AI Now Costs More Than the Workers It's Replacing. Companies Are Cutting Anyway. That's the Whole Problem.

Here's the state of play. Companies are spending more on AI than on the people who use it. They're simultaneously cutting headcount in the name of AI efficiency. And the financial returns from those cuts are, according to the most comprehensive data available, nonexistent.

That's not a contradiction the market will ignore forever.

Bryan Catanzaro, Nvidia's vice president of applied deep learning, recently told Axios something that should be pinned to the wall of every boardroom running an AI transformation: "For my team, the cost of compute is far beyond the costs of the employees."

That quote is striking for what it is. It's devastating for where it comes from. Nvidia is the company selling the infrastructure. They are the picks-and-shovels supplier for every AI gold rush happening right now. And their own VP is saying the shovels cost more than the miners.

The spending is real. The returns are not.

Worldwide IT spending is projected to hit $6.31 trillion in 2026, up 13.5% from 2025, according to Gartner's latest forecast. Data center systems spending alone is expected to surge 55.8%, reaching nearly $788 billion. GenAI model development spending is forecast to more than double year-over-year. The money is moving fast.

And it's not just infrastructure. At the application layer, token costs are creating a new category of budget crisis that nobody modeled for. Uber's CTO Praveen Neppalli Naga told The Information that his company burned through its entire 2026 AI budget in four months, driven by explosive adoption of AI coding tools across the engineering org. Engineers were racking up monthly API costs between $500 and $2,000 per person. By April, the full-year budget was gone. Uber is now, in Naga's words, "back to the drawing board."

Uber's R&D expenses hit $3.4 billion in 2025, up 9% year-over-year. AI-related costs at the company have risen roughly 6x since 2024. This isn't a company that stumbled into adoption. They actively incentivized it with internal leaderboards ranking teams by AI tool usage.

Meanwhile, on LinkedIn, Swan AI CEO Amos Bar-Joseph posted about his company's $113,000 Anthropic bill like it was a badge of honor, writing: "We're building the first autonomous business, scaling with intelligence, not headcount." The post went viral. The bill will keep growing.

Jensen Huang, Nvidia's CEO, went further at GTC 2026. He said he'd be "deeply alarmed" if a $500,000 engineer didn't consume at least $250,000 worth of AI tokens per year. When asked if Nvidia is spending around $2 billion annually on tokens for its engineering team, Huang answered simply: "We're trying to." He's now pitching token budgets as a recruiting tool, a fourth component of compensation alongside salary, bonus, and equity.

That framing tells you everything about where this is headed. AI is not reducing costs. It's creating an entirely new cost center that scales with usage, not headcount.

The layoffs are accelerating anyway

While compute costs are exploding, companies are cutting people at an accelerating pace and citing AI as the reason.

In 2025, companies directly attributed 55,000 job cuts to AI, more than 12 times the number of AI-linked layoffs just two years earlier, according to outplacement firm Challenger, Gray & Christmas. Over 100,000 employees were impacted by AI-driven layoffs in 2025, per Programs.com. In 2026, that number has already surpassed 70,000 and the year isn't half over.

The names are not small. Accenture is cutting at least 11,000 employees in an AI-focused restructuring. Block laid off 4,000 workers, roughly 40% of its global workforce, in February. Coinbase cut nearly 700 in an "AI-native" restructuring in May. Freshworks announced 500 cuts as part of AI-led changes. Cognizant is eliminating 4,000 roles under its "Project Leap" AI initiative. Pinterest cut 15% of its workforce in January, explicitly stating it was "reallocating resources" to AI-focused teams. Meta is cutting another 1,000 in Reality Labs. HP announced plans to reduce headcount by 4,000 to 6,000, projecting $1 billion in savings by 2028.

A Resume.org survey of 1,000 U.S. business leaders found that 37% of companies expect to have replaced jobs with AI by the end of 2026. Nearly three in ten said they've already done it. Half have pulled back on hiring. Fifty-eight percent believe layoffs are likely in 2026.

Here's the part that should concern every executive making these decisions. Forrester reports that 55% of employers already regret laying off workers for AI. Many companies are cutting roles for AI capabilities that don't exist yet, betting on future promises rather than proven technology.

And the data from the companies that have actually tried it says the bets aren't paying off. As we reported earlier this month, Gartner surveyed 350 global enterprises with at least $1 billion in annual revenue that were already deploying autonomous AI capabilities. Eighty percent had made workforce reductions tied to AI. And the companies that cut the most showed nearly identical financial returns to the companies that cut the least. In several cases, the ones that cut less performed better.

Helen Poitevin, Distinguished VP Analyst at Gartner, was direct: "Many CEOs turn to layoffs to demonstrate quick AI returns; however, this disposition is misplaced. Workforce reductions may create budget room, but they do not create return."

The economic viability question nobody wants to answer

The assumption driving every AI-for-headcount swap is that AI is cheaper than people. An MIT study from 2024 examined that assumption rigorously and found that AI automation was economically viable in only 23% of roles where computer vision is a key component of the work. In the remaining 77%, human labor was cheaper.

