Uber COO Struggles to Justify AI Token Costs
Key insights
- Uber's COO cannot link 70% AI-generated code or 95% engineer adoption to measurable consumer product improvements.
- Uber exhausted its entire 2026 Claude Code and Cursor budget within four months of the year starting.
- Uber joins Microsoft and Duolingo in publicly questioning whether token volume translates to business outcomes.
Why this matters
Enterprise AI adoption has crossed a threshold where high usage rates are no longer the story, and Uber's admission signals that the measurement problem is now urgent enough for C-suite executives to raise publicly. For AI tooling vendors like Anthropic and Cursor, this creates pricing and contract pressure as procurement teams demand outcome-linked justification rather than seat-count renewals. Technical leaders evaluating AI coding tool budgets now have a named, high-profile precedent for demanding ROI frameworks before committing multi-year spend.
Summary
Uber COO Andrew Macdonald has publicly admitted the company cannot draw a clear line between its surging AI token spend and measurable improvements in consumer products, even as 95% of Uber engineers use AI tools monthly and 70% of committed code is now AI-generated.
The admission carries weight because it follows Uber CTO Praveen Neppalli Naga's earlier disclosure that Uber burned through its entire 2026 budget for Claude Code and Cursor in just four months. The company is spending aggressively on AI tooling, seeing adoption at near-total scale internally, and still cannot point to the revenue or product outcome that justifies the line item.
Essentially: (Uber, Microsoft, Duolingo) are arriving at the same uncomfortable place: high AI adoption rates do not automatically produce legible ROI.
- 95% of Uber engineers use AI tools monthly, yet the COO cannot connect that usage to consumer-facing gains.
- The company exhausted its full-year Claude Code and Cursor budget by April 2026.
- Macdonald's statement is among a growing set of enterprise signals pushing back on treating token consumption as a proxy for productivity.
As enterprises normalize high AI adoption, the next competitive pressure point shifts from access to measurement: who can actually prove the spend is working.
Potential risks and opportunities
Risks
- Anthropic and Cursor face contract renegotiations or non-renewals at large enterprise accounts if Uber's public framing emboldens other COOs to demand outcome-linked pricing in Q3 2026 budget cycles.
- Uber engineering leadership risks internal credibility loss if the 70% AI-generated code figure cannot be tied to shipping velocity or defect reduction before the next board review.
- A broader enterprise pullback on AI coding tool budgets could suppress token demand forecasts, directly affecting Anthropic's revenue projections for the second half of 2026.
Opportunities
- AI observability and attribution vendors (Langfuse, Honeycomb, Datadog's LLM monitoring suite) gain a direct sales narrative: Uber's problem is their product.
- Consulting firms with AI ROI measurement practices (McKinsey QuantumBlack, Accenture AI) can use the Uber disclosure as a wedge to sell measurement engagements to similarly exposed enterprises.
- Outcome-based AI coding tool startups with usage-linked pricing models gain differentiation against Cursor and GitHub Copilot if the ROI measurement gap widens into a procurement requirement.
What we don't know yet
- Whether Uber has since implemented any token-consumption guardrails or usage caps after blowing the 2026 budget by April.
- Which specific consumer product metrics Uber is using to attempt ROI attribution, and why they are failing to show signal.
- Whether Anthropic or Cursor have offered Uber revised contract structures (e.g., outcome-based pricing) in response to the budget overrun.
Originally reported by businessinsider.com
Read the original article →Original headline: Uber COO Says It's Getting Harder to Justify AI Token Spending — 70% of Committed Code Is AI-Generated but ROI Line Won't Draw