The numbers alone are enough to raise eyebrows. Uber, a company deeply integrated with artificial intelligence to power everything from ride pricing to route optimization, reportedly exhausted its entire 2026 budget for AI coding tools in just four months. This rapid consumption followed an internal initiative that incentivized employees to adopt the technology, even ranking teams based on their AI tool usage. Now, Andrew Macdonald, Uber’s president and chief operating officer, is openly questioning the tangible benefits of this accelerated spending, suggesting a disconnect between increased AI use and demonstrable improvements for consumers.
Macdonald articulated this concern during a recent interview on the Rapid Response podcast, stating that the link between rising AI expenditures, particularly on tools like Claude Code, and innovations directly serving the company’s vast user base is “not there yet.” He elaborated that while more features might implicitly be shipping, it remains challenging to draw a direct correlation between these statistics and a significant uplift in useful consumer features, perhaps a 25% increase, for example. This sentiment highlights a growing dilemma for enterprises embracing AI: the accelerating costs associated with its adoption, even as the per-unit pricing of AI tokens trends downward.
The rideshare giant is hardly alone in navigating this complex landscape. Microsoft, for instance, has reportedly begun scaling back direct Claude Code licenses, steering its engineers towards GitHub Copilot CLI instead. This shift suggests a broader industry reevaluation of AI investment strategies. Other business leaders, initially enthusiastic about AI’s potential, have also tempered their expectations. Luis von Ahn, CEO of Duolingo, previously held a bullish outlook but has since revised his stance, indicating he no longer foresees AI replacing tasks performed by his employees. This evolving perspective underscores the ongoing search for clear, quantifiable returns on substantial AI investments.
Uber’s CEO, Dara Khosrowshahi, mentioned in a recent earnings call that autonomous agents now build approximately 10% of the company’s committed code. He further noted that the company’s AI adoption extends beyond software engineering, with teams in legal, marketing, and development also utilizing these tools. Khosrowshahi framed this as empowering employees with “superpowers.” However, the financial implications of this widespread adoption are considerable. Research firm Gartner projects that while the cost of inference on highly sophisticated AI models will decrease by 90% between 2025 and 2030, this reduction in token cost might not translate into cheaper enterprise AI. This is because agentic models demand significantly more tokens per task than standard models, and AI providers are unlikely to fully pass on these cost savings to consumers.
Indeed, some AI firms are already adapting their pricing structures to capture this increased usage. Anthropic, for instance, transitioned from a flat-fee model to a usage-based one, charging for compute use per token. Sam Altman, OpenAI CEO, articulated this industry trajectory in a March interview, envisioning a future where intelligence functions as a utility, much like electricity or water, which users purchase on a metered basis. Gartner further forecasts that spending on AI agent software will surge to nearly $207 billion in 2026, marking a substantial 139% increase from the $86.4 billion spent in 2025.
Uber’s research and development spending reflects this trend. The company allocated 3.4% of its budget to R&D in 2025, a 9% increase over the previous year. In the first quarter of 2026 alone, Uber spent $951 million on R&D, representing a nearly 17% jump compared to the same period a year prior. Despite the current expenditure concerns, the company is not retreating from technological advancement. Macdonald indicated that Uber is heavily investing in autonomous driving, a technology he believes will become commonplace within the next two decades. He even remarked that his young daughters, born today, might never need to obtain a driver’s license, underscoring the company’s long-term commitment to a future shaped by advanced technology.
