Computing costs for artificial intelligence are beginning to outpace what some of the world's largest companies spend on their entire workforces, a shift that is forcing executives to recalculate whether the technology delivers sufficient returns to justify the expense.
Nvidia's vice president of applied deep learning Bryan Catanzaro described the phenomenon plainly: "For my team, the cost of compute is far beyond the costs of the employees." The strain extends to household names. Uber's chief technology officer exhausted his complete AI budget allocation for 2026 before the year even began, consumed entirely by token costs required to run large language models.
The spending surge reflects a broader explosion in enterprise technology investment. Global IT budgets are projected to reach 6.31 trillion dollars in 2026, a 13.5 percent jump from 2025, with AI infrastructure, cloud services, and software subscriptions driving nearly all of that growth. For many companies, this means AI has become their single largest technology expense.
Amos Bar-Joseph, CEO of Swan AI, captured the bullish sentiment in a viral LinkedIn post, announcing his company as "the first autonomous business, scaling with intelligence, not headcount." That confidence, however, faces mounting pressure. Companies now must prove their AI investments actually move the needle. Productivity gains, revenue increases, or measurable efficiency improvements have become non-negotiable for defending these costs to shareholders during earnings calls.
Brad Owens, vice president of digital labor strategy at Asymbl, a firm specializing in workforce management, noted the conversation is shifting fundamentally. "The tone is shifting a bit more into what is the true value of a worker, human or digital?" he said.
The economics are becoming precarious. As pricing pressure mounts from major AI labs, what once looked like a strategic advantage could quickly transform into a liability. Anthropic has already adjusted its pricing structure to reflect soaring demand. An OpenAI investor flagged another factor: the efficiency game itself. Some labs are better at extracting value per token than others, meaning companies could see their bills fluctuate based on which AI platforms they choose.
The calculus growing clearer each quarter: sky-high AI spending without documented business returns may not survive the next round of cost-cutting.
Author James Rodriguez: "Companies are learning the hard way that having the fanciest AI doesn't matter if the quarterly earnings can't justify the bill."
Comments