For years, artificial intelligence was marketed as the ultimate cost-cutting tool—a way to empower employees while simultaneously slashing operating expenses. Now, a different reality is emerging: at some companies, the cost of operating AI systems has surpassed salary expenditures, overturning long-held beliefs about the economics of automation, according to Axios.

According to the publication, corporate IT budgets are "bursting at the seams" as AI infrastructure spending overtakes payroll in several firms. The article cited Bryan Catanzaro, Nvidia’s VP of Applied Deep Learning Research, who noted that for his team, "the cost of compute is significantly higher than the cost of the employees." This admission is particularly notable coming from an executive at the company whose GPUs power the majority of the world’s AI tasks.

Uber Exhausts Entire AI Budget in Months

This trend has already led to a high-profile budget crisis. Uber’s CTO, Praveen Neppalli Naga, told The Information that the company burned through its entire 2026 AI budget just months into the year—driven by the rapid adoption of Anthropic’s Claude Code among the company’s engineers. "I have to start over because the budget I expected to use is already gone," Naga stated.

The overspend wasn't due to a failed project; on the contrary, the tools proved too effective. After Uber granted access to Claude Code to approximately 5,000 engineers in December 2025, usage nearly doubled by February. By March, 84% of developers were categorized as agentic programming users, and about 70% of recorded code was AI-generated. Roughly 1,800 code changes were being submitted weekly without direct human intervention. The cost issue stems from per-token pricing: expenses skyrocket as engineers run parallel agents and perform full codebase refactors, driving monthly API costs to between $500 and $2,000 per engineer.

A Trillion-Dollar Shift

The broader spending landscape reflects these same trends. According to Gartner's latest quarterly forecast released in late April, global IT spending is set to reach $6.31 trillion in 2026, a 13.5% increase from the previous year. Data center spending alone is projected to grow by 55.8%, approaching $788 billion—fueled by demand for compute power and high-performance processors for AI tasks.

The Question of Sustainability

Skyrocketing costs are forcing executives to ask: is mass AI adoption financially justified? Despite a $3.4 billion R&D budget, Uber failed to anticipate how quickly the technology would be embraced. Catanzaro, whose own teams at Nvidia struggle to secure enough GPUs amid global shortages, put the problem simply: "We are all supply-constrained." For companies racing to embed AI into their processes, the uncomfortable arithmetic is becoming clear—machines may be productive, but they don't come cheap.