AI didn't just lead venture in 2025 — it became the market. AI startups pulled in a record $192.7 billion, more than half of every venture dollar for the first time ever, while 498 private AI unicorns reached a combined ~$2.7 trillion in value. Here are the numbers behind the boom, with sources — and the multiples that make it a risk as much as a revolution.
a new global record
Anthropic + OpenAI dwarf the entire rest of the field.
Source: Company filings / press (mid-2026)
AI crossed half of all venture funding for the first time.
Source: PitchBook
AI raises at a ~5x premium to SaaS — the core risk.
Source: Aventis Advisors
Up from 65% in early 2024 — the demand is real.
Source: McKinsey
A handful of names hold most of the $2.7T.
Source: Company filings (mid-2026)
a new global record
first year above half
worth ~$2.7T combined (mid-2025)
75 new AI unicorns, up from 52%
vs ~6–7x for public SaaS
up from 65% in early 2024
highest in YC history
vs 23% before ChatGPT
In 2025, AI startups raised a record $192.7 billion globally — crossing 50% of all venture dollars for the first time, off roughly a third of all completed deals. The concentration is even sharper at the top of the cycle: in the most recent quarter, AI took about 62.7% of US venture dollars and 53.2% globally.
| Metric | Value |
|---|---|
| Total AI VC invested (2025) | $192.7B |
| AI share of all VC dollars | >50% |
| AI share of all deals | ~33% |
| AI share of US dollars (latest quarter) | 62.7% |
| AI share of global dollars (latest quarter) | 53.2% |
As of mid-2025 there were 498 private AI unicorns worth roughly $2.7 trillion combined, and AI made up 61% of all new unicorns minted that year. The top of the table dwarfs everything below it: as of mid-2026, Anthropic was valued at roughly $965B and OpenAI at ~$852B — a single pair worth more than most of the rest of the list put together.
| Company | Valuation | Note |
|---|---|---|
| Anthropic | ~$965B | Series H, May 2026 |
| OpenAI | ~$852B | after ~$122B raised in Q1 2026 |
| Cursor (Anysphere) | $60B | pending SpaceX acquisition |
| Scale AI | ~$29B | Meta ~49% stake |
| Perplexity | ~$20B | Sept 2025 round |
AI startups raised at roughly 25–37x revenue in 2025, versus about 6–7x for public SaaS — a several-fold premium. That gap is the central risk in the whole category: these valuations price in extraordinary, sustained growth, and any slowdown re-rates them hard. It's why the same numbers read as 'revolution' to bulls and 'bubble' to bears.
The whole AI trade rests on one assumption: that ~30x revenue multiples are justified by growth that hasn't happened yet. If that growth slows even slightly, the re-rating won't be gentle.
The demand side is real, not just hype: about 71% of organizations now regularly use generative AI in at least one business function, up from 65% in early 2024 (McKinsey). And the next generation of startups is almost entirely AI — 88% of Y Combinator's Summer 2025 batch was AI-native, the highest share in YC's history, and AI took 42% of all seed capital. The funnel from seed to frontier is now overwhelmingly AI-first.
AI startups raised a record $192.7 billion globally in 2025 — more than half of all venture dollars for the first time, and roughly a third of all completed deals. In the US, AI took about 62.7% of dollars invested in the latest quarter (PitchBook).
There were about 498 private AI unicorns worth roughly $2.7 trillion combined as of mid-2025, with AI making up 61% of all new unicorns minted that year (CB Insights). The most valuable are Anthropic (~$965B) and OpenAI (~$852B) as of mid-2026.
AI startups raised at roughly 25–37x revenue in 2025, versus about 6–7x for public SaaS — a several-fold premium driven by growth expectations. That gap is also the key risk: it prices in extraordinary future growth (Aventis Advisors).
About 71% of organizations now regularly use generative AI in at least one business function, up from 65% in early 2024 (McKinsey). On the startup side, 88% of YC's Summer 2025 batch was AI-native and AI took 42% of all seed capital.
Compiled and maintained by Trace Cohen · @Trace_Cohen · t@nyvp.com. Free to cite with attribution (CC BY 4.0).