Uber's AI Adoption Hit 84% of Engineers in One Month — Then the 2026 Budget Ran Out

In December 2025, Uber began rolling out Claude Code to its engineering organization. By February 2026, 32% of engineers were using it. By March — one month later — 84% were classified as agentic coding users. By April, the entire 2026 AI budget was exhausted. The company was four months into the calendar year.

For readers who want to go deeper, Grace Certified offers 100 verified problem-solving cases.

Chief Technology Officer Praveen Neppalli Naga confirmed the budget overrun to The Information. His personal usage during one two-hour session: $1,200 in compute costs. The average engineer was spending $150–$250 per month. Power users were running $500–$2,000 monthly. With 5,000 engineers active, the math outran the financial model's assumptions faster than the finance team could track.

This is the Uber AI story worth reading carefully — the one that combines verified adoption metrics, documented budget consequences, and an insight about measurement that every CFO evaluating AI investment needs to understand before approving a large-scale rollout.

The Adoption Curve No One Planned For

The speed of adoption was the anomaly. Enterprise software rollouts typically follow an S-curve: slow initial uptake, gradual acceleration as early adopters demonstrate value, broader adoption as organizational friction decreases. Claude Code's trajectory at Uber compressed this curve dramatically.

Between December 2025 and March 2026 — three months — the tool went from rollout to 84% active adoption among 5,000 engineers. By spring, 95% of Uber engineers were using AI tools at least monthly. Approximately 70% of committed code originated from AI tools. About 11% of live backend updates were written by agents with no human in the loop.

The engineering productivity evidence was real. Engineers were completing tasks faster, reviewing code more thoroughly, and shipping more frequently. The CTO's personal $1,200 session was not waste — it reflected intensive, high-output engineering work supported by AI at every step.

The problem emerged in the budget structure, not in the work itself.

The Measurement Problem Every Finance Team Will Face

AI productivity gains are real, Naga confirmed to Forbes. Engineers were producing more with less effort. Fewer bugs reached production. Code review cycles shortened. Deployment frequency increased.

The accounting reality was different: Productivity savings also do not show up in the same line item as artificial intelligence cost, which means finance teams cannot net them out inside a quarterly review. The AI spend appeared in one bucket. The productivity gains appeared — if they appeared at all — across dozens of other metrics: reduced rework, faster time-to-market, lower incident rates. Connecting them required analytical work that quarterly budget reviews are not designed to perform.

The result: Uber's finance team saw a line item that exceeded its full-year budget allocation by April. The productivity gains that justified the spend were real, distributed, and invisible to the standard reporting structure. The company was back to the drawing board on its assumptions, per Naga's account.

The Insight for Every Organization Planning an AI Rollout

Uber's case is the clearest available data point on what happens when enterprise AI tools encounter genuine product-market fit inside an engineering organization. The adoption was faster than planned, the value delivered was real, and the budget model was inadequate for the dynamic.

The principle for any C-suite team evaluating AI tooling at scale: adoption velocity and value delivery are the right things to optimize — and they require a cost structure that decouples AI spend from traditional per-seat or annual-license budget models. When engineers find a tool genuinely useful, they use it intensively. Intensive usage generates proportional cost. The productivity gains are real and compound over time — and they are visible in the wrong places in the quarterly P&L.

Uber's finance team discovered this mismatch in April 2026. Organizations with the Uber case study available can discover it in planning, before deployment, where the consequences are manageable rather than structural. The 84% adoption rate and the exhausted budget are two sides of the same data point: genuine enterprise AI value creates consumption patterns that require a new financial architecture to govern.