Why AI Bills Are Exploding Without Delivering ROI

The AI Budget Crisis

Uber's CTO made headlines in April 2026 after revealing the company had already exhausted its annual AI budget. The ride-sharing giant isn't alone—CEOs and CIOs across industries are watching AI expenses climb while struggling to identify corresponding returns on investment.

The cost explosion stems from a collision of market dynamics and organizational behavior. AI providers initially offered extremely low prices to drive adoption, while simultaneously, agentic AI systems began consuming dramatically more tokens—the basic unit of AI usage. Even as per-token costs decreased, total consumption increased enough to drive net costs higher.

Meanwhile, companies aggressively promoted AI adoption, sometimes tying usage to promotions or employment security. This created perverse incentives that prioritized consumption over value creation.

Why It Matters

The AI spending crisis reveals a fundamental gap between technology adoption and business strategy. Organizations that rushed to embrace AI without establishing clear metrics tied to business outcomes now face budget overruns, security risks from unauthorized tools, and a workforce that may be using AI extensively but not effectively. For technology leaders, this represents both a financial challenge and an opportunity to establish more disciplined AI governance before costs spiral further.

The Tokenmaxxing Problem

A troubling practice has emerged: tokenmaxxing, where employees compete to use the most AI tokens possible. When organizations mandate AI usage without specifying how it should drive business value, workers respond by maximizing consumption to demonstrate compliance. This behavior inflates costs without improving outcomes.

Compounding the issue is the proliferation of "Shadow AIs"—unauthorized AI tools employees adopt outside official IT channels. Like the shadow IT phenomenon before it, these tools create security vulnerabilities and make cost management nearly impossible.

What Technology Leaders Should Do

The first step is recognizing that widespread AI adoption, while valuable, represents only the beginning of transformation. The real work involves connecting usage to measurable business outcomes.

Organizations need to move beyond token-based metrics toward measurements that reflect actual business value. Token consumption is easy to track because it ties directly to billing, but it reveals nothing about whether AI is being used effectively. Better metrics should reward employees for business outcomes and efficient AI use that amplifies their capabilities.

Employee training must evolve beyond basic AI literacy to focus on effective application within specific roles and domains. Workers need to understand not just how to use AI, but when to use it, what tasks benefit most, and how their AI-augmented work connects to organizational objectives.

Treating AI as a Strategic Budget Item

AI costs should be managed as a major budget line item comparable to facilities or travel, with dedicated cross-functional ownership. The rapid evolution of AI pricing and capabilities makes forecasting challenging, but organizations must establish processes for tracking consumption, negotiating with providers, and understanding the shifting cost landscape.

Procurement strategies should tier AI services by business criticality, reserving the most expensive options for mission-critical applications while providing reasonable allocations for general team use.

The Path Forward

Achieving widespread AI adoption is genuinely difficult and represents real progress. However, adoption without strategic discipline leads to inflated bills rather than improved outcomes. Business metrics—the true non-negotiables—haven't changed. AI is a means to improve them, not an end in itself.

These details were first reported by Nisha Talagala, CEO of Schovia, writing in Forbes.