Nvidia's Brian Catanzaro: AI costs now outstrip engineer salaries

The awkward economics here run counter to the promise that AI would replace expensive human labor. According to the article, companies are cutting staff to fund AI tools that, for now, cost more than the employees they dismissed — a reversal that would be comic if it did not threaten so many livelihoods.

The concrete examples pile up. Uber's engineering ranks moved fast: by March, 84% of its engineers were using Claude Code, and roughly 70% of committed code is now AI-generated. Yet COO Andrew MacDonald said publicly that token usage did not map directly onto useful features delivered to users. Microsoft, which has invested about $13 billion in OpenAI and writes up to 30% of its code with generative AI, reportedly told engineers in key divisions to stop using AI coding assistants. Axios reported that at one company, management forgot to set a usage cap, and a single month's Claude bill hit $500 million.

Nvidia sits at the top of the supply chain, and its message is blunt. Jensen Huang has said an engineer earning $500,000 a year should consume at least $250,000 worth of AI tokens annually, and the company is working toward a $2 billion annual token budget for engineers. Whether spending at that scale translates into products customers actually want appears to be the open question.