Uber COO Says AI Spending Hard to Justify as Productivity Gains Lag

Why It Matters

The Uber episode underscores a pivotal moment for CIOs tasked with justifying large AI investments. When token‑driven costs outpace demonstrable productivity gains, enterprises risk eroding stakeholder confidence and inflating operating expenses without delivering customer value. This tension forces CIOs to develop more rigorous measurement frameworks, tying AI usage directly to feature delivery, bug reduction, or revenue impact. If the industry does not shift toward outcome‑based AI procurement, the current wave of token‑maxxing could lead to broader budgetary pull‑backs, slower hiring, and a reevaluation of AI‑first strategies. Uber’s public acknowledgment may accelerate vendor reforms, prompting AI tool providers to offer clearer ROI metrics and flexible pricing models that align with enterprise performance goals.

Key Takeaways

• Uber burned through its entire 2026 AI budget in four months, according to CTO Praveen Naga.
• 95% of Uber engineers now use AI tools monthly; 70% of code commits are AI‑generated.
• COO Andrew Macdonald said there is no clear link between token consumption and a 25% increase in useful consumer features.
• Microsoft has also halted Claude Code usage, citing rising token costs without productivity gains.
• CIOs may need to shift from seat‑based licensing to outcome‑based AI contracts to control spend.

Pulse Analysis

Uber’s candid admission signals a maturation of the enterprise AI market. Early hype promised ten‑fold productivity gains, but the reality of token‑based pricing has turned AI spend into a line‑item that must be defended in boardrooms. The company’s experience illustrates how high adoption rates can mask underlying inefficiencies; without a robust attribution model, AI becomes a cost center rather than a growth engine.

Historically, technology investments that lack clear ROI tend to trigger a cycle of budget tightening and strategic reassessment. In the case of AI, the shift is likely to be two‑fold: first, enterprises will demand granular analytics that map token usage to specific outcomes such as reduced time‑to‑market or defect rates. Second, AI vendors will be pressured to redesign pricing structures, moving away from flat‑fee or per‑seat models toward usage‑based or performance‑linked fees. This could spur a new wave of AI platforms that embed built‑in ROI dashboards, making it easier for CIOs to justify spend.

Looking ahead, the key question is whether the industry can deliver the promised productivity uplift fast enough to avoid a prolonged pull‑back. If firms like Uber can translate token consumption into measurable product improvements, the AI boom may regain momentum. If not, we could see a recalibration of AI ambitions, with CIOs prioritizing selective, high‑impact use cases over blanket deployment. The next earnings season will likely reveal whether Uber’s adjustments are enough to restore confidence or whether the broader enterprise sector will adopt a more cautious, ROI‑first approach to AI.