AI Venture Funding Boom Leaves 220+ Fallen Unicorns Behind

A generation of startups stuck in 2021

The venture market now runs on two tracks. Artificial intelligence companies attract record sums, while a large group of startups that raised money before ChatGPT are stuck with outdated valuations and few ways to cash out.

PitchBook counts 857 U.S. startups valued at $1 billion or more. Nearly half of them had not raised new funding in the three years through the end of 2025, so their headline valuations reflect market conditions that no longer exist. 1 More than 220 companies that once reached unicorn status are now classed as "fallen unicorns." The list includes consumer brands such as Glossier, Savage X Fenty, AG1 and The Farmer's Dog. 1 The markdowns are large. Startups whose last round was in 2021 are worth 68% less on average, and those that last raised in 2022 have dropped 52%. 1

A separate analysis says 537 American unicorns still carry valuations set during the 2021–22 bubble. 2 That figure is higher than PitchBook's "nearly half" of 857, which would be roughly 430 companies. The gap probably comes from different methods, such as different cutoff dates or different definitions of a stale valuation. Both counts point to the same thing: hundreds of companies have not had their prices tested against today's market.

Where the money went

The main cause is concentration. More than $250 billion has gone into OpenAI and Anthropic alone, and that has reset how investors value whole categories of startups. 1 When so much capital flows to a few frontier labs and the companies built on top of them, everyone else is judged against them. A non-AI direct-to-consumer brand or a mid-tier SaaS tool simply looks less compelling by comparison.

Being an AI company is no longer enough to escape this. One analysis argues that calling a startup "AI native" is turning from an advantage into a weakness, because investors now ask what remains once the AI label is removed. 2 The scrutiny shows up in prices. Seed-stage companies without protectable intellectual property reportedly trade at a 20% to 30% discount to more defensible peers, and the gap widens to 30% to 40% by Series A. 2 The same source describes the market splitting into two winners: giant platforms and very small, lean software companies. Mid-sized single-purpose tools are left in a weak position between them. 2

The exit freeze is the real constraint

Low valuations would hurt less if companies could exit. They mostly cannot. PitchBook-NVCA's Venture Monitor reports that only 18 companies went public in the third quarter. 3 Unicorns that do exit are taking heavy losses. Airtable reportedly sold for $1.3 billion and Miro for $1.4 billion, both close to 90% below their earlier valuations. 3 Those sales show what a repricing actually looks like. The 68% average markdown PitchBook estimates for 2021-vintage startups may be a generous estimate for companies that have to sell. 13

The lack of exits is also changing who gets to raise venture funds. Established firms took 88.2% of all U.S. venture fundraising so far this year, and Andreessen Horowitz alone closed $23.8 billion, more than a fifth of the total. 3 Limited partners are putting their money behind a few well-known managers. That leaves less for newer firms that might back non-AI founders who are struggling to get from seed to Series A. 3

What it adds up to

These numbers connect into a cycle:

• AI absorbs most of the new capital. 1
• Older unicorns cannot raise without accepting a lower valuation. 1
• They also cannot exit without taking similar losses. 3
• Investors who see no returns from the 2021 cohort send their next commitments to the largest firms, which in turn favor AI. 3

Each step makes the others worse.

In my reading, the "fallen unicorn" label understates the problem. Losing billion-dollar status is a symptom. The bigger issue is a mismatch between how many companies were valued at the peak and how much liquidity exists to clear them. With 18 IPOs in a quarter, there is no realistic route for hundreds of stale-priced companies to go public. Most of them face one of three paths: a down round, a sale at a deep discount like Airtable's or Miro's, or a slow decline.

There is also a caution for the AI winners. The discounts on AI startups without defensible IP suggest investors are already separating real moats from AI branding. 2 Today's capital concentration could produce its own set of fallen unicorns if exits stay frozen when those companies' valuations are tested. The 2021 cohort shows that being well funded does not guarantee a company can get its investors' money back out.