Why Uber Burned Through Its 2026 AI Budget in Just Four Months

Uber has spent its entire year-long artificial intelligence budget in only four months, according to company president Andrew Macdonald. In an interview with Rapid Response, Macdonald said the ride-hailing company is struggling to show that all the money it's pouring into AI — particularly a tool called Claude Code — is actually making the platform better for riders or drivers.

Uber spent $3.4 billion on research and development in 2025, a 9 percent jump from the year before. The company's chief technology officer had already warned in April that the budget set aside for Claude Code usage alone was exhausted for 2026. Now Macdonald's comments confirm the problem is even wider.

The Core Issue: Spending Without Clear Results

Macdonald told the publication that despite using far more AI than expected, Uber hasn't found a clear link between that increased usage and new features or improvements that actually benefit customers. Think of it this way: the company is burning fuel, but it's unclear whether the engine is running better or faster.

This gap between how much a company spends on AI and what customers actually see is a problem across the entire tech industry right now. When you measure AI usage in "tokens" — essentially counting how many times the system processes information — you get a clear picture of cost. But figuring out whether that spending leads to better, more useful features is much harder to track.

Cutting Headcount to Fund AI

To offset these rising AI costs, Uber's CEO Dara Khosrowshahi has indicated the company is slowing down human hiring. This is part of a broader trend in tech: companies betting that AI systems will do work that humans used to do, so they need fewer people on staff.

Whether this trade-off makes sense depends partly on whether the AI spending eventually pays off. Right now, with Uber's AI budget already exhausted and unclear returns, that equation looks risky.

A Pattern We've Seen Before

This situation echoes something that happened with other big technology shifts. When companies rushed to build out the internet in the late 1990s, or when they raced to migrate to cloud computing a decade ago, the pattern was similar: heavy upfront spending followed by a harder look at whether the money was actually returning value. Eventually, companies got more disciplined about tracking results. Uber may be entering that rationalization phase now.

The Autonomous Vehicle Question

Adding pressure to all this is Uber's push to develop autonomous vehicles — self-driving cars. Macdonald has called this essential to Uber's future survival. But he's also been vague about the timeline, saying it could take anywhere from a couple of years to a couple of decades.

That kind of uncertainty matters because both the immediate AI improvements and the long-term autonomous vehicle research compete for the same pool of money and computing power. The budget crunch forces hard choices about which projects get resources.

Growing Beyond Rides

Separately, Uber is expanding into hotels and hospitality services. This diversification gives the company new sources of income if ride-hailing faces disruption — from autonomous vehicles or otherwise. But it also means the company needs to invest in new technology and platforms at the exact moment when AI spending is already outrunning expectations.

What Comes Next

For Uber, the immediate problem is figuring out how to fund the rest of 2026 while keeping development moving on both customer-facing features and self-driving research. The bigger question is whether the company can prove that its AI investments actually work — because the answer will likely determine how much money it's willing to spend on AI in 2027 and beyond.

Across the industry, enterprises are facing similar pressure. Building AI systems is relatively easy; showing that they deliver real value to users is harder. How Uber solves that problem will matter not just for the company, but as a test case for others wrestling with the same issue.