For the record, this is happening all over the AI space. Usage limits are tightening significantly, not only for free users but paid plans too. The major players are upping their costs per token, reducing token limits on subscription plans, and increasingly talking about moving all of their customers to paying direct inference costs instead of subscriptions.

This will kill the industry, and they don’t have a way out of it. Regular consumers aren’t going to pay per token. You think about the people you know or hear stories about, the people who love ChatGPT because it “listens to all their problems” and “gives them real advice.” The people who use it as a therapist or ask it what to eat at a restaurant. Now imagine those people having to mentally budget every conversation they have with an LLM. No way. Not a chance.

And business users are looking at their skyrocketing AI bills, with zero meaningful productivity gains on the other side of the balance sheet, and they’re calling bullshit. It’s doing nothing for them and they’re paying outrageous prices for the privilege. And right now those prices are still subsidized.

This has always been the fatal flaw in the entire industry. In the entire concept. They could have survived the hallucinations, the low quality output, the scandals and deaths and AI psychosis, if only the damn thing was cheap to run. But the cost is what will kill them.