In H1 2026, global venture capital reached $510B — a single half-year that surpassed all of 2025's $440B. But break that number down: OpenAI and Anthropic alone took 43%, U.S. deals of $100M+ accounted for 87.5% of the capital, while North American seed rounds were −27% year-over-year. The total belongs to the market; the record belongs to fewer than twenty companies — the exit window has opened and fundraising has come back, but both are flowing first to the same names. Every cell in the live dashboard can expand into a data table, backfilled quarterly.
Right now: AI has swallowed almost the entire VC market. Q1 came in at $305B, Q2 at $205B, together $510B for the first half, above the prior record set in H2 2021 of $375B. AI accounted for 80% in Q1 and 70% in Q2, with the US at 86% for H1 — up from under half a year ago. In Q3, AI's share fell for a second straight quarter to 64% (80%→70%→64%), even as deals over $1 billion hit a record. What to watch next: whether AI's share keeps falling in Q4, or the money is just concentrating into fewer, bigger deals. Non-AI companies are facing a market as tight as 2023 — it's just been drowned out by the big headline totals.
Valuation data has also split into two markets. Carta's share of down rounds fell below 12%, with median valuations rising across stages; but Series A shows a two-track pattern: AI foundation model companies have a median valuation of about $300M, versus about $55M for non-AI companies. Using the overall median to price a non-AI deal would systematically overestimate it. Europe's and Asia's rebounds are likewise concentrated rebounds — Europe's Q2 reached $24B (+67% year-over-year), with 65% going to just 42 companies; Asia reached $42.8B, even as deal count fell to a multi-year low.
Anthropic went from a $380 billion valuation in February to $965B in its May funding round — about 2.5x in three months — with run-rate revenue above $65B. By October, market talk of an IPO valuation was still hovering above $1 trillion (see Thesis Two), but the direction is downward, not a one-way climb: the markdown on Oct 15 already pulled expectations back from $2 trillion to $1.5 trillion. The pricing logic at this level is "compute financing" rather than traditional venture investing: the money mostly turns into data-center commitments, and returns hinge on how fast inference demand grows. Q3 actual revenue was $16.8 billion but growth slowed, and the Oct 14 disclosure became one of the triggers for that day's AI stock selloff — the market has started pricing these companies on growth rate rather than scale.
On June 12, SpaceX's IPO raised $75B at a valuation of $1.77T, popping +19% on day one — the largest ever; it then acquired Cursor's parent company Anysphere in an all-stock deal worth $60B, closing on August 14, the largest startup acquisition on record. In Q2, 32 VC-backed companies valued at $1 billion or more went public, and 24 M&A deals above $1 billion totaled $113B — the highest quarter ever. Anthropic plans to go public November, while OpenAI leans toward pushing its listing to 2027.
But the medians tell a different story: Carta recorded 421 M&A exits in H1, the busiest half-year on record, yet the median deal size was just $71M; SpaceX alone skewed total exit value by roughly $1.7 trillion. The liquidity is actually being absorbed by the secondary market — global private secondary transactions topped $120B in H1 (+20% year-over-year), with GP-led deals at 53.7%, an all-time high; single-asset continuation funds reached $34B. "Keep the best assets and sell them again" has become a standard GP playbook for managing DPI (distributions to paid-in capital) pressure.
Now: exit proceeds haven't really reached LPs' pockets yet. For funds launched in 2021, DPI (cash returned as a share of capital invested) is still under 0.1x five years in, with holding periods stretching past 7 years; secondary markets have already started discounting for the AI shock. Anthropic's IPO timeline is a hard constraint: the S-1 was filed with the SEC on Oct 26 (ticker ANT), the roadshow is set for the week of Nov 9, pricing is targeted before Thanksgiving, and valuation expectations have dropped from $2 trillion to $1.5 trillion; the prospectus shows first-half revenue of $16.286 billion and a net loss of $80.748 billion, with the loss running near five times revenue. Secondary-market comparables are broadly down: 5 of 6 tracked names are lower year-to-date, with only the new-issue-tracking ETF still up 2.29% this year — the volatility hasn't settled, and going public doesn't mean the money is actually in hand yet.
U.S. fund formation in H1 reached $74.8B, already surpassing all of 2025's $66B (the weakest year since 2017); but a16z, Thrive, and Founders Fund alone accounted for 48.1%. The a16z fund family topped $15B, the largest single raise on record; Thrive's Fund X reached $10B. The pace of first-time fund closings is the lowest since 2016.
H1 fundraising reached ¥1.01T (YoY −4.9%), with 4,606 new funds established (YoY +134%) — small funds surged while large funds shrank. State-owned capital platforms accounted for 41% of funding, up 424% YoY — state capital isn't filling a gap, it's taking over the entire funding chain. The national venture guidance fund runs on a 20-year horizon, aiming to mobilize trillions in scale; guidance funds are shifting from new formation to consolidating existing holdings, with return-investment requirements loosened and fault-tolerance provisions taking effect.
H1 investment reached ¥828.5B (+44% year-over-year) across 6,317 deals; AI took 48.6% of the capital while accounting for only 22.5% of deal events, with DeepSeek's round alone at ¥51B. On the exit side, 156 companies IPO'd, with paper returns of ¥366.8B (+247% year-over-year); 87 companies listed in Hong Kong, raising HK$210B, with average first-day returns over 60%.
Now: the money is coming from state-backed sources, flowing into AI, and exiting through Hong Kong listings — this is the capital-flow evidence behind the "Manufacturing Winners" approved-list system. Exit paths for companies on the list are open, but not easy: Transsion Holdings listed on the Hong Kong Stock Exchange on Oct 15, raising about $390 million, and closed down 5.72% from its offer price on debut day. Secondary-market comparables remain broadly lower, with 5 of 6 tracked names down year-to-date; but the primary market hasn't cooled in response — Form D (private placement) filings over the past 30 days total about $12.40 billion. By wangvc's count: there have been 68 mega-rounds of $1 billion or more this year, totaling about $429.09 billion, with the top five companies taking 70.43% and no new ones added for 3 straight days; in Q4 so far there have been 6 such rounds worth about $33.33 billion, with AI still accounting for roughly 84%. What to watch next: whether this pause is a brief gap or a genuine slowdown among listed companies, and whether the gap between primary pricing and actual secondary-market take-up prices widens.
Venture totals can no longer be read apart from a short list of names: the earlier concentration figures already told the story — among this year's billion-dollar-plus mega-rounds in the wangvc database, the top five firms took 70.43% of the total. Widen the list to twenty and the share only rises, never falls. Strip out that list, and what's left isn't a shortage of dollars — it's a shortage of deal count. The line that's genuinely back to 2023 levels is the non-AI one.
If, after excluding the dozen or so companies at the top, the remaining market's funding amount and deal count were to both return to 2021 levels, then the claim that "the total belongs to the market, the record belongs to a few companies" would no longer hold — concentration would just be a temporary phenomenon.