In February, U.S. startups raised $174 billion. In March, they raised approximately $13 billion. The 92 percent decline is not a crash. It is an accounting fact that reveals how concentrated the venture capital market has become. [1]

This paper reported yesterday that the venture industry's apparent health was a mirage produced by a handful of AI megarounds. Crunchbase's updated March data, published Monday, makes that diagnosis sharper. February's $174 billion included OpenAI's $110 billion round — the largest private funding round in history — plus substantial raises by Anthropic and Waymo. Remove those three transactions and February shrinks to roughly $25 billion. March's $13 billion, while low, is not far from that adjusted baseline. [1][2]

The structural picture is what matters. Crunchbase's analysis shows that AI companies accounted for more than 60 percent of all venture deals above $100 million in the first quarter of 2026. The rest of the startup ecosystem — fintech, biotech, consumer software, enterprise SaaS — is raising capital at rates that would have been considered anemic even in the post-2022 downturn. The venture market has not recovered from its 2022 correction so much as it has been masked by AI's gravitational pull. [1]

Crunchbase cautioned against drawing structural conclusions from one month's figures. "The March slowdown is almost entirely due to fewer giant AI megarounds closing this month," the report noted. The implication — that the market is fundamentally sound and merely experiencing a pause between large closings — is plausible but incomplete. [1]

The war in Iran has introduced a new variable. Several investors told Fortune that limited partners have begun pulling back from illiquid commitments as energy costs and interest rate uncertainty cloud the macroeconomic outlook. LP appetite for venture capital is finite, and when it must compete with war-driven inflation and potential rate hikes, the smaller end of the startup spectrum loses first. [2]

The deeper question is whether the venture capital model works for non-AI companies in 2026. If the entire industry's headline numbers depend on three or four companies raising historically unprecedented sums, and if removing those companies reveals a market that has been flat or declining for three years, then venture capital is not so much an asset class as it is a delivery mechanism for concentrated AI bets. [1]

That is not necessarily a failure. It may simply be an honest description of where the technology economy's value creation is concentrated. But it raises uncomfortable questions for the thousands of startups that are not building foundation models — and for the investors who back them.

-- DAVID CHEN, Beijing