Research

We track the AI landscape so we can pick the right tools for each job. This is the data we work from.

Last updated 12 June 2026 · Funding figures reflect publicly disclosed data as of publication.

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 (self-funded, no external valuation).

Total external capital raised. Google DeepMind, Meta AI, DeepSeek, and Qwen excluded (internal or self-funded).

OpenAI
$189.6B
Anthropic
$132.3B
xAI
$36.1B
Kimi
$4.6B
Mistral AI
$3.1B
Perplexity
$1.7B
OpenAI
Valuation$852B
Total Raised$189.6B
Key ModelGPT-5.5
Model AccessClosed
Revenue~$25B ARR
Compute Spend~$7B+/yr est.
Anthropic
Valuation$965B
Total Raised$132.3B
Key ModelClaude Fable 5 / Opus 4.8
Model AccessClosed
Revenue~$47B run-rate
Compute Spend~$3B+/yr est.
Google DeepMind
Parent Cap~$4.3T
Key ModelGemini 3.5 Flash / 3.1 Pro
Model AccessClosed
Revenue~$44B Cloud ARR
Compute Spend~$75B+/yr CapEx
Meta AI
Parent Cap~$1.4T
Key ModelLLaMA 4
Model AccessOpen Weights
RevenueIn $164B total rev.
Compute Spend~$60B+/yr CapEx
xAI
Valuation$250B
Total Raised$36.1B
Key ModelGrok 4.1 / 4.20
Model AccessPartial (Grok)
Revenue~$3.2B (2025), loss-making
Compute Spend~$3B+/yr est.
DeepSeek
ValuationSelf-funded
Total Raised$0 VC
Key ModelDeepSeek V4 (preview)
Model AccessOpen Weights (MIT)
RevenueMinimal (low-cost API)
Compute Spend~$1B+/yr est.
Qwen (Alibaba)
ParentAlibaba Cloud
Key ModelQwen 3.5
Model AccessOpen Weights (Apache 2.0)
Revenue~$16B Cloud ARR
Compute Spend~$21B/yr CapEx
Perplexity
Valuation$21B
Total Raised$1.7B
Key Modelpplx-api
Model AccessClosed
Revenue~$500M ARR
Compute SpendNot disclosed
Kimi (Moonshot AI)
Valuation$20B+
Total Raised$4.6B+
Key ModelKimi 2.5
Model AccessOpen Weights
Revenue~$200M ARR
Compute SpendNot disclosed
Mistral AI
Valuation$14B
Total Raised$3.1B
Key ModelMistral Medium 3.5
Model AccessOpen Weights (Apache 2.0)
Revenue~$100M ARR
Compute Spend~€240M/yr est.

Revenue and compute estimates are based on publicly reported data. "Not disclosed" means no public figures are available. Open Weights = model weights downloadable for use, but training data and code are not shared.

Cumulative funding raised

Total external capital raised over time for VC-funded companies. Click any line or legend label to highlight. Google DeepMind, Meta AI, DeepSeek, and Qwen excluded (internal/self-funded).

Company profiles

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

Anthropic
Key model: Claude Fable 5
Funding: $132.3B across 18 rounds
Model access: Closed

Safety-focused AI lab, now the most valuable private AI company at $965B (May 2026 Series H, $65B round co-led by Altimeter, Dragoneer, Greenoaks, and Sequoia). Confidentially filed for an IPO on 1 June 2026. Founded by former OpenAI VP of Research Dario Amodei and President Daniela Amodei. Unique governance: Long-Term Benefit Trust with escalating board control. Amazon ($8B+) and Google ($2B+) are early investors with zero voting rights. Revenue run-rate crossed ~$47B in May 2026, driven heavily by enterprise Claude Code adoption. Released Claude Opus 4.8 in late May and Claude Fable 5 (its first publicly available Mythos-class model) on 9 June 2026.