A follow-up MIT study released in late 2025 expanded the analysis across the full U.S. labor market and found that current AI systems could take over tasks tied to about 11.7% of the workforce, representing roughly $1.2 trillion in wages. That's meaningful. But it also means that for nearly 90% of the labor market, AI replacement is not yet cost-effective.

Those studies were conducted before the current wave of token cost inflation. The costs of running AI agents at scale, where systems read entire codebases, orchestrate multi-step workflows, run tests, and open pull requests autonomously, consume tokens at rates that bear no resemblance to a simple per-seat subscription. As Uber discovered, agentic workflows create nonlinear cost curves that make traditional budgeting models obsolete.

Ben May, director of global macro research at Oxford Economics, raised another possibility in a CBS News report: that some companies could be using AI as a pretext for job cuts that would have happened regardless. The AI narrative provides better optics than "we overexpanded" or "the market shifted." It's a strategy that works until investors start looking at the numbers.

And they are. Goldman Sachs noted that investors are now reacting negatively to layoff announcements, even when framed as strategic automation moves. The era of cutting headcount and getting a stock bump for it may already be ending.

The brand cost that doesn't show up on a balance sheet

At The State of Brand, we've been tracking this pattern all year. Klarna replaced 700 customer service roles with AI, watched quality decline, and started rehiring. IBM automated large parts of its HR function and reversed course. The Commonwealth Bank of Australia reversed AI-driven layoffs after concluding the roles were never redundant.

Now we have the macro data to match the anecdotes. Cutting people for AI is not improving financial performance.

But even if it were, the brand calculus runs differently. When Gartner's data says financial returns are equivalent whether you cut or not, it doesn't measure the second-order effects that compound over time.

The talent signal. When a company publicly eliminates roles for AI, it tells every prospective hire that people are a variable cost to be optimized away. In a market where Gartner predicts companies will pay a 15% premium to rehire the same talent they cut, that signal has a price. It just shows up two years late.

The customer trust deficit. When buying committees evaluating a seven-figure purchase see a vendor cutting the customer success team and replacing it with chatbots, they don't see innovation. They see risk. They see a company that might not have the team to support them a year from now. In B2B, where deals close on trust and relationships, that's not a branding problem. It's a pipeline problem.

The narrative vacuum. Companies that announce layoffs "for AI" cede narrative control to everyone else: analysts, journalists, former employees, competitors. The story becomes about cost-cutting, not transformation. About subtraction, not capability. And in an era where AI search surfaces brand narratives in real time, that framing calcifies. It becomes the answer when a prospect asks ChatGPT or Perplexity what your company is doing with AI.

The institutional knowledge gap. The mid-career professionals being cut, the people who understand the customers, the edge cases, the reasons the process works the way it does, cannot be replaced by a model trained on internal documentation. When they leave, the documentation becomes the only source of truth. And anyone who has worked at a company of any size knows exactly how reliable internal documentation is.

What the smart companies are doing instead

The companies seeing real returns from AI are not playing the layoff game. They're playing the leverage game.

They're using AI to amplify what their people can do, not to eliminate the people doing it. They're redesigning roles around what humans do well (judgment, relationships, creative problem-solving, navigating ambiguity) and what AI does well (speed, pattern recognition, processing scale, tireless execution of defined workflows). They're building operating models where humans guide autonomous systems rather than getting replaced by them.

Poitevin was pointed about this in the Gartner analysis: "Those who only look to the workforce tend to be the 'laggards,' because they're not going after the broader set of value that they can get to."

The difference is between using AI to do the same work with fewer people and using AI to unlock work that was previously impossible. The first approach saves money on paper. The second compounds over time. And it is the second group that Gartner's data associates with stronger ROI.

Brad Owens, VP of digital labor strategy at Asymbl, captured the moment: "The tone is shifting a bit more into what is the true value of a worker… human or digital?"

That question is the right one. The answer companies land on will define their brands for the next decade.

The bottom line

Nvidia's own VP says AI costs more than his employees. Uber blew its entire AI budget in four months. MIT says AI replacement isn't economically viable for the vast majority of roles. Gartner says the companies cutting people aren't seeing better returns than the ones keeping them. Forrester says more than half of companies already regret the cuts they made.

And yet, the layoffs continue. More than 70,000 AI-attributed job cuts in 2026 so far. Forty-five CEOs and counting have publicly tied workforce reductions to AI.

At some point, the market will reconcile the narrative with the numbers. AI spending will need to produce returns that justify the costs. Not just savings from eliminating the people who used to do the work, but actual revenue growth, faster time to market, and measurable operational transformation.

The companies that figure this out will build the defining brands of the next era. The ones that treat AI as a headcount reduction program will spend the next five years rehiring, rebuilding trust, and explaining to their boards why the savings never materialized.

The cost of compute is far beyond the cost of the employees. The cost of getting the strategy wrong is far beyond both.