Investors poured $297 billion into startups in the first three months of 2026. That single quarter eclipses every full year of global venture activity before 2019. And four companies — OpenAI, Anthropic, xAI, and Waymo — captured $188 billion of it, according to Crunchbase’s Q1 report.
That means 65% of all venture capital invested on the planet went to four companies. AI overall consumed 81% of the total, up from 55% a year earlier.
The Numbers Are Absurd
OpenAI’s $122 billion round closed March 31 at an $852 billion valuation. Amazon invested up to $50 billion, Nvidia $30 billion, and SoftBank $30 billion. OpenAI also raised $3 billion from retail investors — ordinary people buying into a company that generates $2 billion per month in revenue but is still burning cash and is not profitable.
Anthropic’s $30 billion Series G valued the company at $380 billion. It was the second-largest venture deal of all time. Run-rate revenue: $14 billion. Total raised since 2021: nearly $64 billion.
xAI raised $20 billion in its Series E. Waymo pulled in $16 billion. Combined with OpenAI and Anthropic, these four frontier labs collected more money in a single quarter than the entire U.S. venture industry invested in the whole of 2020.
The Concentration Problem
Here’s the part that should worry anyone building an AI startup — or investing in one. Late-stage deals drove $244 billion of the quarter’s volume, a 200%+ year-over-year jump. Meanwhile, seed funding told a very different story.
Seed rounds totaled $12 billion, up 31% year over year in dollar terms. Sounds healthy until you look closer: deal counts fell 30% to 3,800. The top 20 seed deals captured more than half of all seed dollars. The next 300 split another quarter. Everyone else fought over scraps.
Seed valuations for AI companies have ballooned to $10 million rounds at $40–45 million post-money. You used to get that kind of valuation with a working product and paying customers. Now it buys you a pitch deck about “autonomous agents.”
U.S.-based companies took 83% of global venture capital in Q1, up from an already-high 71% a year earlier. China came in second at $16.1 billion, followed by the U.K. at $7.4 billion. The rest of the world split what was left.
The Gap Nobody Talks About
There’s a number that makes all these funding records look fragile: $690 billion.
That’s what Amazon, Alphabet, Meta, Microsoft, and Oracle plan to spend on AI infrastructure in 2026 alone, up from roughly $400 billion in 2025. Against that spending, enterprise AI revenue sits around $100 billion — generous estimates put generative AI revenue at $30–37 billion.
JPMorgan has done the math. The industry needs $650 billion per year in AI revenue to justify current infrastructure spending. We’re generating somewhere between $30 billion and $100 billion, depending on how broadly you define “AI revenue.” That’s somewhere between a 6x and 18x gap between spending and revenue.
Some of the companies making these infrastructure bets now have negative free cash flow as a result. They’re betting that demand materializes before the bills come due.
Who Gets Hurt
The biggest risk isn’t to OpenAI or Anthropic. They’ve locked in billions and can survive for years regardless. The risk falls on:
Mid-tier AI startups without clear differentiation. When enterprises consolidate spending to fewer vendors — and VCs predict they will — companies selling features that OpenAI or Google can add next week are the first casualties.
AI wrapper companies whose entire product is an API call with a UI on top. Investors have gotten wise to this. Seed funding now requires demonstrating defensibility: proprietary data, vertical depth, or meaningful switching costs. A chatbot skin doesn’t cut it anymore.
Retail investors who piled into OpenAI’s round. Buying into a private company at an $852 billion valuation that isn’t profitable is the kind of trade that looks either visionary or catastrophic in hindsight. There’s no secondary market to bail out of if things turn.
Non-AI startups competing for the same pool of capital. When 81% of VC money goes to AI, that leaves 19% for everything else — biotech, climate, fintech, all of it.
What This Means
We wrote about this concentration problem in early March when February’s numbers came in. The Q1 totals confirm it wasn’t a one-month anomaly. It’s the new normal.
The venture capital industry has effectively become an AI infrastructure financing mechanism. Four companies receive the majority of capital. Those companies buy compute from a handful of chip makers and cloud providers. The cloud providers spend hundreds of billions on data centers. And the whole thing needs AI revenue to grow 10x to 20x within a few years to pencil out.
Investors polled by GeekWire largely reject the word “bubble” but acknowledge that “capital and valuations are running well ahead of fundamentals.” That distinction may matter less than they think if the revenue doesn’t catch up to the spending.
The comparison that keeps coming up is the late-1990s telecom buildout, where companies spent hundreds of billions laying fiber optic cable that eventually carried the internet — after most of the companies that built it went bankrupt. The infrastructure survived. The investors didn’t.
Whether this cycle ends the same way depends on a simple question: can AI companies generate enough revenue to justify what’s being spent? So far, the answer is not yet. The bet is that it will be. And $297 billion per quarter is one hell of a bet.