AI Token Budgets Now Rival the Salaries of Laid-Off Workers

When the Cost of AI Surpasses the Employee It Replaced

The meme arrived before the analysis, as uncomfortable truths often do. A CFO with a look of stunned disbelief, 39,000 likes, and a single sentence that captures a structural contradiction few executives are willing to verbalize in board meetings: the AI token budget is about to exceed the salary of the employees who were laid off to fund it.

This is not hyperbole. It is arithmetic.

Jason Calacanis and Chamath Palihapitiya debated this openly on the All-In podcast. “We, with our agents, hit $300 per day per agent using the Claude API, almost instantly,” Calacanis admitted. “That’s $100,000 per year per agent.” Palihapitiya was more direct: “We’re getting to a place where we have to basically say, what is the token budget we’re willing to give our best developers? And when you aggregate it across all people, you can clearly see a trend where you think, well, now they need to be at least 2x as productive as another employee. Otherwise I’ll run out of money.” X

That is not a theoretical debate. It is the real equation behind tens of thousands of layoffs in 2025 and 2026.

The Accounting Illusion of the Efficient Layoff

The official narrative reads as follows: AI increases productivity, which allows companies to do more with fewer people, and the payroll savings fund the technology investment. It is a clean, convenient, and partially false argument.

A March 2026 survey by ResumeBuilder.com of 866 U.S. business leaders found that 54% of companies have or will reduce employee compensation to free up capital for AI spending in 2026. Of those making these cuts, 88% said the weak job market makes it easier to reduce compensation without losing talent. Technocracy News

This is not optimization. It is extraction. Companies are not reinvesting payroll savings into productive infrastructure that benefits remaining employees. They are redirecting that capital toward language models that, in turn, generate operating costs no one accurately projected.

Raw token costs are small, tens or hundreds of dollars per seat per year. But McKinsey’s platform and governance costs range from $200,000 to $500,000 annually. As one analysis concluded: “AI is cheaper claims are misleading unless you factor in the full lifecycle costs and human oversight.” Yahoo Finance

The layoff turned out to be the easy part. The AI bill is the part nobody budgeted for.

Tokens as the New Compensation Line Item

While some CFOs discover that compute spending is exceeding projections, Silicon Valley CEOs are turning that same spending into a recruitment tool. Jensen Huang proposed it without ambiguity at the GTC conference in March 2026.

Huang floated a compensation model in which engineers would receive a token budget on top of their base salary, effectively paying them to deploy AI agents as productivity multipliers. “They’re going to make a few hundred thousand dollars a year in base pay. I’m going to give them probably half of that on top as tokens,” he declared at GTC. CNBC

The proposal captured the industry’s imagination but also revealed a fundamental tension. One Ericsson engineer in Stockholm told the New York Times he probably spends more on Claude than he earns in salary, though his employer covers the bill. And at the point where a company’s token spend per employee approaches or exceeds that employee’s salary, the financial logic of headcount starts to look different to its finance team. If the compute is doing the work, the question of how many humans are needed to coordinate it becomes harder to avoid. TechCrunch

This is the contradiction the meme captures with brutal precision: the CFO who laid off employees to save money discovers that the agent replacing them costs, in terms of total infrastructure, as much or more than the salary they sought to eliminate.

Displacement as Accounting Rebalancing, Not Real Efficiency

The 2026 layoff data tells a more complex story than the narrative of liberating automation.

More than 100,000 employees were impacted by AI-attributed layoffs in 2025. In the first months of 2026, more than 61,000 had already been affected. At least 45 CEOs announced cuts citing AI efficiencies, and layoffs extended beyond the tech sector into finance, logistics, consulting, media, retail, and manufacturing. Programs

Meta set its capital expenditure guidance for 2026 at $115 to $135 billion, nearly double the $72.2 billion it spent in 2025. CFO Susan Li warned of a “significant acceleration in infrastructure expense growth.” For roughly 8,000 people losing their jobs on May 20, the acceleration is already here. TNW | Facebook

The five largest U.S. cloud and AI infrastructure providers, including Microsoft, Alphabet, Amazon, Meta, and Oracle, have collectively committed to spending between $660 and $690 billion on capital expenditures in 2026. Capital expenditures among major tech companies have more than doubled over the last two years, reaching $427 billion in 2025. Technocracy News

The Productivity Paradox CFOs Did Not Anticipate

A Duke CFO survey, conducted in partnership with the Federal Reserve Banks of Atlanta and Richmond, found a significant gap between the perceived productivity gains from AI and the realized benefits. Perceptions of AI’s gains are larger than the reality, a lag the researchers say likely reflects a delay in realized revenue, consistent with what economists have observed about AI productivity gains. Fortune

In concrete terms: companies laid off people based on productivity projections that AI has not yet materialized, while the operating costs of the model meant to replace those people are already landing on the balance sheet. It is the worst possible fiscal scenario: the human asset was eliminated before the technological asset was ready to substitute it, and the compute bill keeps running in the meantime.

The term AI-washing has emerged to describe situations where companies over-attribute layoffs to AI. The argument is not that AI is irrelevant, but that it is sometimes used as a simplified narrative that reduces public backlash, signals modernization, and reassures investors that leadership is acting decisively. Blockchain Council

What the Meme Implies but Does Not Say

The bewildered CFO in the meme is not a victim of technology. He is the architect of a poorly calibrated decision. Laying off employees to fund AI, without rigorously modeling the total cost of ownership of that AI, is a failure of financial analysis, not an efficiency adjustment.

The token compensation structure introduces new complexities around equity and access. Engineers at well-funded startups might receive lavish token packages. Those at bootstrapped companies or non-tech firms might receive nothing, forced to rely on free tiers or personal subscriptions. The disparity could become another form of talent stratification. Techbuzz

The question no CFO poses to their board is this: if the AI agent costs, in annual infrastructure terms, the same or more than the employee it replaced, what was the real benefit of the layoff? The honest answer, in many cases, is that the benefit was accounting-based and temporary, and the deferred bill is arriving now.