acquisition. No external investors. All AI investment comes from Alphabet's balance sheet: $91.4B of CapEx in 2025 and $195B to $205B guided for 2026.

Meta AI
Key model: Muse Spark 1.3
Parent market cap: ~$1.9T
Model access: Mixed (Muse Glimmer open weights)

Built its name on open-weight Llama models, then moved its frontier work behind an API. Meta Superintelligence Labs, led by Chief AI Officer Alexandr Wang, released Muse Spark 1.3 on 2 September 2026 as a closed model, and Muse Glimmer, a 30B open-weight model under Apache 2.0, in August 2026. Meta launched Muse, a personal AI agent app, on 8 September 2026, and it reached No. 1 on the US App Store on 18 September. Yann LeCun left in November 2025 to found AMI Labs. Meta's revenue was about $228B over the twelve months to June 2026, with AI embedded across ad targeting and recommendations. CapEx was $72.2B in 2025 and is guided to $130B to $145B for 2026.

2013FAIR foundedn/an/a
Feb 2023LLaMA open releasen/an/a
2024Meta CapEx$39.2Bn/a
Jun 2025Scale AI (49% stake)$14.3Bn/a
2025Meta CapEx$72.2Bn/a
Aug 2026Muse Glimmer (open weights)n/an/a
Sep 2026Muse Spark 1.3 (closed)n/an/a
Sep 2026Muse agent app launchedn/an/a
2026Meta CapEx (guidance)$130B to $145Bn/a

Parent Company

Meta Platforms Inc. (META): 100% ownership

AI Investments

Scale AI ($14.3B for a 49% stake, Jun 2025)

Meta AI is a wholly owned division of Meta Platforms. All funding is via corporate R&D budget and CapEx ($72.2B in 2025, $130B to $145B guided for 2026). No external VC investors.

xAI
Key model: Grok 4.7
Valuation: $250B (SpaceX acquisition)
Model access: Mostly closed

Elon Musk's AI lab, founded 2023 and now operating as SpaceXAI. SpaceX acquired it in an all-stock deal on 2 February 2026 that valued xAI at $250B, a month after xAI closed a $20B Series E. xAI is now part of the listed SpaceX, which priced its Nasdaq IPO (SPCX) at $135 a share on 11 June 2026 and began trading on 12 June, raising $75B ($85.7B with the over-allotment) at about $1.75T. SpaceX was worth about $2T in late September 2026. Built the Colossus supercomputers in Memphis. Grok is integrated into X. SpaceX's IPO prospectus shows the AI segment earned $3.2B of revenue in 2025 against a $6.4B operating loss. Grok 4.7, released on 21 September 2026, is the current flagship. Weights for older models are public: Grok-1 under Apache 2.0, and Grok 2.5 under a more restrictive community licence.

Nov 2023Series A$134.7M~$0.7B
May 2024Series B$6B$24B
Dec 2024Series C$6B$50B
Mar 2025X merger (all-stock)n/a$80B (xAI)
Jun 2025Equity + debt (Morgan Stanley)$5B + $5B debtn/a
Jan 2026Series E$20B~$230B (reported)
Feb 2026SpaceX acquires xAI (all-stock)n/a$250B
Jun 2026SpaceX IPO (Nasdaq: SPCX)$75B raised~$1.75T parent

The $37.1B total counts equity rounds only; the $5B of June 2025 debt is excluded.

Lead Investors

Valor Equity Partners Nvidia a16z

Strategic

SpaceX (acquirer, now 100% owner) Tesla (~$2B, Series E) Cisco Investments

Other

Fidelity Qatar Investment Authority MGX (Abu Dhabi) StepStone Baron Capital

SpaceX acquired xAI in February 2026 at a $250B valuation, making it a wholly owned subsidiary. Previous xAI investors received SpaceX equity. Elon Musk maintains control through SpaceX.

DeepSeek
Key model: DeepSeek V4-Pro / V4.1-Flash
Funding: ~$7.4B (first external round, 2026)
Model access: Open Weights (MIT)

