Presumably they'll aim for IPO or getting acquihired before then. Most startups these days aren't meant to build anything lasting, they're designed to line up an exit from day zero. It's an established, mature business model now.
The whole business model is poisonous and extractive. 90% of the businesses launched aren't valid, and then the few that are get bought up by oligopolists/monopolists.
I'd bet on them. Brand and models will get cheaper. Can engineer to use cheaper open models. Secret sauce can be in that context engineering and becoming a "cloud" themselves to get cheaper inference.
How is that different than literally any other SaaS raising big rounds? Profitability is apparently only important for AI companies, and even being more profitable than most SV startups since the inception of this site is not good enough.
Regular SaaS businesses are not capital intensive once launched, they're mostly pure software plays. Their marginal user cost trends towards 0.