Burning Billions: Why Uber is Slamming the Brakes on Blind AI Spending

Silicon Valley’s favorite trend is hitting a massive financial wall. On Tuesday, May 26, 2026, Uber’s leadership publicly questioned the actual value of massive automation investments. In a remarkably candid interview, Uber president and chief operating officer Andrew Macdonald admitted that drawing a straight line between software token consumption and actual worker output is becoming incredibly difficult. After reportedly burning through its entire annual artificial intelligence budget in just the first four months of the year, the ride-sharing giant is starting to demand real financial accountability.

Macdonald made his position clear, stating that the company can no longer justify throwing millions of dollars at computing infrastructure without seeing concrete returns. Uber has integrated various machine learning systems, such as Anthropic’s coding assistants, to streamline operations and build new tools for passengers. Yet, despite massive spikes in server usage, the leadership team is not seeing a parallel jump in overall company performance or profit margins.

The Token Cost Explosion

The corporate pivot comes after a historic spending spree. Uber directed a staggering $3.4 billion toward research and development efforts in 2025, marking a 9% increase over its infrastructure spending from the previous year. Company chief executive officer Dara Khosrowshahi previously defended these heavy investments by cutting costs elsewhere, notably by slowing down human recruitment and hiring fewer office employees. The strategy centered on a bet that automated software could easily replace standard administrative and programming roles.

Now, that bet is looking highly unstable. Macdonald noted that while technical metrics look great on paper, the practical connection to consumer satisfaction remains invisible. He argued that the tech sector has focused way too much on tracking software token consumption as a metric for progress. In reality, this approach simply rewards systems that churn out massive mountains of code or text, running up eye-watering bills with cloud computing providers like Amazon and Google without improving the core product.

Shifting Focus to Real Value

To fix this financial drain, Uber is overhauling how it evaluates technology projects. Moving forward, the company plans to freeze any automated software implementations that fail to demonstrate an immediate reduction in operational costs or a clear boost in user retention. Macdonald wants to force engineering teams to pivot away from generic chat assistants and focus instead on shipping specific, practical features that users actively value.

The ride-sharing company is also rethinking its aggressive reduction in human headcount. While automated systems can handle predictable data tasks, they completely fail when navigating complex, real-world customer service crises or local regulatory disputes. Uber is sending a clear warning shot to the entire tech ecosystem: the era of writing blank checks for data centers is officially over. If a technology cannot prove it makes a business more profitable, it does not deserve a spot on the corporate balance sheet.