Matthias Balk | Image Alliance | Getty Photographs
5 years in the past, enterprise capitalists had been pouring cash into American startups promoting the whole lot from lingerie subscriptions to scheduling software program, anointing them with billion-dollar valuations earlier than most even turned a revenue.
It was a frothy period for startups, fueled by a mix of low cost cash and pandemic-boosted demand. However even after the Federal Reserve took some froth off by beginning to increase rates of interest in 2022, many founders believed that they might develop into their inflated valuations, buyers advised CNBC.
Then, an app known as ChatGPT arrived.
“The ChatGPT second was when folks stated, ‘Holy smokes, the following era of entrepreneurs, their coding language is spoken English,'” stated Samir Kaul, a associate on the enterprise agency Khosla Ventures, an early backer of OpenAI.
“Now you are seeing 50 engineers do what it might’ve taken 500 engineers to do 5 years in the past,” Kaul stated. “We needed to utterly reshuffle how we valued these firms.”
Whereas the shares of public software program firms like Salesforce, ServiceNow and Workday acquired hammered this yr due to the menace from synthetic intelligence, a quieter reckoning has been unfolding within the non-public markets.
The AI increase that funneled greater than $250 billion into OpenAI and Anthropic forward of their anticipated mega-IPOs this yr has left a whole bunch of startups constructed earlier than ChatGPT’s arrival in 2022 stranded — successfully lower off from enterprise funding due to their inflated valuations and outdated know-how, but not worthwhile sufficient for the general public markets.
There are 857 U.S. startups valued at $1 billion or extra, the edge for being deemed a “unicorn” firm, based on PitchBook knowledge. However practically half of that group hadn’t raised contemporary funding within the final three years as of the top 2025, making lots of these valuations stale, based on the non-public markets knowledge agency.
Startups that final raised in 2021 had been price 68% much less on common on the finish of final yr, whereas those who final raised in 2022 noticed a 52% decline, based on Pitchbook’s personal valuation estimates.
In consequence, greater than 220 firms that had reached billion-dollar valuations within the enterprise increase had been deemed fallen unicorns, based on PitchBook, which offered a listing of the businesses completely to CNBC. The estimates are based mostly on components together with head depend progress and comparisons with public firms.
“Numerous these firms are pre-AI, not simply of their value construction, but additionally of their merchandise,” Mercury CEO Immad Akhund advised CNBC. His firm, which raised $200 million in funding final month, supplies banking providers to a 3rd of early-stage U.S. venture-backed companies.
“They’re positively in a tough spot,” he stated. “All the eye’s on AI, so for those who’re not an AI-first firm, you want actually robust numbers to lift.”
Glossier, Brooklinen, AG1
The listing of fallen unicorns consists of well-known manufacturers like Glossier, The Farmer’s Canine, Rothy’s, Brooklinen and Savage X Fenty, the lingerie firm based by musician Rihanna. The businesses had been a part of a wave of direct-to-consumer companies constructed on the hope that digital retailers may earn software-like margins.
Additionally included are mainstays of podcast commercials together with the powder complement maker AG1 and the robo-advisor pioneer Betterment in addition to the net ticket market SeatGeek.
These firms got here of age in an surroundings that rewarded progress at nose-bleed valuations based mostly on two broad assumptions: rates of interest would stay low and a startup may at all times be acquired for its engineering expertise.
However the arrival of generative AI has redrawn the enterprise panorama, redirecting capital towards AI-native companies whereas making it unimaginable for a lot of older startups to justify their earlier valuations.
Hit hardest are enterprise software program firms like scheduling startup Calendly, which signify the only largest class among the many fallen unicorns. There are 75 software-as-a-service, or SaaS, companies showing on PitchBook’s listing, which is double the variety of fintech firms, the next-biggest group.
That displays each the large valuations that software program startups commanded throughout the 2021 enterprise increase and the diploma to which generative AI has destabilized assumptions underpinning the sector.
David Zhu, an ex-DoorDash head of engineering, stated that after the “ChatGPT second” he appeared throughout the software program panorama — from startups to medium-sized companies funded with non-public credit score to the biggest public SaaS firms — and noticed a seismic shift on the horizon.
“The thesis I had was that every one workflow-driven enterprise SaaS firms might be both disrupted or lifeless within the subsequent decade,” Zhu advised CNBC.
The SaaS mannequin, the place firms embed themselves in worker workflows and sometimes cost by the consumer, is very threatened by the rise of autonomous brokers. After leaving DoorDash, the place he led greater than 200 engineers, Zhu based Reevo, an AI platform that automates company gross sales and advertising groups.
