AI Reckoning Strands 220 Pre-ChatGPT Unicorn Startups

More than 220 U.S. startups that crossed the billion-dollar mark during the venture boom have quietly slipped below it, casualties of an AI shift that has redrawn how investors value pre-ChatGPT companies. New data from the private-markets research firm PitchBook shows nearly half of America’s 857 unicorns have not raised fresh money in three years, leaving their valuations frozen at prices the market no longer believes.

The reckoning is not about a single bad quarter. It is the bill arriving for two assumptions that propped up the last decade of venture investing: that interest rates would stay near zero, and that a struggling startup could always be sold for its engineers. Generative AI broke the second one, and the Federal Reserve broke the first.

The Floor Under Startup Valuations Just Gave Way

For most of the 2010s, a software founder who could not grow into a lofty valuation still had a parachute. A bigger technology company would buy the startup for its engineering team, often paying roughly $2 million per coder. A firm with 100 engineers was therefore worth $200 million to $300 million almost regardless of its revenue, a number that quietly set a floor under thousands of private valuations.

That floor is gone. AI coding tools now let a handful of developers ship what once took a full department, so the talent a buyer would have paid a premium for is no longer scarce. “Now you’re seeing 50 engineers do what it would’ve taken 500 engineers to do five years ago,” said Samir Kaul, a partner at Khosla Ventures and an early backer of OpenAI. “We had to completely reshuffle how we valued these companies.”

The change hit a market that was already cooling. When the Fed began raising rates in 2022, the cheap capital that inflated 2021 valuations dried up, and many founders bet they could grow into their numbers before they had to raise again. Then an app called ChatGPT arrived, and the engineering-talent premium that underwrote their fallback exit evaporated within months.

What is left is a cohort of companies that are neither fundable nor sellable at anything close to their old marks. They are too richly valued for new venture money and not profitable enough for the public markets, stranded in between.

Why Nearly Half the Unicorns Went Quiet

A valuation only means something when someone agrees to pay it. Across the unicorn class, almost half have not tested that question since 2021 or 2022, and PitchBook’s modeling, built on head-count trends and comparisons with public peers, suggests the silence is hiding steep markdowns.

The estimates are blunt. Startups whose last round closed in 2021 are now worth about 68% less on average, while those that last raised in 2022 are down roughly 52%. Software companies dominate the wreckage: 75 software-as-a-service firms (SaaS, the per-seat subscription model that bills by the user) appear on PitchBook’s fallen-unicorn list, double the number of fintech names behind them.

“When we see companies not raising, it’s a red flag,” said Andrew Akers, an analyst at PitchBook, noting that prolonged silence usually means growth has gone flat or negative. A few quiet firms are simply profitable and uninterested in diluting, he added, but that is the exception. You can see the other side of the same trend in PitchBook’s tally of US unicorn value, where a handful of AI leaders now hold the majority of the herd’s worth.

Enterprise Software Took the Hardest Hit

The category bleeding most is the one that looked safest two years ago. Workflow software embedded itself in how employees did their jobs and charged a fee for every seat, a model that scaled beautifully when headcounts grew. Autonomous AI agents threaten the math directly, because software that does the work needs fewer humans logged in to pay for it.

The Per-Seat Model Under Pressure

David Zhu, a former head of engineering at DoorDash who led more than 200 engineers there, surveyed the field after generative AI landed and reached a stark conclusion. “The thesis I had was that all workflow-driven enterprise SaaS companies will be either disrupted or dead in the next decade,” Zhu said. He left to found Reevo, an AI platform that automates corporate sales and marketing teams, betting on the disruption rather than against it.

Why Incumbents Struggle to Pivot

The trouble for older startups is structural, not cosmetic. Their cost base assumes large teams, and their products were architected for a pre-AI world, so bolting on a chatbot does not fix the underlying model. “Unless they make a stark, 180-degree pivot to rebuild the exact same thing from scratch, they’re going to slowly fail,” Zhu said.