Open-weights lab from Hangzhou, China, founded in July 2023 by Liang Wenfeng, co-founder of the High-Flyer quantitative hedge fund. High-Flyer funded it alone until mid 2026, when DeepSeek closed its first external round: about RMB 50B ($7.4B) at a post-money valuation above RMB 350B (about $52B), according to a regulatory filing reported by Caixin. Liang reportedly contributed a large share, alongside Tencent and CATL, and keeps control. On 24 September 2026 Reuters, citing The Information, reported that DeepSeek's annualised revenue run rate had reached $1B and that it is seeking a second round of RMB 50B at a RMB 500B valuation (about $74B), targeted to close by the end of October. That round had not closed at the time of writing. DeepSeek-V3 (Dec 2024) uses a 671B-parameter Mixture-of-Experts architecture with 37B active per token; its final training run reportedly cost about $5.6M. DeepSeek-R1 (Jan 2025) matched leading reasoning models of the time and helped trigger a sell-off in which Nvidia lost about $589B of market value in one day. V4-Pro reached general availability in August 2026 and V4.1-Flash followed on 10 September 2026, both with MIT-licensed weights. API prices remain far below US frontier models. Operates under Chinese AI rules and US export controls on advanced Nvidia chips. Controversies include questions about training on other labs' model outputs and national security concerns raised by US lawmakers.

Jul 2023Founded by Liang Wenfengn/an/a
Nov 2023DeepSeek Coder, then DeepSeek LLM 67Bn/an/a
May 2024DeepSeek-V2 (MoE)n/an/a
Dec 2024DeepSeek-V3 (671B MoE)~$5.6M final runn/a
Jan 2025R1 release (top App Store)n/an/a
Apr 2026DeepSeek-V4 previewn/an/a
Jul 2026First external round (filing)~$7.4B~$52B
Aug 2026V4-Pro general availabilityn/an/a
Sep 2026V4.1-Flash (MIT weights)n/an/a

R1 became the #1 free app on Apple's App Store within days of launch. On 27 January 2025 Nvidia lost about $589B of market value, the largest one-day loss for any company at the time.

Founder and Backers

Liang Wenfeng (founder, CEO and controller) High-Flyer (founding backer)

2026 Round

Tencent CATL Others reported: JD.com, NetEase, IDG Capital, National AI Industry Investment Fund

Most of the 2026 round reportedly came through a partnership managed by Liang, and most outside investors reportedly have no voting rights. Individual amounts are not in a public filing, so we do not list them.

Qwen (Alibaba)
Key model: Qwen3.8-Max
AI capex plan: RMB 380B (~$53B, 3 yrs)
Model access: Open Weights (Apache 2.0)

Alibaba's model family. Since June 2026 Qwen has sat in Alibaba's Token Foundry division, led by group CEO Eddie Wu. Zhou Jingren, who led Qwen's development, handed the Alibaba Cloud CTO role to Li Feifei in April 2026 and is now Alibaba's Chief Scientist. Qwen3.8-Max, the current flagship, went on sale through Alibaba Cloud on 3 August 2026 and its open weights followed in mid August: a 2.4T-parameter Mixture-of-Experts model with 95B active parameters and a 1M-token context. The Qwen series is among the most downloaded open-weight families globally. Alibaba's AI cloud and compute revenue grew 45% to RMB 48.4B (about $7.1B) in the June 2026 quarter, about $28B a year at that pace. Alibaba has committed RMB 380B (about $53B) over three years to AI and cloud infrastructure and had spent RMB 190B of it by June 2026.

Aug 2023Qwen-7B first releasen/an/a
Jun 2024Qwen2 series (Apache 2.0)n/an/a
Sep 2024Qwen2.5 (7 sizes, 0.5B to 72B)n/an/a
Nov 2024QwQ reasoning modeln/an/a
Feb 2025AI and cloud capex planRMB 380B (3 yr)n/a
Feb 2026Qwen3.5 (397B open weights)n/an/a
Jun 2026Qwen moves into Token Foundry divisionn/an/a
Aug 2026Qwen3.8-Max (2.4T MoE, open weights)n/an/a

Qwen is funded by Alibaba. The RMB 380B plan covers 2025 to 2028; half had been spent by the June 2026 quarter.

Parent Company

Alibaba Group (BABA/9988.HK): 100% ownership

Division

Token Foundry (since June 2026) Alibaba Cloud Intelligence (infrastructure)

Qwen is wholly owned by Alibaba, which is spending RMB 380B (about $53B) over three years on AI and cloud infrastructure. No external VC investors.

Perplexity
Key product: Agent API / Comet
Funding: ~$1.7B across 8 rounds
Model access: Closed

AI-native answer engine, now also selling the Comet browser and an Agent API that replaces its Sonar API. Last confirmed valuation $20B (September 2025). In August 2026 Reuters reported that Nvidia was in talks to invest in a new round valuing it above $30B. Founded in 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski. Annualised revenue passed $750M by August 2026, per the same reports. Faces copyright suits from publishers including Dow Jones, The New York Times, and Encyclopaedia Britannica, plus disputes with Reddit and Amazon.

