Uber blew its entire 2026 AI budget in 4 months — and one 2-hour coding session alone cost $1,200

Corporate America is pouring hundreds of billions of dollars into artificial intelligence. The exertion is expected to assistance companies prevention clip and money. But for immoderate businesses, simply paying for it is becoming a large expense.

Uber burned done (1) its full artificial quality fund for 2026 successful conscionable 4 months. One enforcement managed to rack up a $1,200 measure during a azygous two-hour coding session.

Must Read

• Jeff Bezos backs a level that lets anyone put successful rental homes for arsenic small arsenic $100 — 6 ways to physique wealthiness similar a landlord without really being 1

• Dave Ramsey warns astir 50% of Americans are making 1 large Social Security mistake — here's what it is and 3 elemental steps to hole it ASAP

• Millionaires nether 43 clasp lone 25% of their wealthiness successful stocks. Here's wherever their wealth is really going

The spending surge came aft Uber rolled retired Anthropic's Claude Code to thousands of engineers, who adopted the AI coding instrumentality overmuch faster than the company's fiscal models anticipated. It shows however rapidly AI costs tin escalate erstwhile companies promote wide usage of tools that complaint according to depletion alternatively than a predictable monthly price.

It besides raises a larger question facing firm America: Are the billions of dollars pouring into AI producing results that warrant the expense?

How Uber's engineers burned done the fund

Uber introduced Claude Code crossed its engineering enactment successful December 2025. By February, 32% of its engineers were utilizing agentic coding tools, but that fig had climbed to 84% by March. By the spring, 95% were utilizing AI tools monthly, portion astir 70% of committed codification originated from those tools, according to Forbes (1)⁠.

Claude Code tin bash much than suggest idiosyncratic lines of code. Engineers tin nonstop AI agents to enactment connected aggregate tasks, refactor ample sections of software, make tests and nutrient backend code, each of which tin devour important computing resources.

The mean monthly outgo reportedly ranged from $150 to $250 per engineer. Heavy users generated bills ranging from $500 to $2,000, Forbes reported (1)⁠.

Uber besides reportedly maintained interior leaderboards that ranked engineers according to their Claude Code usage. That encouraged employees to devour much AI resources adjacent though the teams promoting adoption weren't liable for controlling the budget.

The productivity gains stay hard to measure. Uber President and Chief Operating Officer Andrew Macdonald aboriginal said it was becoming harder to warrant rising token costs without grounds that they were producing much utile features for customers, according to The Verge (2)⁠.