In February, U.S. startups raised $174 billion. In March, they raised roughly $13 billion. The decline is not a typo. It is not a rounding error. It is the sound of a market that has been running on a very specific kind of fuel — AI megarounds of $500 million and above — discovering what happens when that fuel stops flowing for a few weeks. [1][2]

Crunchbase News published the March data on Monday and immediately cautioned against drawing structural conclusions from a single month. The slowdown, the outlet wrote, "is almost entirely due to fewer giant AI megarounds closing this month." The underlying message was clear: this is a calendar effect, not a market collapse. But the underlying data tell a more interesting story than the one Crunchbase chose to lead with. [2]

The February number was itself an anomaly. OpenAI closed a $110 billion round that month — the largest private funding round in history — and that single transaction accounted for more than half of the month's total. Anthropic and Waymo added their own substantial raises. Three companies accounted for the overwhelming majority of $189 billion in global venture capital. Remove them, and February's number drops to something much closer to March's. [1]

This is the structural reality that the March figure makes visible. The venture capital market in 2025 and 2026 has been operating on a two-tier basis: a handful of AI companies raising historically unprecedented sums, and everything else. TechCrunch reported last week that AI startups accounted for 41 percent of the $128 billion in venture dollars raised by companies on Carta in 2025 — a record-high annual share. In the first quarter of 2026, AI's share of deals over $100 million climbed to 62 percent. The industry is not so much investing in artificial intelligence as it is being consumed by it. [3]

The question that March's number poses is what happens to the rest. Crunchbase has tracked a six-year decline in late-stage rounds of $30 million and under — the bread-and-butter raises that sustain enterprise software companies, fintech startups, and the vast middle class of the venture ecosystem. These companies have not disappeared, but their access to capital has narrowed in direct proportion to AI's expansion. The money is finite. The AI bets are not. [4]

There is a case, which Crunchbase makes and which is not unreasonable, that March is simply a pause between mega-deals. April or May could produce another billion-dollar AI close — CoreWeave, Databricks, or any of the half-dozen companies rumoured to be raising — and the monthly totals would spike again. The monthly data in venture capital has always been volatile. What matters is the trend, and the trend, depending on how you choose to read it, either shows a market in robust health or a market that has become dangerously concentrated around a single thesis. [2]

The concentration risk is real. When 41 percent of an industry's capital flows to companies built around a single technology, the industry's returns become a bet on that technology's commercialisation timeline. If large language models generate the revenue their valuations imply within the next three to five years, the AI megarounds will look prescient. If they do not — if the revenue materialises more slowly, or at lower margins, or not at all — the correction will be felt not just by the AI companies themselves but by every startup that did not receive funding because the capital had already been allocated elsewhere. [3]

Seed-stage investing has not been immune. Crunchbase reported in January that over 40 percent of seed and Series A investment in 2026 went to rounds of $100 million or more — a figure that would have been inconceivable five years ago. The traditional venture model — small bets on many companies, with returns driven by outlier exits — is being replaced by something that looks more like private equity: large bets on a few companies, with returns driven by continued fundraising at higher valuations. [5]

For the moment, the March figure is a data point, not a verdict. The venture capital industry will process it, contextualise it, and move on. The AI megarounds will resume. The monthly totals will recover. But the $13 billion number will sit in the record as a reminder of what the market looks like when the largest bets take a breath, and the answer is: considerably smaller than anyone would prefer to admit.

-- DAVID CHEN, Beijing