The problem with hypergrowth AI startups

Many hypergrowth AI startups have achieved rapid revenue growth by reselling inference from large model providers at very low or negative margins. This business model is precarious because the value provided is often a thin layer on top of the resold tokens, making it vulnerable to competition. As AI tokens become commoditized and enterprises increasingly want to bring their own inference (BYO), these startups will face significant downward pressure on their revenue. This could put companies that raised capital based on this top-line growth in a difficult position, suggesting a potential return to more traditional, high-margin business models.