That logic is why money is flowing to founders building AI-native infrastructure like Anthropic’s Claude models rather than into rescues of yesterday’s leaders. Investors, Zhu noted, would rather wager on new entrepreneurs at lower entry prices than double down on an aging cap table.

Public software giants are feeling the same chill. Shares of Salesforce, ServiceNow and Workday all sank this year as investors priced in the same agent threat that is gutting private SaaS valuations.

The Direct-to-Consumer Class Reprices

Software is not the only casualty. A whole generation of direct-to-consumer brands (DTC, companies that sell straight to shoppers online and skip traditional retail) raised at software-style multiples on the promise of software-style margins that never fully arrived.

The fallen-unicorn list reads like a roll call of podcast-ad mainstays and venture darlings, brands that came of age when growth mattered more than profit:

• Glossier, Rothy’s and Brooklinen, DTC consumer names once valued like high-growth tech
• Savage X Fenty, the lingerie label founded by the musician Rihanna
• The Farmer’s Dog, the fresh pet-food subscription service
• AG1, the powder supplement advertised across half the podcast economy
• Betterment, the robo-advisor pioneer, and the ticket marketplace SeatGeek

The pressure is already forcing hands. After CNBC asked about AG1’s place on the list, Reuters reported the supplement maker was exploring a sale of part or all of itself at a $2 billion valuation, a figure said to include the company’s debt.

Where the Money Rotated Instead

The capital did not vanish; it moved. More than $250 billion has poured into OpenAI and Anthropic ahead of their expected mega-listings, and the rare pre-AI startups posting strong numbers are still getting funded handsomely. The dividing line is whether a company looks built for the AI era or stranded before it.

Mercury, which provides banking services to about a third of early-stage U.S. venture-backed firms, is the clearest example of the new winners. The fintech closed a $200 million round last month, detailed in Mercury’s Series D announcement, at a valuation that jumped sharply as AI-driven company formation fed its growth.

• $5.2 billion valuation for Mercury, up 49% in roughly 14 months
• $650 million in annualized revenue, with four straight years of profitability
• $4.4 billion valuation for the drone maker Skydio after a $110 million Series F led by existing backers

Skydio’s raise was notable for its restraint. “The most significant fact in our Series F is how little we are raising,” chief executive Adam Bry said in Skydio’s Series F financing details, arguing that a strong core business now funds most of its own growth. The contrast with frozen unicorns is the point: the firms raising today often need the money least. The same gravitational pull is visible in the public-market pipeline, where Anthropic’s confidential IPO filing signals that investors are reserving their biggest bets for AI-first names.

Mercury’s chief executive, Immad Akhund, sees the split plainly from his vantage point banking thousands of young companies. “All the attention’s on AI, so if you’re not an AI-first company, you need really strong numbers to raise,” he said.

The Exit Math Founders Now Face

For a stranded unicorn, the realistic outcome is no longer a triumphant IPO or a richly priced acqui-hire. It is a sale at a fraction of the money investors put in. Two recent deals show the new discount in hard numbers.

In February, the investing app Stash was bought by the Singapore-based super-app Grab at an enterprise value of $425 million, below the roughly $660 million backers had poured into the company; the terms are laid out in Grab’s acquisition of Stash Financial. The same month, the savings startup Step sold to the YouTube star MrBeast for an undisclosed sum that investors suspect ran well below the roughly $500 million it had raised.

Ryan Falvey of Restive Ventures, which backs fintech firms, frames the repricing in multiples. Valuations have compressed about sixfold from the 2021 peak of 50 times forward revenue, he said, meaning a company with identical revenue is worth roughly 85% less than it was five years ago. Kaul puts the survival test more bluntly.

The question I ask every time one of them presents is, why can’t OpenAI, Anthropic or Google do this? For most of them, the answer is, they can.

The dominoes, as PitchBook’s Akers put it, are only beginning to fall. If outcome-based pricing and AI-native architecture become the price of admission, the back half of the fallen-unicorn list spends 2026 choosing between a discount sale and a slow fade. If a handful instead prove they can rebuild from scratch, they will reset what a survivable pivot actually looks like, and the rest will be measured against them.