TL;DR: AI startups raised about $202B in 2025, roughly half of all global venture capital. Anthropic ($965B) and OpenAI ($852B) now lead the most valuable startups on Earth.
AI startups pulled in about $202 billion of venture capital in 2025, close to half of all global VC and up 75% from $114 billion in 2024, according to Crunchbase. Foundation model labs took the biggest share. OpenAI and Anthropic alone absorbed roughly 14% of every venture dollar invested worldwide. Anthropic closed a $65 billion round in May 2026 at a $965 billion valuation, making it the most valuable private company on the planet, just ahead of OpenAI at $852 billion.
The figures here come from Crunchbase end-of-year 2025 data, the CB Insights State of Venture 2025 report, the OECD, company funding announcements, and Second Talent placement data from our network of 100,000+ engineers across Asia. This post covers totals, mega-rounds, valuations, sectors, regions, exits, and the bubble debate. It does not forecast share prices or rate individual startups.
One trend defines this market. Capital is concentrating fast. A handful of frontier labs now soak up most of the money, while smaller AI teams compete for the rest and for the scarce engineers who can ship real products. That makes where and how you hire AI talent a strategic question, not just a budget line.
Key takeaways
- AI startups raised about $202B in 2025, roughly 50% of all global venture capital, up from 34% in 2024.
- Foundation models and LLMs took around $80B, about 40% of all AI funding and more than double 2024.
- Anthropic ($965B) and OpenAI ($852B) are now the two most valuable startups in the world.
- US startups captured 79% of global AI funding in 2025, with the San Francisco Bay Area alone at $122B.
- The risk is not too little money. It is too much, in too few companies, far ahead of revenue.
Global AI Startup Funding Overview
The 2025 AI funding map is lopsided. North America raised the vast majority of capital, while Europe and Asia-Pacific grew but stayed far behind. China declined under regulatory and export pressure. The table below uses Crunchbase end-of-year 2025 figures, with regional shares cross-checked against the OECD.
| Region | AI Funding 2025 | Global Share | 2024 Comparison | Trend |
|---|---|---|---|---|
| United States | $159B | 79% | $77B | Surging |
| Europe | $15.8B | ~8% | $13B | Steady growth |
| China | $13.9B | ~5% | $18B | Declining |
| Asia-Pacific (ex-China) | $9B | ~4% | $8B | Slow growth |
| Rest of World | $4B | ~2% | $3B | Emerging |
| Global Total | ~$202B | 100% | $114B | +75% YoY |
The concentration is even sharper inside the US. Crunchbase reports the San Francisco Bay Area alone raised $122 billion, about 76% of the US total. The early 2026 data shows no slowdown. AI startups took $242 billion in the first quarter of 2026, about 80% of all global venture capital that quarter, a new record.
AI’s Share of Venture Capital Keeps Climbing
AI is no longer one sector among many. It is the venture market. CB Insights put AI at about 48% of all global VC in 2025, the largest share on record. The OECD framed it even higher at 61%. Either way, the direction is the same, and Q1 2026 pushed the figure to roughly 80%.
This is good news for AI founders raising capital. It is harder news for everyone else. Non-AI startups now compete for the shrinking half of the pool. It also means investors expect AI inside the pitch. A SaaS company with no AI story struggles to stand out in a market where capital chases models, agents, and data.
Just raised, or about to? Pick your next hiring move.
Funded AI startups burn most of their round on talent. Here is where to start.
Funding Stage Distribution and Trends
Money is moving to the late stages. Mega-rounds of $500 million or more accounted for about 58% of all AI dollars in 2025. Seed and Series A are still active, but the headline capital sits with proven labs. The table below shows the rough split across stages.
| Funding Stage | Share of AI Capital | Typical Round | Median Valuation | 2025 Trend |
|---|---|---|---|---|
| Pre-Seed / Seed | ~6% | $2M to $8M | $20M | Strong, agent-led |
| Series A | ~10% | $10M to $30M | $90M | Selective |
| Series B | ~12% | $30M to $80M | $350M | Steady |
| Series C+ | ~14% | $100M to $500M | $1.5B | Premium |
| Mega-rounds ($500M+) | ~58% | $1B to $122B | $30B+ | Dominant |
By investor type, private equity led with about $63 billion deployed into AI, ahead of traditional VC at $38 billion. The rest came from corporate venture arms, sovereign funds, and crossover hedge funds chasing the frontier labs.
