Summary: Crunchbase put 2025 global venture funding at $425 billion, with AI startups receiving $211 billion, or about half. Foundation model companies drew roughly $80 billion. OpenAI and Anthropic together absorbed 14% of all global venture dollars.
Early January 2026 brought the final Crunchbase tally: global venture funding closed 2025 at $425 billion across more than 24,000 companies, a 30% climb over 2024’s $328 billion and the third-largest year ever recorded. AI startups claimed $211 billion of that pool — an 85% surge from $114 billion — which lands at roughly half of all venture dollars, versus a 34% share in 2024. CB Insights runs its own count slightly higher, at $226 billion and about 48%. The numbers disagree at the margin; the conclusion does not: AI captured close to half the world’s venture capital and sat atop the funding charts for the fourth consecutive year.
These figures matter because funding structure is what sets the 2026 product environment. Compute, talent, and distribution trail the money — and for application-layer startups, the year repriced both how hard it is to raise and what the exit menu looks like.
The Totals: $211B, About Half of Venture
Pull the total apart and the concentration starts leaking out. US companies took $274 billion, 64% of the global figure (up from 56% in 2024). Q4 alone deployed more than $113 billion — $66.5 billion late-stage, $37 billion early-stage, $9.9 billion seed. Seed climbing 12% year over year sounds respectable, yet the gap against late-stage keeps stretching — for any startup not on a megaround trajectory, the market on the ground still reads as winter.
Valuations swelled on the same curve. The Crunchbase Unicorn Board pushed toward $7.5 trillion in aggregate value, a gain of more than $2 trillion in twelve months, and Google’s $32 billion buyout of Wiz became the largest venture-backed M&A exit on record.
Foundation Labs Take Nearly Half the Money
Crunchbase’s year-end chart pack cuts the AI total one layer deeper: foundation model companies collected roughly $80 billion, some 40% of all AI funding — better than double 2024’s $31 billion, which had itself represented around 27%. The SoftBank-led $40 billion round into OpenAI ranked as the year’s biggest single deal and the largest private funding round ever recorded, lifting OpenAI to a $500 billion valuation and the title of the world’s most valuable private company.
How concentrated is that? OpenAI and Anthropic between them absorbed 14% of every venture dollar invested globally in 2025. A mere five companies raised more than $5 billion apiece — OpenAI, Scale AI, Anthropic, Project Prometheus, and xAI — and together they banked $84 billion, a full fifth of the year’s venture total. Meta’s $14.3 billion position in Scale AI, which carried CEO Alexandr Wang and his team in-house, wrote the emblematic footnote on corporate capital taking the field directly.
Megarounds and Bay Area Concentration
By deal shape, 58% of AI funding landed as megarounds of $500 million or more. The geography squeezes even harder: San Francisco Bay Area companies captured about $122 billion, more than three-quarters of US AI funding. The investor mix is being recast as well — private equity and alternative investors led roughly 300 sole-led rounds for a combined $63 billion, while venture capitalists led some 1,600 of them for $38 billion, defending their presence on deal count rather than dollars.
Revenue and Capex: The Other Two Yardsticks
Funding on its own inflates the bubble narrative; revenue on its own understates the buildout. Menlo Ventures pegs 2025 enterprise AI revenue at $37 billion, more than a threefold jump, split between roughly $19 billion of user-facing products and $18 billion of infrastructure. Across the same stretch, hyperscalers committed an estimated $300 billion-plus in capital expenditure. Revenue is giving chase, but the capex slope remains far steeper — and that widening scissors gap is the single biggest open question hanging over 2026.
Three Questions for 2026
One: does the application layer get fed? Seed at $9.9 billion against late-stage at $66.5 billion is a structure in which the small-team share keeps eroding. Two: can megaround valuations be redeemed by revenue? $37 billion of enterprise AI revenue set against $211 billion raised in a single year means the market will sooner or later audit retention and margins rather than demos. Three: do rules and energy costs redirect the flows? Transparency obligations such as California’s SB 53 are now in force, and power and data center capacity have become the binding constraints — the other thread woven through our opening outlook for the year.