AI Infrastructure Costs Now Exceed Salaries of Replaced Workers

The economics of AI aren't adding up

Tech companies face an uncomfortable arithmetic problem: the artificial intelligence tools meant to replace human workers currently cost more than the employees they've displaced. Uber burned through its entire 2026 AI coding budget by April—just four months into the year—despite unclear productivity gains, according to reporting by Forbes contributor Jemma Green.

By March, 84 percent of Uber's engineers had adopted Claude Code, with roughly 70 percent of committed code now originating from AI. Yet Uber's COO Andrew Macdonald publicly acknowledged that token usage didn't correlate directly with useful features shipped to users.

Microsoft, after investing approximately $13 billion in OpenAI, instructed engineers in a major division to stop using an AI coding assistant because costs became untenable. One unnamed company ran up a $500 million Claude bill in a single month after management forgot to set a usage cap.

Bryan Catanzaro, Nvidia's vice president of applied deep learning, stated bluntly that compute costs for his team now exceed what the company spends on the employees using it. Meanwhile, Nvidia CEO Jensen Huang tells the industry that a $500,000 engineer should consume at least $250,000 worth of AI tokens annually, with Nvidia targeting a $2 billion annual token budget for its engineering workforce.

Why it matters

This cost structure reveals a fundamental market distortion: AI providers are pricing inference below the cost of serving it, burning venture capital to capture market share. OpenAI spends nearly two dollars for every dollar it earns on inference and loses money on its $200 monthly subscriptions. When this subsidy model unwinds—as it began to in 2026—enterprise AI bills could rise another 30 to 50 percent. Companies that laid off workers to fund AI adoption may find themselves unable to afford either.

The human cost of misaligned incentives

More than 115,000 tech workers lost their jobs in 2026 across over 150 companies, with AI reallocation cited as a primary rationale. Meta eliminated 8,000 positions. SentinelOne cut 8 percent of its workforce. Atlassian shed 1,600 jobs.

Yet an MIT study found AI automation is economically viable in only 23 percent of roles. For the remaining 77 percent, humans remain cheaper. Goldman Sachs' chief economist stated he does not view AI investment as strongly growth-positive, while Sequoia Capital estimates AI companies need roughly $600 billion in annual revenue to justify current infrastructure spending—a gap that widened through mid-2026.

The cultural dimension compounds the problem. Amazon built an internal leaderboard called KiroRank to track AI usage among engineering teams, then quietly removed it after employees gamed the system by burning tokens on meaningless tasks to climb rankings. When consumption becomes the performance metric, spending becomes the output.

Market correction underway

In June 2026, chipmakers lost roughly $1.3 trillion in market value in a single session—the steepest one-day drop for the PHLX semiconductor index since March 2020. South Korea's benchmark index fell 10 percent and briefly halted trading. The selloff reflected investor recognition that spending and returns have diverged.

Anthropic moved enterprise customers from flat-rate plans to usage-based billing in April 2026. GitHub followed with the same shift for Copilot after quietly absorbing up to eight times the subscription value for heavy users. OpenAI projects $14 billion in losses this year, with $44 billion in cumulative losses before any profit appears in 2029.

Lisa Emme, co-founder of Inversion AI, argues the path forward requires architectural change: "AI-native companies rebuild around the model—and once you do, you stop paying frontier prices for work a specialized model does better and cheaper. The winners this decade won't run the biggest model. They'll have built systems where the right model runs the right task."

The industry must now answer whether AI pays for itself before funding runs out. These details were first reported by Dr. Jemma Green, cofounder and chairman of Powerledger, writing for Forbes.