Uber just proved that telling thousands of engineers to use AI “as much as possible” without a spending cap is the corporate equivalent of handing your teenager a credit card and saying “have fun.”
The rideshare giant burned through its entire 2026 AI coding budget in just four months. Let me repeat that timeline for clarity: a twelve-month budget, gone in four. The company has now slapped a $1,500 monthly cap per employee per agentic coding tool. And somewhere in Uber’s finance department, someone is updating their resume.
What Actually Happened
According to reports from Bloomberg and TechCrunch, Uber had previously encouraged its staff to use AI tools extensively. The company wanted to be at the front of the AI adoption curve, pushing employees to integrate these tools into their daily workflows. Standard Silicon Valley playbook — move fast, figure out the costs later.
The problem? They figured out the costs later. And those costs were astronomical.
Uber’s CTO reportedly said he’s “back to the drawing board,” which is executive-speak for “we did not see this bill coming.” The kicker? The tool that apparently did the most damage costs around $200 a month per seat. But when you factor in token consumption from agentic coding tools — where AI agents run autonomously, burning through API calls and compute cycles — that $200 baseline becomes a rounding error.
The $1,500 Cap Tells You Everything
The new monthly limit of $1,500 per employee per AI coding tool is interesting for a few reasons. First, it confirms that some employees were spending far more than that. Possibly multiples of that figure. When agentic tools run loops, retry failed attempts, and generate massive context windows, token costs can spiral in ways that traditional SaaS subscriptions never did.
Second, it reveals a fundamental misunderstanding that many companies still have about AI tool costs. This is not like buying a Jira license. AI tools — especially agentic ones — have variable costs that scale with usage intensity. An engineer who runs an AI agent for eight hours straight on a complex refactoring task will consume dramatically more resources than someone using it for autocomplete suggestions.
Third, and this is the part that interests me most as someone who reviews these tools daily: $1,500 per month is still a generous allowance. Most individual developers spending their own money on AI coding tools are working with $20 to $200 monthly budgets. Uber is still spending heavily — they’re just no longer spending recklessly.
This Is Going to Happen Everywhere
Uber is not unique here. They’re just the first major company to publicly face the consequences of uncapped AI spending. Every enterprise that has encouraged broad AI adoption without usage governance is sitting on the same time bomb.
The pattern is predictable:
- Company announces bold AI adoption strategy
- Leadership tells employees to experiment freely
- Finance discovers that “freely” has a very specific dollar amount attached
- Caps get introduced with corporate messaging about “responsible usage”
We’ll see this play out at dozens of companies over the next twelve months. The agentic AI tools that are gaining popularity right now — the ones that run autonomously rather than waiting for prompts — are particularly expensive to operate at scale. They’re useful, but they eat tokens like a GPU cluster eats electricity.
My Take
I review AI tools for a living, and I’ve been warning about this exact scenario for months. The pricing models for agentic coding tools are designed to look cheap at the individual level while becoming budget-destroyers at enterprise scale. A tool that costs $200 per seat can easily generate thousands in token costs when an agent is running autonomously across large codebases.
Uber’s mistake wasn’t adopting AI tools. It was treating variable-cost infrastructure like fixed-cost software licenses. That’s a CFO problem, not an engineering problem. The engineers did exactly what they were told — they used AI as much as possible. Turns out “as much as possible” meant burning a year’s budget in a quarter.
The $1,500 cap is a reasonable correction, not a retreat from AI. But it should serve as a warning to every company currently in the “encourage unlimited usage” phase: you need usage monitoring and spending alerts before you need a budget postmortem.
Uber learned this lesson publicly. Everyone else gets to learn it from watching.
🕒 Published: