Kevin Surace explains why companies mandating AI for every task faced shocking bills when providers switched to per-token pricing. A single user racked up $100,000 a month; heavy AI usage costs two to three times a developer's salary. The math flipped: AI adoption became more expensive than hiring engineers.
Companies told their employees to use AI for everything. Then the bill came in and some of them nearly had a heart attack. Here's the mistake they're making. Just in the last couple of months, you start to see these companies who all push their employees, use, use, use, use, use. We're going to measure you on usage. Then they got the bill and all of a sudden they had $120 million token bill from Anthropic. They go, whoa, we could have hired engineers for that. And it turns out that at some point, you know, each coder or each developer is using 2 or 3 times their salary in tokens. And you go, I thought they're supposed to save money. And what happened? A couple things happened. One, people finally figured out how to use it for everything. And the second thing is, is that all of the Frontier models stopped doing a $20 a month eat-all-you-can model, right? And they go, we're gonna go broke doing this. So they started charging by the token, and all of a sudden you had a $100,000 a month bill for a single person instead of a $20 a month bill, and everybody choked.