Largest AI Funding Rounds of 2025 and 2026
The mega-rounds of this cycle broke every prior record. OpenAI’s $40 billion round in March 2025 was the largest private tech financing ever, then its March 2026 round more than tripled it. Anthropic raised three times in nine months and ended up the most valuable startup in the world.
| Company | Round | Date | Amount | Valuation | Lead Investors |
|---|---|---|---|---|---|
| OpenAI | Primary | Mar 2026 | $122B | $852B | SoftBank, a16z, MGX |
| Anthropic | Series H | May 2026 | $65B | $965B | Sequoia, Altimeter |
| OpenAI | SoftBank-led | Mar 2025 | $40B | $300B | SoftBank, Microsoft |
| Anthropic | Series G | Feb 2026 | $30B | $380B | GIC, Coatue |
| xAI | Series E | Jan 2026 | $20B | ~$230B | Valor, Fidelity, Nvidia |
| Meta / Scale AI | Minority stake | Jun 2025 | $14.3B | ~$29B | Meta (49%, non-voting) |
| Anthropic | Series F | Sep 2025 | $13B | $183B | ICONIQ, Fidelity |
| Databricks | Series L | Feb 2026 | ~$5B | $134B | Insight, Fidelity |
| Anduril | Follow-on | May 2026 | $5B | $61B | Thrive, a16z |
| Anysphere (Cursor) | Series D | Nov 2025 | $2.3B | $29.3B | Accel, Coatue |
The Meta and Scale AI deal is often mislabeled. Meta paid $14.3 billion for a 49% non-voting stake, not a takeover. Scale stays independent, though its CEO left to run Meta’s Superintelligence Labs. Other big raises included Mistral AI’s 1.7 billion euro round led by chipmaker ASML, and Thinking Machines Lab, the startup from former OpenAI CTO Mira Murati, which raised $2 billion at a $12 billion valuation before a later up-round stalled.
AI Startup Valuations in 2026
Valuations have detached from the rest of the market. Anthropic at $965 billion and OpenAI at $852 billion are worth more than almost every public company. The gap between them and the next tier is wide, and the gap between valuations and revenue is wider still.
| Company | Valuation (2026) | Focus | Reported Revenue Run-Rate | Last Round |
|---|---|---|---|---|
| Anthropic | $965B | Frontier models (Claude) | ~$47B (May 2026) | Series H, May 2026 |
| OpenAI | $852B | Frontier models (ChatGPT) | ~$13B (end 2025) | Primary, Mar 2026 |
| xAI | ~$230B | Frontier models (Grok) | ~$500M (2025) | Series E, Jan 2026 |
| Databricks | $134B | Data and AI platform | $3B+ ARR | Series L, Feb 2026 |
| Safe Superintelligence | $32B | AI safety research | Pre-revenue | Apr 2025 |
| Anysphere (Cursor) | $29.3B | AI coding | $1B+ ARR | Series D, Nov 2025 |
| Perplexity | $20B | AI search | ~$150M ARR | Sep 2025 |
| Mistral AI | ~$13B | Open-weight models | Not disclosed | Series C, Sep 2025 |
Anthropic’s roughly $47 billion run-rate is the standout, the closest any lab comes to justifying its price. Most others trade on growth, not profit. For founders, the lesson is simple. The market rewards real revenue and defensible data, not just a good demo. That is also why proven AI and machine learning engineers command the salaries they do.
Which AI Sectors Attracted the Most Capital
Foundation models and LLMs dominate. They took about $80 billion in 2025, around 40% of all AI funding and more than double 2024’s $31 billion. After that, the money spreads into infrastructure, defense, coding tools, and healthcare.
| AI Sector | 2025 Funding | Share of AI Total | Notable Players | Trend |
|---|---|---|---|---|
| Foundation Models & LLMs | ~$80B | ~40% | OpenAI, Anthropic, xAI, Mistral | Dominant |
| AI Infrastructure & Compute | High (capex-led) | Large | CoreWeave, Databricks, Together AI | Surging |
| Defense AI | ~$49B | ~24% | Anduril, Shield AI, Helsing | Breakout |
| AI Coding & Agents | Strong | Growing | Anysphere, Cognition, Glean | Hot M&A |
| Healthcare AI | Strong | Steady | Tempus, Abridge, OpenEvidence | Resilient |
| Robotics & Humanoids | Strong | Growing | Figure AI, Physical Intelligence | Rising |
Defense AI was the surprise of 2025, with a near-record year anchored by Anduril’s $2.5 billion round at $30.5 billion. AI coding tools became the hottest area for acquisitions and acqui-hires, with Anysphere passing $1 billion in annual recurring revenue. Demand for engineers who can build and deploy these systems, including data scientists and MLOps specialists, climbed alongside the funding.
