Uber’s chief operating officer, Andrew Macdonald, has publicly questioned the company’s $500 million AI investment, calling the return on spend unclear after four months of rapid adoption. The admission comes as enterprises grapple with ballooning AI costs—even as per-token pricing drops—and raises questions about whether the hype has outpaced the value.
The AI Budget Burn Rate
Uber blew through its entire 2026 budget for AI coding tools—including Claude Code—in just four months, according to Fortune. The company incentivized adoption with an internal leaderboard ranking teams by AI tool usage, but Macdonald now says the link between spending and measurable consumer benefits is tenuous. “That link is not there yet,” he told the Rapid Response podcast.
The revelation underscores a broader industry reckoning. While AI infrastructure costs are plummeting—Gartner projects inference expenses for sophisticated models will drop 90% by 2030—enterprises face a catch-22: agentic AI models require far more tokens per task, and providers aren’t passing cost savings to consumers. Uber’s experience mirrors Microsoft’s recent pivot, where the tech giant reportedly canceled most of its Claude Code licenses in favor of GitHub Copilot CLI, a move that reflects growing skepticism about vendor lock-in and ROI.
CEO Dara Khosrowshahi’s Contradiction
Uber CEO Dara Khosrowshahi painted a rosier picture during the company’s May earnings call, where he claimed 10% of Uber’s committed code is now written by autonomous agents. “We’re seeing uptake of these tools, whether it’s our legal team or marketing team or developers,” he said. “We think it’s creating employees with superpowers.” But the disconnect between Macdonald’s caution and Khosrowshahi’s optimism highlights a critical tension: while AI adoption is accelerating across functions, the business case remains murky.
“If you’re not actually able to draw a direct line to how [many] useful features and functionality you’re shipping to your users, that trade becomes harder to justify.”
The quote cuts to the heart of the problem. Companies like Uber are racing to adopt AI tools—not just for coding, but for legal review, marketing automation, and even customer support—yet few can quantify the tangible impact on revenue or efficiency. Macdonald’s skepticism aligns with a growing chorus of executives who’ve scaled back AI ambitions. Duolingo’s CEO, Luis von Ahn, reportedly reversed course last year, admitting AI isn’t replacing the tasks his employees perform. The pattern suggests that while AI adoption is inevitable, its financial justification is far from settled.
Why the AI Binge Could Backfire
- Token inflation: Agentic models demand exponentially more computational resources than traditional AI. A single complex query might require 100x the tokens of a simple prompt, eroding cost savings from cheaper per-token pricing.
- Vendor opacity: AI providers like Anthropic (Claude) and OpenAI aren’t transparent about how they allocate cost reductions. Enterprises pay for access, not raw compute, leaving them vulnerable to hidden fees.
- Measurement gaps: Most companies lack the tools to track AI’s direct impact on revenue or productivity. Macdonald’s comment about “useful consumer features” reflects a broader industry struggle: how do you prove AI is worth the price tag?
The risk isn’t just financial. Over-investment in unproven AI could distract from core business priorities. Uber’s 2026 budget burn suggests that without clear ROI, executives may shift spending back to revenue-generating areas—like driver incentives or market expansion—where the impact is easier to measure.
What Comes Next: The AI Audit
Companies are beginning to demand hard evidence before greenlighting further AI investments. Yahoo Finance reported this week that enterprise AI budgets are under scrutiny, with CFOs pushing for pilot programs with defined KPIs before scaling. The shift mirrors what happened in cloud computing a decade ago: early adopters overbuilt, then consolidated as costs became clearer.
For Uber, the next 90 days will be critical.
- Pause new AI tool purchases until ROI is demonstrated.
- Redirect funds to high-impact use cases (e.g., dynamic pricing algorithms with proven uplift).
- Push vendors like Anthropic to offer usage-based pricing tied to measurable outcomes.
The broader implication? The AI gold rush may be over. What’s left is a sober reassessment: not every dollar spent on AI yields a dollar in value. For now, Uber’s $500 million burn is a warning to the rest of Silicon Valley: the hype curve is flattening, and the bills are due.
Correction (May 31, 2026): An earlier version of this article misstated the timeline for Uber’s AI budget depletion. The company exhausted its 2026 allocation in four months, not six.
