محمد يحي
Uber is facing growing internal questions over the effectiveness of its artificial intelligence investments after reports revealed the company exhausted its entire 2026 AI coding tools budget within just four months.
The concerns were highlighted by Uber President and Chief Operating Officer Andrew Macdonald during an appearance on the Rapid Response podcast, where he admitted it remains difficult to directly connect increased AI adoption with meaningful consumer-facing innovation.
“That link is not there yet,”
Macdonald said, referring to the company’s growing use of AI coding tools such as Claude Code.
“Maybe implicitly there’s more that is getting shipped, but it’s very hard to draw a line between one of those stats and actually producing significantly more useful consumer features.”
According to reports, Uber encouraged employees to increase their use of AI tools through internal leaderboards ranking teams by overall AI usage, accelerating the company’s spending on AI infrastructure and coding assistants.
The situation reflects a broader challenge facing enterprises adopting AI technologies at scale.
While the cost of AI computation and models is gradually declining, overall corporate spending continues to rise as businesses rely more heavily on advanced autonomous agents that require substantially greater computational resources.
Macdonald noted that rising AI expenses become harder to justify when companies cannot clearly demonstrate measurable improvements in products and user experiences.
Uber is not alone in facing this dilemma.
Reports recently indicated that Microsoft has started reducing direct Claude Code licenses in favor of GitHub Copilot CLI, while several tech executives have softened their earlier optimism regarding AI’s ability to replace human workers.
Last year, Luis von Ahn, CEO of Duolingo, also revised his public stance, saying he no longer believes AI will replace the work performed by employees at the company.
Despite concerns surrounding costs, Uber continues expanding AI usage across multiple departments beyond software engineering. During the company’s recent earnings call, Uber CEO Dara Khosrowshahi stated that nearly 10% of the company’s committed code is now generated by autonomous AI agents.
“We’re seeing uptake of these tools, whether it’s our legal team or marketing team or developers,” Khosrowshahi said. “We think it’s creating employees with superpowers.”
Meanwhile, research firm Gartner estimates that inference costs for highly sophisticated AI models could decline by as much as 90% by 2030 compared to 2025 levels.
However, the firm warned that lower per-token costs may not necessarily reduce enterprise AI spending because agentic AI systems consume significantly more computational tokens per task.
AI companies are also revising pricing models to capitalize on growing usage demand. Anthropic recently shifted from flat-rate pricing to a usage-based structure tied directly to compute consumption.
Similarly, Sam Altman, CEO of OpenAI, previously described AI’s future as a utility-based service model where intelligence would be consumed “like electricity or water.”
Another Gartner study forecasts that global spending on AI agent software will reach nearly $207 billion in 2026, up from $86.4 billion in 2025, representing growth of more than 139%.
Uber increased its research and development spending by 9% in 2025 compared to the previous year. The company also spent approximately $951 million on research and development during the first quarter of 2026 alone, marking a nearly 17% year-over-year increase.
Nevertheless, Uber remains committed to long-term technological innovation, particularly in autonomous driving. Macdonald said he believes self-driving transportation will become standard within the next couple of decades.
He said:
“I don’t think my daughters, who are little kids today, will end up getting a driver’s license”.
Read the article in