Firms constructed earlier than generative AI are weighed down by bloated staffing fashions and software program designed for a pre-AI world, based on Zhu, making it laborious for them to remodel themselves.
“Until they make a stark, 180-degree pivot to rebuild the very same factor from scratch, they will slowly fail,” Zhu stated. “What which means is that buyers would moderately simply wager on new entrepreneurs at decrease valuations moderately than double down on older startups.”
‘Dominoes to fall’
A lot of the 20 firms highlighted by CNBC both did not reply to a number of requests for remark or declined to remark.
A spokesperson for the drone maker Skydio — estimated by PitchBook to have dropped in worth from $2.5 billion to $509 million as of December 2025 — stated in a press release, “This third-party hypothesis is fake and never based mostly on Skydio’s operations or the exponential progress we’re seeing in income and prospects.”
In April, Skydio introduced that it had raised $110 million via present buyers, elevating its valuation to $4.4 billion.
An AG1 spokesperson did not present a press release for this text, however after CNBC’s inquiry, Reuters reported that the complement maker was trying to promote half or the entire firm at a $2 billion valuation. That determine would come with AG1’s debt, the report stated.
After publication of this text, a spokesperson for The Farmers Canine stated that the PitchBook estimate was incorrect and that the corporate’s valuation “has solely elevated since our final funding spherical in 2022.”
If an organization hasn’t raised funding since 2021 or 2022, its unlikely it will ever accomplish that once more, say buyers and founders. With out entry to enterprise funding or a believable preliminary public providing ramp, the most certainly exit for a lot of fallen unicorns is an acquisition at a fraction of their outdated valuation, they are saying.
“Once we see firms not elevating, it is a crimson flag,” stated PitchBook analyst Andrew Akers, including that it often means their progress is tepid and even destructive.
Whereas some startups may’ve averted fundraising as a result of they’re producing sturdy income, that’s the exception to the rule, he stated.
“Beneath the floor, I believe there are a variety of dominoes to fall,” Akers stated.
Collapsing flooring
There have been glimmers of a reset amongst some startups this yr.
In February, Stash, the funding and financial savings app, was acquired by Singapore-based the whole lot app Seize at an enterprise worth of $425 million, under the roughly $660 million that buyers put into the corporate throughout its lifetime.
That very same month, one other fintech, Step, was acquired by the YouTube star MrBeast for an undisclosed quantity, main buyers to take a position that the acquisition worth was far under the approximate $500 million the startup raised earlier than the deal.
“Many of those companies simply aren’t price that a lot anymore, which is why you are seeing them get acquired at steep reductions,” stated Ryan Falvey of Restive Ventures, which invests in fintech companies.
Valuations have compressed by about sixfold from the 2021 peak of fifty instances future revenues, that means that an organization with the identical income is price about 85% much less in at the moment’s market than 5 years in the past, Falvey advised CNBC.
Earlier than the reset, a startup may typically be offered to a bigger know-how firm trying to purchase the smaller agency’s engineers for roughly $2 million per coder, based on Khosla Ventures’ Kaul. A agency with 100 engineers could be price at the very least $200 million to $300 million, he stated.
However that assumption, which offered a flooring underneath startup valuations throughout the increase, evaporated after AI coding instruments allowed far smaller groups to construct merchandise — leaving exit alternatives few and much between.
‘OpenAI, Anthropic or Google’
The result’s that post-GPT startups are operating laps round their older opponents, based on Falvey. He known as investments revamped the previous three years “undoubtedly one of the best” his agency has made.
“We seen by 2023 that the businesses we invested in post-ChatGPT had been already making extra money than a lot of the firms we invested in earlier than ChatGPT,” Falvey stated.
Generative AI could in the end scale back the quantity of capital required to construct profitable software program firms, difficult one of many core assumptions that fueled the enterprise increase of the previous decade.
The shakeout might be simply starting, because the affect of AI reverberates throughout the enterprise funding ecosystem, from enterprise to non-public credit score to public giants.
Older software program companies, Kaul stated, nonetheless depend on enterprise fashions constructed round charging prospects based mostly on the variety of staff utilizing their merchandise, an strategy he believes AI will undermine as firms automate extra white-collar work.
Software program suppliers might want to shift towards outcome-based pricing fashions and AI-native infrastructure to outlive, he stated.
“The query I ask each time considered one of them presents is, why cannot OpenAI, Anthropic or Google do that?” Kaul stated. “For many of them, the reply is, ‘They’ll.'”
