Uber president says AI spending is getting ‘harder to justify’

The Verge reports that Uber, with only four months elapsed in 2026, has reportedly already exhausted its annual AI budget, prompting the company internally to begin questioning whether this spending truly delivers sufficient returns. Uber President and COO Andrew Macdonald said in a Rapid Response interview that the company currently does not see rising Claude Code token consumption that can be clearly linked to more meaningful outcomes. In other words, increased usage of AI tools does not necessarily mean that development efficiency, product output, or business value rise in lockstep. The key point of this news is not that Uber is abandoning AI, but rather that large enterprises, as generative AI adoption enters its second phase, are beginning to scrutinize the relationship between cost and output more rigorously. For development teams, coding-assistance tools like Claude Code may indeed speed up certain tasks, but without clear metrics—such as shortened delivery times, lower defect rates, higher engineer productivity, or improved operational efficiency—rapidly rising token costs become financial pressure. For investors and management, this also shows that AI budgets are no longer merely strategic investments or symbols of innovation, but need to be scrutinized like other cloud, SaaS, or labor costs. For Taiwanese readers, this article serves as a reminder that when enterprises adopt AI, they cannot look only at whether employees are using tools heavily, nor treat token usage as a success metric; a more practical approach is to establish trackable workflow metrics that distinguish which tasks truly benefit from AI and which merely shift costs to model inference fees. Overall, Uber's remarks reflect that the early enthusiasm for generative AI on the enterprise side is gradually cooling, and the next focus will be governance, budget control, and quantifiable ROI.