2023Series A$25.6Mn/a
Jan 2024Series B$73.6M$520M
Apr 2024Series B-1$62.7M$1.04B
Jun 2024Series C$250M$3B
Dec 2024Series D (IVP)$500M$9B
May 2025Series E (Accel)$500M$14B
Jul 2025Extension$100M$18B
Sep 2025Funding round$200M$20B
Aug 2026Nvidia talks (reported)n/a>$30B

Lead Investors

SoftBank Vision Fund 2 IVP Accel NEA

Strategic

Nvidia Jeff Bezos (Bezos Expeditions)

Other

Databricks Ventures Tobi Lütke Elad Gil

VC-backed with no single controlling investor. About $1.7B raised across 8 rounds. The founders retain significant equity.

Kimi (Moonshot AI)
Key model: Kimi K3
Funding: $7.7B+ across the 5 rounds listed
Model access: Open Weights (Kimi K3 licence)

Beijing-based lab founded by Yang Zhilin, known for long-context models. It closed a $3.5B round at a $35B valuation in late July 2026, led by China's National AI Industry Investment Fund, after raising about $2B at $20B in May and $700M+ at $10B in February. Kimi K3, released on 16 July 2026, is a 2.8T-parameter Mixture-of-Experts model with a 1M-token context, and Moonshot published its weights in late July under its own Kimi K3 licence. Backed by Alibaba, Tencent, and Meituan. Anthropic named Moonshot in a February 2026 report on distillation of Claude outputs. Annualised revenue reportedly passed $1B in August 2026, up from about $300M in June. On 3 September 2026 Reuters reported that Moonshot had filed confidentially for a Hong Kong IPO and was valued at about $50B in a funding round still under way.

Feb 2024Alibaba-led round$1B$2.5B
Late 2025Funding round$500M$4.3B
Feb 2026Funding round$700M+$10B
May 2026Meituan-led round~$2B$20B+
Jul 2026National AI Industry Investment Fund-led$3.5B$35B

Earlier 2023 rounds are not publicly documented, so the $7.7B+ total counts only the rounds listed.

Lead Investors

National AI Industry Investment Fund (led Jul 2026) Alibaba (led Feb 2024 round) Meituan (led May 2026 round)

Other

Tencent IDG Capital China Mobile 5Y Capital HongShan CPE Tsinghua Capital

Yang Zhilin (founder) retains significant equity. Unusually broad Chinese backing, with Alibaba, Tencent, and a state AI fund all on the cap table.

Mistral AI
Key model: Mistral Medium 3.5
Funding: ~$6.6B equity across 6 rounds
Model access: Open Weights (Apache 2.0 / modified MIT)

Europe's AI champion, with a French sovereignty angle. Raised a €3B Series D at more than €21B (about $24B) post-money on 8 September 2026, led by Samsung Electronics with EQT's Scaleup Europe Fund and PSG Equity as co-leads, the largest equity round for a European tech company. ASML led the September 2025 Series C. Secured €722M ($830M) of debt in March 2026 for its first large data centre near Paris. Shipped Voxtral TTS (speech) in March and Mistral Medium 3.5 (open weights, modified MIT licence, 256K context) in April 2026; Mistral Large 3 (December 2025) is Apache 2.0. EU AI Act compliance built in. Its CFO told The Wall Street Journal on 8 September 2026 that annual recurring revenue was slightly above $1B. Reuters quoted him the same day as saying Mistral was on track for $1B by the end of the year.

Jun 2023Seed€105Mn/a
Dec 2023Series A€385M~$2B
Feb 2024Microsoft€15Mn/a
Jun 2024Series B (incl. €132M debt)€600M€5.8B
Sep 2025Series C (ASML-led)€1.7B€11.7B
Mar 2026Debt (data centre)€722Mn/a
Sep 2026Series D (Samsung-led)€3B€21B+

The ~$6.6B total converts equity only (about €5.7B) at the rate on each round's date; the 2024 and 2026 debt is excluded.

Lead Investors

Samsung Electronics (led Series D) ASML (led Series C, ~11%) EQT Scaleup Europe Fund PSG Equity a16z General Catalyst Lightspeed

Strategic

Microsoft (minority stake) Nvidia Salesforce Ventures

Other

Bpifrance DST Global BNP Paribas BlackRock Advent Grand Duchy of Luxembourg

Co-founders Arthur Mensch, Timothée Lacroix, and Guillaume Lample retain significant equity. Bpifrance (French state investment bank) is a key backer supporting European AI sovereignty. About $6.6B of equity raised across 6 rounds, plus €722M of data centre debt.