Unicorns, Decacorns, and Mega-Valuations
AI minted unicorns at a record pace in 2025. CB Insights counted 75 new AI unicorns, about 61% of all new unicorns created that year, up from 52% in 2024. Crunchbase, using a stricter definition, counted 47. Either way, AI is now the default path to a billion-dollar valuation.
| Metric | 2025 Result | 2024 Comparison | Notes |
|---|---|---|---|
| New AI unicorns | 75 | ~60 | 61% of all new unicorns |
| Total decacorns ($10B+) | 82 | 70 | 24 minted in 2025 |
| Most valuable startup | Anthropic $965B | OpenAI $157B | Passed SpaceX |
| Fastest to unicorn | Lovable (~8 months) | ~1.5 years | AI tools scale faster |
| Unicorn board value | ~$7 trillion | ~$5.4 trillion | OpenAI + Anthropic ~10% |
The speed is striking. Crunchbase’s unicorn board reached about $7 trillion in value by the end of 2025, and just two companies, OpenAI and Anthropic, account for roughly a tenth of it. AI tools also reach unicorn status faster than past startups because the products scale with little extra headcount.
Geographic Investment Patterns
The United States is not just leading. It is taking almost everything. US startups captured 79% of global AI funding in 2025, with the San Francisco Bay Area alone responsible for $122 billion. Europe and Asia grew in absolute terms but lost ground in share.
This concentration creates a talent paradox. The capital is in California, but the engineers are everywhere. Salaries in the Bay Area have climbed past what most funded startups outside the top labs can pay. That is why more teams build distributed AI engineering teams across Asia, where the skills exist at a fraction of the cost. Singapore, Vietnam, and the Philippines have become real AI delivery hubs, not just back-office locations.
| Hub | Strength | Senior AI Engineer (Monthly) |
|---|---|---|
| San Francisco Bay Area | Frontier labs, capital | $18,000 to $30,000 |
| New York | Enterprise and fintech AI | $15,000 to $24,000 |
| London | DeepMind heritage, research | $10,000 to $16,000 |
| Singapore | Asia HQ, applied AI | $6,000 to $9,000 |
| Vietnam | Engineering depth, low cost | $4,000 to $6,000 |
| Philippines | English fluency, data and AI ops | $3,500 to $5,500 |
A funded startup that hires three senior AI engineers in Vietnam instead of San Francisco can save more than $400,000 a year. In a market where most rounds go to talent, that is months of extra runway. Our work helping companies find the talent they need across these hubs shows the gap is about price and availability, not skill.
Exits, M&A, and IPOs
Exits returned in force. Global M&A reached about $4.8 trillion in 2025, the second-highest year ever, and tech M&A jumped 76% to roughly $478 billion. Most large strategic tech deals now involve an AI-native target. The IPO window also reopened for AI infrastructure.
| Deal | Type | Date | Value | Note |
|---|---|---|---|---|
| CoreWeave | IPO | Mar 2025 | ~$1.5B raised | AI cloud, ~$23B debut |
| Cerebras | IPO | May 2026 | ~$5.55B raised | AI chips, ~$95B, +68% day one |
| OpenAI / io (Jony Ive) | Acquisition | May 2025 | ~$6.5B | OpenAI’s largest acquisition |
| Google / Windsurf | Reverse acqui-hire | Jul 2025 | ~$2.4B | Licensing plus team, no equity |
| Meta / Scale AI | Minority stake | Jun 2025 | $14.3B | 49% non-voting |
The Cerebras IPO in May 2026 was the largest US tech listing since 2020, a clear sign that public investors want AI hardware exposure. Strategic buyers, meanwhile, increasingly use license-and-hire structures, as Google did with Windsurf, to acquire AI teams while sidestepping antitrust review.
Is AI Funding a Bubble?
