After a year of urging workers to experiment freely with artificial intelligence, big corporations are suddenly slamming on the brakes. The massive bills associated with running advanced AI models have caught Corporate America by surprise, forcing executives to aggressively ration access to the technology to protect their bottom lines.

Throughout last year, companies across all industries spent lavishly on AI tools, eager to signal to Wall Street that they were capitalizing on the technology’s disruptive potential. However, that unbridled enthusiasm has led to skyrocketing costs for “tokens”—the basic units of measurement for AI data processing—as model providers struggle to balance supply and demand.

Blowing Through Budgets in Record Time

The financial reality of widespread AI adoption has turned out to be far more expensive than most Chief Financial Officers anticipated. Rather than saving companies money, the technology is introducing staggering new infrastructure costs:

• Vanishing Annual Budgets: Some enterprises have burned through their entire annual AI budget in as little as three to four months. Uber Technologies, for example, completely depleted its yearly AI coding assistance budget in just about four months.
• Exploding Operational Bills: Multiple organizations have reported seeing their monthly AI vendor invoices double or triple unexpectedly.
• Compute vs. Talent: The financial weight of processing power has shifted drastically. Nvidia VP Bryan Catanzaro recently noted that the cost of computing power has actually surpassed employee salary expenses for his team.

Enterprise spending on AI is projected to absorb roughly 1.7% of total corporate revenue this year—more than double the allocation seen last year. Despite these massive investments, fewer than 1% of corporate executives report seeing a significant return on investment (ROI) of 20% or more. The vast majority report cost savings of under 10% in the specific departments where the tools have been deployed.

Pulling Back on Access

To bring these surging expenses under control, top technical executives at companies like Uber, Meta, Microsoft, Salesforce, and DoorDash are implementing strict new optimization strategies.

Instead of treating AI like an all-you-can-eat buffet, companies are shifting to a heavily restricted model. Management is actively monitoring usage to ensure that queries contribute directly to measurable productivity. In many cases, organizations are completely pulling premium tool licenses from certain groups of employees, forcing them toward cheaper, homegrown tools or free alternatives. Others are restricting access to premium models like Claude or heavily capping the number of advanced requests a developer can make in a given week.

This sudden pivot toward financial discipline marks a new phase in the AI boom. While tech providers continue to pitch artificial intelligence as the definitive path forward, Corporate America is learning the hard way that intelligence comes with a premium price tag—and for now, they are rationing it like printer ink.