Not long ago, we were laughing at open-source scripts for meaningless burning of AI tokens for corporate KPIs and studying the unbreakable economics of this whole generative sandbox.
Well, reality has set in — investors have sobered up, and big tech has started cutting costs.
1️⃣ Uber: Tokens Don't Convert to Features
Uber's COO stated outright that it's becoming harder to justify AI spending. In April, it turned out that Uber engineers had already burned through the entire annual budget for Claude Code.
And guess what brilliant conclusion management reached? It turns out that the exponential growth of burned tokens does not correlate at all with the growth of useful business features.
Seems obvious: if you attach a powerful LLM to bad processes, you just get very expensive bad processes. But businesses needed to spend millions to realize that unlimited API access doesn't make a developer 5x more productive at the snap of a finger. To offset these AI losses, Uber is now slowing down hiring.
By the way, the corporate craze with "tokenmaxxing" is also winding down. Duolingo, for instance, has removed AI metrics from performance reviews.
2️⃣ Microsoft: Eat What You're Given
The second news is even funnier. Microsoft is mass-revoking Claude Code licenses from its own engineers and forcibly migrating everyone to its own GitHub Copilot CLI.
The reason is trivial: June 30 is the end of their fiscal year. Paying Anthropic bills for thousands of developers turned out to be too painful for operating expenses. Never mind that engineers objectively liked Claude Code more, and Microsoft itself is Anthropic's largest partner. Cost savings trump convenience.
And the "good" news is that Russian companies generally haven't even managed to actively integrate AI into development yet, and it's already being rolled back. We didn't live well, so why start? 😅
Comments
0No comments yet.
Sign in to join the discussion.