AI Is Killing the Unicorns: 220+ Pre-ChatGPT Startups Are Now Worth Half Their Peak Value

In November 2022, when OpenAI released ChatGPT to the public, the venture capital industry divided into two eras without anyone fully realizing it. There was before, and there was after. Three and a half years later, the financial consequences of that division are becoming impossible to ignore.

A new analysis finds that more than 220 companies that once commanded billion-dollar valuations — the unicorns that defined the 2021 boom — have now seen their estimated worth fall below 50% of their peak. Nearly half of all 857 US-based unicorns haven’t raised a new financing round in three or more years, a period of dormancy that, in a rapidly shifting technology environment, functions as a slow-motion valuation reset.

The Numbers Behind the Collapse

The magnitude of the decline correlates tightly with timing. Companies that last raised capital in 2021 — when the combination of near-zero interest rates and post-pandemic digital acceleration pushed valuations to historic highs — have seen an average value erosion of 68% from their peak. Those that last raised in 2022, just before ChatGPT launched and the rate cycle turned, have declined 52% on average.

Several companies illustrate the scale of individual falls. Articulate, which built corporate training and e-learning software and raised at a $3.75 billion valuation, is now estimated to be worth $683 million — a decline of roughly 82%. Calendly, the scheduling software that became ubiquitous during pandemic-era remote work and raised at $3 billion, is now valued at approximately $793 million, down 74%. Chegg, once worth over $14 billion as an education technology platform, has seen its public market value fall below $150 million — nearly a 99% destruction of peak value.

Consumer brands that happened to carry unicorn valuations fared little better: Glossier, Brooklinen, and Rothy’s all appear on what analysts are calling the “fallen unicorn” list, alongside marketplace businesses and consumer fintech platforms that raised during the zero-rate era on projections that now look like fantasy.

The Structural Cause: AI Changes the Unit Economics of Software

The headline numbers are dramatic, but the underlying dynamic is structural rather than cyclical. This is not simply a correction in overvalued growth stocks; it is a technology disruption that is systematically altering the cost structure of software-building and therefore the addressable market for software-selling.

In the pre-ChatGPT era, building enterprise software required large engineering teams to develop features, manage integrations, and handle the edge cases that enterprise customers invariably discover. A typical Series B SaaS company might employ 150–300 engineers to maintain a competitive product. The high cost of that labor was precisely what justified high valuations: barriers to competition were real, and customers were paying for the moat that engineering talent created.

AI is dissolving that moat. “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. The implications cascade through the entire software ecosystem. Smaller teams building AI-native products can enter markets that previously required substantial capital and headcount to penetrate. Products that previously took two years to build can be built in two months. And customers are increasingly choosing AI-native alternatives over incumbents whose workflows were designed around human execution rather than AI augmentation.

The category most visibly disrupted is enterprise SaaS. Seventy-five SaaS companies appear on the fallen unicorn list — roughly double the number of fintech firms in the same category, the next-largest cohort. The pattern makes sense: SaaS businesses typically built value through workflow lock-in, integrations, and the switching costs that came from having thousands of employees trained on a specific product. AI assistants that can replicate those workflows with a natural-language interface, or AI-native competitors that rebuild the product from scratch with a fraction of the headcount, attack the lock-in directly.

The Capital Concentration Problem

The valuation collapse among pre-ChatGPT companies is inseparable from a dramatic concentration of venture capital into AI. In the first quarter of 2026 alone, AI startups raised $255.5 billion globally — a figure that would have been the full year’s total global venture investment just four years ago. AI deals accounted for 81% of all venture funding during the quarter, leaving roughly 20% of capital for everything else.

For founders who raised in 2021 and need to raise again in 2026, the arithmetic is brutal. They are competing for a smaller slice of non-AI capital against a larger field of competitors, at valuations that no longer have the support of the zero-rate environment that made their original rounds possible. Many are being forced to take “down rounds” — new financing at lower valuations — that dilute earlier investors and trigger negative press cycles that further dampen investor interest.

Others are choosing a different path: restructuring to incorporate AI, pivoting their product positioning, or quietly pursuing acqui-hire exits to larger companies that want their engineering talent without the complexity of a full acquisition. Former DoorDash engineer and startup founder David Zhu put it directly: pre-AI workflow SaaS companies face a binary choice — adapt or become obsolete within the decade as AI-native competitors arrive with smaller teams, lower burn rates, and products that can outperform incumbents on core functionality within months of launch.

The Venture Ecosystem Restructures Around AI

The concentration of capital into AI is producing a venture ecosystem that looks structurally different from anything that preceded it. The median Series A round for an AI startup is now $51.9 million — approximately 30% above the equivalent figure for non-AI companies. Mega-rounds above $100 million, once reserved for late-stage growth rounds, are increasingly appearing at Series A and even seed stage for companies with convincing AI theses.

Investor priorities have realigned around three categories: infrastructure plays — model tooling, custom silicon, compute management — that benefit from the AI capex wave regardless of which application layer wins; enterprise AI products with clear, measurable budget lines that replace identifiable headcount; and what investors are calling “category leaders” — companies building the standards, protocols, or platform primitives that an entire segment of the AI stack will eventually be built upon.

Companies that don’t fit neatly into one of those categories are finding it difficult to attract institutional capital, regardless of their historical traction. The market has effectively bifurcated: extraordinary capital availability for AI-native companies with the right narrative, and a significant funding drought for everything else.

What Comes Next

The divergence is unlikely to narrow quickly. AI’s productivity gains in software development will continue to compound, making it cheaper to build new products and more expensive to defend the moat of existing ones. The 220-plus companies on the fallen unicorn list are not all going to zero — some will successfully adapt, some will be acquired, some will find sustainable niches. But the era when building software at scale was sufficient to justify unicorn valuations is over.

The investors who funded those valuations are facing their own reckoning. Venture funds raised in 2021 at the peak of the boom are now managing portfolios full of companies whose values no longer support the return projections they made to their limited partners. The next few years will see a wave of secondary sales, structured liquidity transactions, and fund-level write-downs that have been delayed by the difficulty of marking private company values to market.

For the founders navigating this environment, the lesson is not that ambition was misplaced — it’s that the technology context in which that ambition operates changes faster than any individual company’s roadmap. The companies that last raised capital before November 2022 raised in a different world. The one they are operating in now is, by almost every relevant measure, a different industry.