The people closest to the money are worried. OpenAI’s Sam Altman said in 2025 that investors are overexcited and compared the moment to the dot-com bubble. Goldman Sachs Research published a note titled “AI: In a Bubble.” The concern is not the technology. It is the math.
| Bubble Signal | 2025-2026 Data | Source |
|---|---|---|
| Capital concentration | OpenAI + Anthropic = 14% of all global VC | Crunchbase |
| Revenue gap | Sequoia’s “$600B question” on AI revenue needed | Sequoia Capital |
| Pilot failure rate | 95% of enterprise GenAI pilots show no P&L impact | MIT Project NANDA |
| Hyperscaler capex | $700B to $900B planned for 2026, up 36% | CreditSights |
| Q1 2026 concentration | 4 labs took 65% of all global VC | Crunchbase |
A 2025 MIT study found that 95% of enterprise generative AI pilots delivered no measurable profit impact. At the same time, the big cloud providers plan to spend up to $900 billion on AI infrastructure in 2026. When capex runs that far ahead of revenue, history says someone loses money. The likely outcome is not a collapse of AI, but a painful sorting between companies that turn models into profit and those that do not.
Future Investment Outlook (2026 to 2030)
Most analysts expect AI funding to stay high but get more selective. The easy money for any team with “AI” in the deck is fading. Capital will keep flowing to revenue, defensible data, and real enterprise adoption. The table below shows where the growth is expected.
| Category | Direction to 2030 | Key Driver |
|---|---|---|
| Foundation models | Concentrating in 4 to 6 labs | Compute and capital scale |
| AI agents and automation | Fastest growth | Enterprise workflow ROI |
| AI infrastructure | Heavy capex, some overbuild | Training and inference demand |
| Vertical AI (health, finance, defense) | Steady, revenue-led | Regulated, sticky markets |
| Applied AI talent | Persistent shortage | Demand outpaces supply |
One thing does not change. Every funded AI company needs engineers who can ship. The World Economic Forum Future of Jobs Report 2025 ranks AI and machine learning specialists as the fastest-growing role in the world. Funding can spike or cool, but the talent shortage behind it is structural.
Frequently Asked Questions
How much did AI startups raise in 2025?
AI startups raised about $202 billion in global venture capital in 2025, according to Crunchbase. That was roughly half of all VC worldwide and a 75% jump from $114 billion in 2024. CB Insights put the figure slightly higher at $226 billion using a broader definition.
What is the most valuable AI startup in 2026?
Anthropic is the most valuable, at a $965 billion valuation after its May 2026 Series H. OpenAI follows at $852 billion. Both are now worth more than almost every public technology company, and both have filed confidentially for eventual IPOs.
Is AI startup funding in a bubble?
Many investors think parts of it are. Sam Altman and Goldman Sachs have both warned of bubble dynamics. The core issue is that valuations and infrastructure spending are far ahead of current revenue. A correction is likely for weaker companies, though the strongest labs with real revenue are better positioned.
Why does so much AI funding go to the US?
US startups took 79% of global AI funding in 2025, mostly in the San Francisco Bay Area, where the frontier labs, top investors, and senior researchers cluster together. This concentration drives up US salaries and pushes many funded teams to hire AI talent in lower-cost regions to extend their runway.
What This Means for Your Hiring Budget
AI funding hit record highs in 2025 and accelerated into 2026, but the money is concentrating in a few labs while everyone else fights for talent. Most of any AI round is spent on engineers, and US salaries have climbed out of reach for many funded startups. That makes how you build your team the real lever on runway.
We help companies build senior AI and engineering teams across Asia for 60 to 75% less than US rates, with no upfront cost and a top 1% acceptance rate. If you have just raised, or are about to, talk to us about how to find the talent you need and make your round last longer.
Data Sources & Methodology
This analysis draws on the following 2025 and 2026 sources:
- Crunchbase – AI funding totals, mega-rounds, unicorn board, Q1 2026 data
- CB Insights State of Venture 2025 – AI share of VC, unicorn counts
- OECD – Regional AI funding shares and corporate VC participation
- Stanford HAI AI Index – Long-run investment and unicorn timing
- Bain & Company M&A Report – 2025 M&A and tech deal activity
- Goldman Sachs Research – AI bubble and capex analysis
- Sequoia Capital – AI revenue gap framework
- Company funding announcements from OpenAI, Anthropic, xAI, Mistral, Databricks, and Anduril
Methodology: Figures aggregated from Q1 2025 through Q1 2026, the latest full quarter available. AI funding totals vary by tracker depending on how “AI” is defined, so headline figures use Crunchbase unless stated. Valuations reflect disclosed post-money figures from named rounds. Some 2026 rounds and IPO plans were reported by major outlets but not company-confirmed, and are described as such. All amounts in USD unless noted. Last updated: June 2026.