May 2021Series A$124M
Apr 2022Series B$580M$4B
Feb 2023Google Strategic$300M$4.6B
May 2023Series C$450M$4.1B
Aug 2023SK Telecom$100M$4.1B
Sep 2023Amazon I$1.25B
Dec 2023Series D$750M$18.4B
Mar 2024Amazon II$2.75B
Jun 2024Series E$2B$18.4B
Nov 2024Amazon III$4B
Jan 2025Series D ext.$2B$60B
Mar 2025Series E$3.5B$61.5B
Jul 2025Google IV$1B
Sep 2025Series F$13B$183B
Nov 2025MSFT/NVDA$5.5B
Feb 2026Series G$30B$380B
May 2026Series H$65B$965B

Lead Investors

Amazon ($8B+ — largest external investor) Google ($2B+ — ~10% est. stake) Altimeter (co-led Series H) Dragoneer (co-led Series H) Greenoaks (co-led Series H) Sequoia (co-led Series H) GIC Coatue D. E. Shaw Spark Capital

Strategic

SK Telecom Microsoft (minority stake) Nvidia Salesforce

Other

Menlo Ventures Lightspeed Sequoia Fidelity Goldman Sachs BlackRock Founders Fund Bessemer Insight Partners

Amazon is the largest external investor at $8B+. Google's $2B+ investment gives it roughly a 10% stake. No single investor holds a controlling interest.

OpenAI
Key model: GPT-5.5
Total raised: $189.6B (latest round $122B, Mar 2026)
Model access: Closed

Pivoted from nonprofit to public benefit corporation. $852B post-money valuation (Mar 2026) — overtaken by Anthropic's $965B in May 2026. Amazon ($50B), Nvidia ($30B), and SoftBank ($30B) led the latest round. Microsoft holds ~27%. Board reshaped after the Nov 2023 crisis. ARR crossed ~$25B in March 2026 with ~900M weekly ChatGPT users, projecting ~$29B for 2026 — though still operating at a significant loss (reported −122% non-GAAP operating margin in Q1 2026). GPT-5.5 released April 2026.

2019Microsoft I$1B
Jan 2023Microsoft II$10B$29B
Oct 2024Series A (PBC)$6.6B$157B
Mar 2025SoftBank$40B$300B
Feb 2026Series B$110B$840B
Mar 2026Series B ext.$122B$852B

Lead Investors

Amazon ($50B) Nvidia ($30B) SoftBank ($30B+) Microsoft ($10B+ — ~27% equity pre-restructure)

Other

a16z D. E. Shaw MGX TPG T. Rowe Price Thrive Capital Tiger Global Khosla Ventures
$852B Post-money
Strategic Corporates46.58% — $396.9B
Nonprofit Foundation25.80% — $219.8B
Employees & Founders19.35% — $164.9B
VC & Institutional7.83% — $63.4B
Individual / Retail0.44% — $3.4B

Estimated / Reconstructed. Not an official disclosure. See full cap table ↓

Google DeepMind
Key model: Gemini 3.5 Flash / 3.1 Pro
Parent market cap: ~$4.3T
Model access: Closed

World's largest AI lab. Merged Google Brain + DeepMind in 2023. Parent Alphabet's market cap reached ~$4.3T by June 2026, among the largest public companies globally. Massive compute infrastructure advantage with $75B+ CapEx planned for 2025. Gemini models integrated across Google products — Gemini 3.5 Flash shipped May 2026 alongside the 3.1 Pro flagship. Revenue (TTM) ~$403B, net income ~$132B. Google Cloud (including AI/ML services) reached ~$44B ARR, the fastest-growing segment.

2010DeepMind founded
Jan 2014Acquired by Google$500M
Apr 2023Merged with Google Brain
2024Alphabet AI CapEx$52.5B
2025Alphabet AI CapEx$75B+

Parent Company

Alphabet Inc. (GOOGL) — 100% ownership

Pre-Acquisition Investors

Founders Fund Horizons Ventures Elon Musk (angel)

Fully owned subsidiary of Alphabet since the $500M acquisition in 2014. No external investors. All AI investment ($75B+ CapEx in 2025) comes from Alphabet's balance sheet.