OpenAI cap table: who owns the $852B

OpenAI published its headline ownership split when it recapitalised in October 2025: the OpenAI Foundation holds about 26%, Microsoft about 27%, and employees and other investors the remaining 47% or so. The March 2026 round then added $122B of new equity at $852B post-money. OpenAI has not published the split after that round, so the illustrative columns below assume the whole $122B was new equity that diluted every earlier holder equally.

$852B Post-money valuation
Employees & Earlier Investors~40.3%
Microsoft~23.1%
OpenAI Foundation~22.3%
March 2026 Round Investors~14.3%
Shareholder Reported Stake (Oct 2025) Illustrative Stake (after Mar 2026) Illustrative Value at $852B Invested (reported) Source
Nonprofit Foundation: ~22.3% (~$190B)
The OpenAI Foundation controls the public benefit corporation and holds equity in it. This block does not represent a cash investment.
OpenAI Foundation~26%~22.3%~$190BNo cash investmentOpenAI, Oct 2025
Microsoft: ~23.1% (~$197B)
OpenAI's longest-standing strategic backer, investing since 2019.
Microsoft~27%~23.1%~$197B~$13BOpenAI and Microsoft, Oct 2025
Employees and Earlier Investors: ~40.3% (~$343B)
Current and former staff plus every investor before 2026, including SoftBank (which led the 2025 round), Thrive Capital, Khosla Ventures, and a16z. OpenAI has not published how this block divides, and CEO Sam Altman's holding is not publicly disclosed.
Employees, founders and pre-2026 investors~47%~40.3%~$343Bn/aOpenAI, Oct 2025
March 2026 Round Investors: ~14.3% (~$122B)
The $122B round closed on 31 March 2026. Commitments by investor are below; how the new shares split between them has not been published.
Amazonn/aPart of ~14.3%n/a$50B committed ($15B at close)Reported, Mar 2026
Nvidian/aPart of ~14.3%n/a$30BReported, Mar 2026
SoftBank GroupWithin ~47%Part of ~14.3%n/a$30B in this roundReported, Mar 2026
a16z, D. E. Shaw, MGX, TPG, T. Rowe Price and othersn/aPart of ~14.3%n/a~$12B combinedTechCrunch, Mar 2026
Grand Total ~100% 100.0% $852.0B

Illustrative estimate. Only the October 2025 split (Foundation about 26%, Microsoft about 27%, others about 47%) comes from OpenAI's own announcement. The illustrative columns apply pro-rata dilution from the $122B March 2026 round and assume every commitment is funded; about $35B of Amazon's commitment depends on conditions. Figures will change when OpenAI files a public prospectus. This table no longer carries an investor-by-investor reconstruction, because most of its rows could not be traced to a public source.

At a glance

This is the landscape we navigate when picking tools for each project. No single model wins everything.

Company Key Model Model Access Valuation Total Raised Revenue Compute Spend
AnthropicClaude Fable 5.1 / Opus 5.5 / Sonnet 5.5Closed$965B~$130B>$65B run-rate>$100B AWS deal (10 yrs)
OpenAIGPT-6 Astra / Sol / LunaClosed$852B$181.6B>$40B run-rateNot disclosed
Google DeepMindGemini 3.1 Pro / 3.8 FlashClosed~$4.2T *n/a~$99B Cloud run-rate$195B to $205B (2026 guide)
Meta AIMuse Spark 1.3 / Muse GlimmerMixed~$1.9T *n/aIn $228B rev. (TTM)$130B to $145B (2026 guide)
xAIGrok 4.7Mostly closed$250B (SpaceX ~$2T *)$37.1B$3.2B (2025)Not disclosed
DeepSeekDeepSeek V4-Pro / V4.1-FlashOpen Weights (MIT)~$52B (round pending at ~$74B)~$7.4B~$1B ARR (reported)~$1.6B (2026, reported)
QwenQwen3.8-MaxOpen Weights (Apache 2.0)n/a *n/a~$28B Cloud run-rate~$18B/yr
PerplexityAgent API / CometClosed$20B~$1.7B>$750M ARRNot disclosed
KimiKimi K3Open Weights (Kimi K3 licence)$35B (in talks at ~$50B)$7.7B+~$1B ARR (reported)Not disclosed
MistralMistral Medium 3.5Open Weights (Apache 2.0 / modified MIT)~$24B~$6.6B~$1B ARRNot disclosed

* Parent company market cap in late September 2026. The AI division has no standalone valuation. Figures as of 27 September 2026 and based on company announcements, filings, and reporting by Bloomberg, Reuters, CNBC, the WSJ and The Information. Compute spend shows CapEx or disclosed commitments where they exist. Open Weights = model weights downloadable, but training data and code not shared.