AI Crushes a Generation of Pre-ChatGPT Startups

The headline is a shakeout, but the load-bearing lesson is about where defensibility now lives. As capital and developer attention concentrate around a handful of frontier models, a polished feature built on a general-purpose model is the easiest thing for that model's next release to absorb. The startups surviving the reset tend to own something the model layer does not: proprietary or hard-to-collect data, deep workflow integration, regulated distribution, or a customer relationship that raw capability cannot dislodge. For founders and operators, that reframes roadmap and fundraising choices away from "train or wrap a model" and toward owning the data pipeline, the integration surface, and the channel.

CNBC reports that an AI investment surge directing more than $250 billion into OpenAI and Anthropic, ahead of their expected mega-IPOs, has left hundreds of startups founded before ChatGPT's 2022 arrival effectively stranded - cut off from new venture rounds because of inflated prior valuations, yet not profitable enough for public markets. Crunchbase counts more than 220 "fallen unicorns" that once cleared $1 billion valuations and no longer qualify.

PitchBook estimates, cited by CNBC, that startups which last raised in 2021 were worth roughly 68 percent less on average by the end of last year, and those that last raised in 2022 about 52 percent less. Crunchbase data shows how lopsided the inflows are: in Q1 2026 about $242 billion, or roughly 80 percent of venture funding, went to AI companies, with four mega-rounds alone accounting for about $188 billion - around 65 percent of all venture funding for the quarter.

David Zhu, a former DoorDash head of engineering, framed the risk bluntly, predicting that "all workflow-driven enterprise SaaS companies will be either disrupted or dead in the next decade." The argument is that seat-based SaaS pricing erodes as AI automates the white-collar work those seats represent. That is a forecast, not a verdict, and incumbents with entrenched data and distribution may adapt - but it captures why investors are repricing model-thin software.

Track venture flows by stage to see whether mid-stage rounds reopen for pre-2022 companies, watch acquisition announcements as platforms buy specialized teams and data, and monitor whether stranded firms pivot toward integration, verticalized data, or managed services rather than competing on raw model quality.