91探花

Reports from earlier this month say that Uber spent a year鈥檚 worth of its AI budget in just four months, according to Bloomberg. This comes after the e-hailing and delivery platform announced that it had been using tools such as Claude Code and Cursor to help run and speed up the company鈥檚 code.

Praveen Neppalli, Uber鈥檚 CTO, explained in an earlier tweet:

鈥淎gentic software engineering adoption is on fire at Uber. 1,800 code changes per week are now written entirely by Uber’s internal background coding agent, and 95% of our engineers now use AI every month across all the tools we track.

鈥淭his is a real reset moment for engineering; it’s one of the most exciting times to lead. This shift requires builders to be curious and hands-on. I鈥檓 incredibly lucky to be surrounded by a team that鈥檚 doing exactly that.

鈥淭he best part is that the strongest adoption isn鈥檛 being pushed top down from leadership announcements; it鈥檚 coming from engineers who are quietly experimenting, quietly shipping, and quietly pushing things forward.

鈥淚 love spending time with those engineers because there鈥檚 no substitute for being close to the work.

鈥淥ver the last few months, we leaned in hard, and the results have been phenomenal.

鈥淭he bigger shift: going agentic.

鈥�84% of AI users are now working with agent-style workflows, not just tab completion. Claude Code usage nearly doubled in 2 months (32% 鈫� 63%), while IDE-based tools have largely plateaued.

鈥淓ngineers are moving from accepting suggestions to delegating tasks. Even within traditional IDEs, ~70% of committed code is now AI-generated.

鈥淏ackground agents are writing code autonomously.

鈥淥ur internal background coding agent went from <1% of all code changes to 8% in just a few months. There is zero human authoring. Engineers review and approve, but the code is written entirely by AI agents.

鈥淭he role of the engineer is shifting – from writing every line to architecting systems and reviewing AI-generated code.

鈥淢ore to come from the [Uber engineering] team in the coming days.鈥�

 

Introducing The Employee Cap: What Does This Mean For Uber?

 

As a result of the budget being blown, the company decided to introduce a $1,500 monthly cap per employee, per platform.

In other words, employees are given that amount on each platform to control how much is spent on AI. This is monitored through a dashboard, and when the cap is hit, there are options to request more funds. While this doesn鈥檛 necessarily restrict spending, the company believes it鈥檒l help keep responsible use in check.

The company told Bloomberg, 鈥淲e think this is all a pretty straightforward way to responsibly encourage agentic AI adoption and experimentation at scale across the company.鈥�
 

 
Sam Wilson, an AI expert calculated how much $1,500 per employee really is compared to his own spending habits, for context. He wrote on his weblog:
 
鈥淚f we assume two actively used tools per engineer that’s $3,000 * 12 = $36,000 cap per engineer per year. Levels.fyi lists the median yearly compensation package for Uber software engineers in the USA at $330,000.

鈥淭hat means each employee’s AI spending cap is ~11% of that median compensation package.

鈥淚 noted that my own token usage comes to about $1,000/month against each of Anthropic and OpenAI – which currently costs me just $100 per provider thanks to their generous subsidised plans for individual subscribers. Those plans are no longer available to larger companies like Uber.

鈥淭heir new policy means if I were working at Uber I’d still have ~$500/month of tokens to spare for each of those tools, given my current usage patterns.鈥�

The Financial Times also reported that companies like Amazon, Walmart, Cisco and Meta have started limiting employee AI use as costs put a strain on budgets. Which brings us to the next point鈥�

 

Are Companies Spending Too Much On AI?

 

The argument here isn鈥檛 as simple as 鈥渃ompanies are overspending on AI鈥�. Jobs have been cut and hiring has been slowing as more companies make room for AI in their budgets (which has cost people their jobs), but the idea that AI helps companies save on labour costs doesn鈥檛 quite stand when AI budgets are being burned.

This news makes me wonder whether companies are starting to spend more than they can measure when it comes to AI鈥檚 return on investment specifically. We鈥檝e spoken about how companies might be aimlessly giving into the AI hype by rushing to try out many different tools while they admit to seeing no long term gains. Using these tools becomes redundant when no actual results are seen in the long run.

Martin Reynolds, Field CTO at Harness, also sees it this way. He said, 鈥淯ber capping AI spend is treating the symptom rather than the cause. The real issue is that many organisations are still measuring AI success through consumption rather than outcomes.

鈥淭racking AI usage made sense when adoption was the priority and engineering teams were sceptical. But many organisations never evolved beyond that stage. Now, with AI embedded in daily workflows, usage metrics start to distort behaviour.

鈥淓mployees are rewarded for generating more prompts, tokens, and model interactions – regardless of whether those activities create meaningful business value. As a result, business leaders can鈥檛 confidently say whether AI usage is driving results or simply inflating the bill.

鈥淎 $1,500 monthly cap doesn鈥檛 do much to tackle the root of the issue. Organisations need to build the type of cost visibility and tagging infrastructure that ties every AI query to a work order, a feature, or a business outcome – highlighting whether it genuinely contributed to customer experience or revenue growth.

鈥淭hat’s how you make AI economically intelligent. Until organisations can measure value, not just spend, they’ll keep swinging between uncapped excess and blunt rationing – and leaving the real ROI on the table.鈥