Meta AI
Key model: LLaMA 4
Parent market cap: ~$1.4T
Model access: Open Weights

Open-weights AI leader. LLaMA models are the most widely used open-weight foundation models. Yann LeCun departed Nov 2025 to found AMI Labs after strategic disagreements with Zuckerberg over LLM direction. Meta Superintelligence Labs now led by Alexandr Wang (Scale AI CEO). The open-weights strategy creates ecosystem lock-in without licensing revenue. Meta's total revenue ~$164B (TTM), with AI embedded across its ad targeting and recommendation systems. AI CapEx budget of $60B+ in 2025.

2013FAIR founded
Feb 2023LLaMA open release
2024Meta AI CapEx$37B
Jun 2025Scale AI ($14.3B)$14.3B
2025Meta AI CapEx$60B+

Parent Company

Meta Platforms Inc. (META) — 100% ownership

AI Acquisitions

Scale AI ($14.3B, Jun 2025)

Meta AI is a wholly owned division of Meta Platforms. All funding is via corporate R&D budget and CapEx ($60B+ in 2025). No external VC investors.

xAI
Key model: Grok 4.1 / 4.20
Valuation: $250B (SpaceX acquisition)
Model access: Partial (Grok)

Elon Musk's AI lab. Founded 2023. Acquired by SpaceX in Feb 2026 at $250B valuation after raising $20B Series E in Jan 2026 — and now part of the listed SpaceX, which began trading on Nasdaq (SPCX) on 12 June 2026 at a ~$1.77T IPO valuation. Built the Memphis Supercluster (100K H100 GPUs). Grok integrated into X (Twitter) platform. "Any lawful use" military stance contrasts with Anthropic's safety focus. The AI unit generated ~$3.2B revenue in 2025 against a ~$6.4B operating loss, per SpaceX's IPO prospectus. Grok-1 was open-sourced, but later models are closed.

Dec 2023Initial raise$134M~$1B
May 2024Series B$6B$18B
Mar 2025X acquisition (all-stock)$113B
Sep 2025Series D$10B$200B
Jan 2026Series E$20B$230B
Feb 2026SpaceX acquires xAI$250B
Jun 2026SpaceX IPO (Nasdaq: SPCX)~$75B raised$1.77T parent

Lead Investors

Nvidia Valor Equity a16z

Strategic

SpaceX (acquirer — now 100% owner) Tesla (~$2B) Cisco

Other

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

SpaceX acquired xAI in Feb 2026 at $250B valuation, making it a wholly owned subsidiary. Previous VC investors were bought out or converted to SpaceX equity. Elon Musk maintains control through SpaceX.

DeepSeek
Key model: DeepSeek V4 (preview)
Funding: Self-funded
Model access: Open Weights (MIT)

Self-funded open-weights lab from Hangzhou, China. ~200 employees. Founded July 2023 by Liang Wenfeng, co-founder of High-Flyer quantitative hedge fund. DeepSeek-V3 (Dec 2024) uses a 671B-parameter Mixture-of-Experts architecture with only 37B active per token. DeepSeek-R1 (Jan 2025) matched GPT-4o on reasoning benchmarks at a reported training cost of ~$5.6M — a fraction of Western equivalents — triggering a $1T market-cap sell-off in US tech stocks. MIT-licensed models. No external VC. Operates under Chinese AI regulations and US export controls restricting access to advanced Nvidia chips. Estimated $1B+ annual compute spend funded entirely from High-Flyer's trading profits. Minimal direct revenue — the API is priced 90-95% below Western competitors. Controversies include questions around data provenance (potential training on proprietary model outputs) and national security concerns raised by US lawmakers.

Jul 2023Founded by Liang Wenfeng
Nov 2023DeepSeek-67B (first model)
May 2024DeepSeek-V2 (MoE breakthrough)
Dec 2024DeepSeek-V3 (671B MoE)~$5.6M train
Jan 2025R1 release (top App Store)
Q2 2026DeepSeek-V4 Preview (1M context)
OngoingSelf-funded via High-Flyer$0 VC

DeepSeek has never taken external venture funding. All capital comes from High-Flyer's quantitative trading profits. R1 became the #1 free app on Apple's App Store within days of launch, triggering a ~$1T sell-off in US tech stocks.

Sole Backer

High-Flyer (幻方量化) — 100% ownership

Founder

Liang Wenfeng (CEO & sole controller)

No external VC, PE, or strategic investors. DeepSeek is entirely self-funded through High-Flyer's quant trading profits. Estimated $1B+ annual compute spend funded from trading revenue.

Qwen (Alibaba)
Key model: Qwen 3.5
Internal investment: $63B
Model access: Open Weights (Apache 2.0)

Alibaba Cloud's AI model division, led by Zhou Jingren (Alibaba Cloud CTO, Ph.D. Columbia, ex-Microsoft Research). Qwen3.5, released on the eve of Chinese New Year 2026, is a 397B-parameter open-weight model with self-reported benchmarks on par with leading closed models; the Qwen series (300+ models, 140,000+ derivatives) is among the most downloaded open-weight families globally. Zhou was elevated to Alibaba's core management committee in 2025, signalling the company's deepening AI commitment. Free API tier available. Alibaba Cloud revenue ~$16B ARR, with AI workloads driving growth. $63B CapEx commitment over 3 years for AI infrastructure (~$21B/yr).

Aug 2023Qwen-7B first release
Jun 2024Qwen2 series (Apache 2.0)
Sep 2024Qwen2.5 (7 sizes, 0.5B-72B)
Nov 2024QwQ reasoning model
2025Zhou Jingren joins mgmt committee
2025-28Alibaba Cloud AI CapEx$63B (3yr)

Qwen is an internal division of Alibaba Cloud. All funding comes from Alibaba's corporate R&D budget ($63B committed over 3 years). Zhou Jingren, the division head, was promoted to Alibaba's core management committee in 2025.

Parent Company

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

Division

Alibaba Cloud Intelligence

Qwen is a wholly owned division of Alibaba Cloud. Alibaba is committing $63B over 3 years to AI infrastructure. No external VC investors.

Perplexity
Key model: pplx-api
Funding: $1.7B across 11 rounds
Model access: Closed

AI-native search engine. $21B valuation (early 2026). Founded by former Google/DeepMind researchers. ARR reached ~$500M by June 2026 — roughly 5x in a year on a barely larger team — against an internal target of $656M by year-end, helped by a pivot into AI agents and usage-based pricing. 780M+ monthly queries. Ongoing copyright disputes with publishers over content usage.

2023Seed rounds$25.6M$150M
Jan 2024Series B$73M$520M
Apr 2024Series B+$165M$1B
Jun 2024Series C$250M$3B
Dec 2024Series D$500M$9B
Jun 2025Series E$500M$14B
Sep 2025Series E-2$200M$20B
Early 2026Series E-6$21.2B

Lead Investors

SoftBank IVP Accel

Strategic

Nvidia Jeff Bezos (personal investment)

Other

NEA Databricks Ventures Tobi Lütke Elad Gil Kindred Ventures

VC-backed with no single controlling investor. $1.7B raised across 11 rounds. Founders Aravind Srinivas and Denis Yarats retain significant equity. SoftBank and IVP are the largest institutional investors.

