AI IPOs to face net revenue retention and gross margin scrutiny, warns Leo AI CEO Maor Farid

Maor Farid's argument is rooted in the boom-and-bust pattern he witnessed firsthand in 2021, when AI companies could raise money on the promise of AI alone but often struggled with retention. Many of those products were generic—traditional software with an AI veneer—or saved time on tasks nobody's job depended on. A 20% improvement on a noncritical task, he notes, makes for a nice demo but a dead renewal. The market, in his view, is now correcting for that, and companies delivering meaningful value are growing faster than anything he has seen in enterprise software.

The standards he expects from public investors are already reshaping his own fundraising conversations. A few years ago, $1 million in ARR was a milestone, but in the AI era a new product can reach that level in about a year and shut down a year later. As a result, ARR alone no longer tells investors enough. What they keep asking about is expansion: whether customers increase spending after the initial deployment. That, Farid suggests, is the closest signal that a product has become embedded in how an organization works.

For founders, the read is that defensibility now rests on domain expertise and proprietary context—knowledge buried in old drawings or in the heads of experienced employees that foundation models cannot access. Companies that are thin layers over someone else's model may find their next funding round much harder than the last. The test ahead is whether public markets actually enforce these standards, but if they do, private investors are likely to follow.