My company is implementing max cost/day on LLM token, has AI usage peaked?

I work at a big tech but not FAANG company, you've very likely used our product this week. We were told to go hard on AI and the majority of the devs I know, including me, don't write code anymore and only review the output of agents. It even became so ingrained in the culture I noticed people wouldn't do anything themselves even "this code looks good, commit and push" so literally spending tokens instead of running git commit, git push. We have internal AI usage dashboards and some people were spending $10k/week in Claude tokens. I always wondered if this was sustainable, if all devs did that the company would be bankrupted. We got the first sign that the powers that be have also noticed as we've been set a limit of $750 per week, meaning many are going to have to adjust their workflows. Could we actually be at the peak of AI usage now? And as tokens become more expensive and cost caps come in, we actually see a return to writing code?