Kimi (Moonshot AI)
Key model: Kimi 2.5
Funding: $4.6B+ across 6+ rounds
Model access: Open Weights

Beijing-based lab. Known for long-context capabilities. Raised a ~$2B Meituan-led round at a $20B+ valuation in May 2026 — up from $10B in February — and is reportedly now targeting up to $30B as China's AI race intensifies. Backed by Alibaba, Tencent, Meituan, and 5Y Capital. Accused of training on Claude outputs. The fastest valuation growth among Chinese AI startups. ARR passed $200M in April 2026, driven by Kimi subscriptions and enterprise model services.

Jun 2023Angel$300M
Feb 2024Series A (Alibaba-led)$1B$2.5B
Late 2025Series C$500M$4.3B
Feb 2026Series D$700M+$10B
May 2026Meituan-led round~$2B$20B+
Jun 2026Reportedly seeking new raise$1–2B targetup to $30B

Lead Investors

Alibaba (led Series A — largest investor) Meituan (led May 2026 round) Tencent 5Y Capital IDG Capital

Other

Meituan Andon HK HongShan

Alibaba led the $1B Series A and is the largest institutional investor. Yang Zhilin (founder) retains significant equity. Unusually broad Chinese tech backing with both Alibaba and Tencent as investors.

Mistral AI
Key model: Mistral Medium 3.5
Funding: $3.1B across 8 rounds
Model access: Open Weights (Apache 2.0)

Europe's AI champion. French sovereignty angle. ~862 employees. $14B valuation (Sep 2025). Secured €722M ($830M) debt financing in Mar 2026 for its first large-scale data centre near Paris with 13,800 Nvidia GB300 GPUs. Shipped Voxtral TTS (open-weight speech model) in March and Mistral Medium 3.5 (open weights, 256K context) in April 2026. EU AI Act compliance built in. Strong enterprise adoption in Europe. Estimated ~$100M ARR from its La Plateforme API and enterprise contracts.

Jun 2023Seed€105M
Dec 2023Series A$415M$2B
Jun 2024Series B€600M$6B
Sep 2025Series C$13.7B
Mar 2026Debt (data centre)€722M

Lead Investors

Bpifrance (French sovereign fund) a16z General Catalyst Lightspeed

Strategic

Microsoft (minority stake) Nvidia Samsung Salesforce

Other

DST Global BNP Paribas HSBC Natixis

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. Microsoft took a minority stake. $3.1B raised across 8 rounds plus €722M in debt financing.

OpenAI cap table — who owns the $852B

A reconstructed ownership breakdown of one of the two most valuable AI startups (alongside Anthropic at $965B), based on publicly reported funding rounds, investor disclosures, and secondary-market data as of March 2026.

