Uber's AI Costs Skyrocket Past Budgets

Uber's rapid embrace of AI-powered software development has resulted in unforeseen and escalating costs, forcing a significant budget reassessment. The company's Chief Technology Officer revealed that AI spending assumptions have already been surpassed due to the swift adoption of advanced coding assistants.

AI is fundamentally reshaping how software is developed at Uber, moving beyond simple assistance to generating code independently. This shift, termed 'agentic software engineering,' sees AI agents writing approximately 1,800 code changes weekly without direct human input.

With nearly 95% of Uber's engineers utilizing AI tools monthly, these systems are responsible for about 70% of committed code. The internal AI agent's contribution to code changes has dramatically increased, rising from less than 1% to roughly 8% in just months.

The financial strain is exacerbated by AI tool pricing models, such as Anthropic's hybrid approach. This model combines per-user fees with mandatory token consumption commitments, requiring companies to forecast AI computation usage accurately. The removal of prior enterprise discounts further contributes to the rising costs.