$852B Post-money valuation
Strategic Corporates46.58%
Nonprofit Foundation25.80%
Employees & Founders19.35%
VC & Institutional7.83%
Individual / Retail0.44%
Shareholder % Ownership Value at $852B Est. Cost Basis Unrealized Gain Return Multiple
Nonprofit Foundation & Governance — 25.80% ($219.8B)
The OpenAI Foundation retains a controlling governance stake to ensure the company remains aligned with its public-benefit mission. This block does not represent a cash investment.
OpenAI Foundation25.80%$219.8B$0.0$219.8B
Strategic Corporate Investors — 46.58% ($396.9B)
Large technology companies that invested for both financial returns and strategic access to OpenAI’s models and infrastructure. Microsoft is by far the largest, having backed OpenAI since 2019.
Microsoft26.79%$228.3B$13.0B$215.3B17.6x
SoftBank Group11.66%$99.3B$69.3B$34.7B1.5x
Amazon.com, Inc4.66%$39.7B$15.0B$24.7B2.6x
NVIDIA Corporation3.47%$29.6B$30.1B-$0.5B1.0x
Employees, Founders & Management — 19.35% ($164.9B)
Equity held by current and former staff through stock grants, options, and secondary sales. CEO Sam Altman’s stake was disclosed as pending restructuring terms.
Current Employees (Pool)15.88%$135.3B~$0~$135B
Former Employees / Alumni3.47%$29.6B~$0~$30B
Sam Altman — CEOStake not publicly disclosed — reportedly under restructuring as part of the PBC conversion. Figures will update when OpenAI files them.
Venture Capital & Institutional Investors — 7.83% ($63.4B)
Professional venture capital, growth equity, and institutional funds. Many entered at earlier (lower) valuations, producing outsized returns. Thrive Capital led the October 2024 PBC round. Khosla Ventures and Sound Ventures were among the earliest backers.
Thrive Capital1.98%$16.9B$3.5B$13.4B4.8x
Andreessen Horowitz (a16z)0.79%$6.8B$2.5B$4.3B2.7x
MGX (Mubadala)0.69%$5.9B$1.5B$4.4B3.9x
D.E. Shaw Ventures0.50%$4.2B$2.0B$2.2B2.1x
TPG Inc.0.45%$3.8B$1.5B$2.3B2.5x
T. Rowe Price Associates0.40%$3.4B$1.0B$2.4B3.4x
Sequoia Capital0.35%$3.0B$1.0B$2.0B3.0x
Blackstone / BlackRock0.35%$3.0B$1.0B$2.0B3.0x
Coatue Management0.30%$2.5B$0.8B$1.7B3.1x
Dragoneer Investment Group0.25%$2.1B$2.8B-$0.7B0.75x
Altimeter Capital0.20%$1.7B$0.6B$1.1B2.8x
Temasek Holdings0.20%$1.7B$1.7B$0.01.0x
Khosla Ventures0.18%$1.5B$0.05B$1.5B30x
Sound Ventures0.15%$1.3B$0.03B~43x
Insight Partners0.15%$1.3B
UC Investments (Univ. of California)0.12%$1.0B
Fidelity Mgmt. & Research Co.0.12%$1.0B
D1/Apolus/Sands/Whale/Goanna/Paragon0.48%$4.1B
Early Angels — Hamid Diboud0.17%$1.4B$0.01B$1.4B~140x
Individual Investors, Retail & Secondary Market — 0.44% ($3.4B)
Smaller allocations available through bank-brokered channels during Series G, plus publicly traded ETF exposure via ARK Invest funds.
Individual Investors (Bank Channel — Series G)0.35%$3.0B$3.0B1.0x
ARK Invest ETFs (ARKK/ARKQ/ARK)0.09%$0.4B$0.4B$0.01.0x
Grand Total 100.0% $852.0B

Estimated / Reconstructed. Not an official disclosure. Cost basis figures are best estimates from public sources. Percentages are approximate and may not sum to exactly 100% due to rounding. Data compiled from SEC filings, press releases, and credible financial reporting as of March 2026.

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 5Closed$965B$132.3B~$47B run-rate~$3B+/yr
OpenAIGPT-5.5Closed$852B$189.6B~$25B ARR~$7B+/yr
Google DeepMindGemini 3.5 Flash / 3.1 ProClosed~$4.3T *~$44B Cloud~$75B+/yr
Meta AILLaMA 4Open Weights~$1.4T *In $164B rev.~$60B+/yr
xAIGrok 4.1 / 4.20Partial (Grok)$250B$36.1B~$3.2B (2025)~$3B+/yr
DeepSeekDeepSeek V4 (preview)Open Weights (MIT)Self-funded$0Minimal~$1B+/yr
QwenQwen 3.5Open Weights (Apache 2.0)— *~$16B Cloud~$21B/yr
Perplexitypplx-apiClosed$21B$1.7B~$500M ARRNot disclosed
KimiKimi 2.5Open Weights$20B+$4.6B+~$200M ARRNot disclosed
MistralMistral Medium 3.5Open Weights (Apache 2.0)$14B$3.1B~$100M ARR~€240M/yr

* Parent company market cap, not standalone AI division valuation. Revenue and compute spend estimates based on publicly reported figures. Open Weights = model weights downloadable, but training data and code not shared.