Key Takeaways:
Corporate America spent 2025 pushing employees to use AI. Now companies are discovering the technology costs more than the workers it was meant to replace.
Tesla will limit employee AI spending to $200 a week starting July 6, according to an internal memo, just months after leadership gamified token consumption to drive adoption — a reversal that mirrors Uber, Microsoft and other companies confronting runaway costs.
"The usage was enormous. The corresponding value was murkier," Uber Chief Operating Officer Andrew Macdonald said publicly, after the ride-hailing company burned through its entire 2026 AI coding budget in four months.
Tesla software engineers were consuming thousands of dollars in AI tokens each week, according to two people familiar with the usage, prompting the $200 cap. The limit excludes beta versions of xAI products, a carve-out that steers heavy users toward Elon Musk's own AI company. Uber, which saw 84 percent of its engineers adopt Claude Code and roughly 70 percent of committed code originate from AI, imposed a $1,500 monthly cap after exhausting its two-year budget by April.
The spending clampdowns come as Big Tech has committed $740 billion in capital expenditure this year, up 69 percent from 2025, while more than 115,000 tech workers have been laid off in 2026. The arithmetic is turning perverse: companies are cutting human labor to fund artificial intelligence that, for many tasks, remains more expensive than the people it replaced.
The Tokenmaxxing Problem
Amazon built an internal leaderboard called KiroRank to track AI usage among engineering teams. It was quietly taken down after employees began gaming it — burning tokens on meaningless tasks to climb the rankings. Meta built a similar tracker called Claudeonomics. Amazon encouraged staff to "tokenmaxx," treating consumption itself as a performance indicator.
The pattern is industry-wide. Roughly 95 percent of enterprise AI usage still runs on the costliest frontier models, even for work that does not demand that sophistication. Microsoft instructed engineers in a major division to stop using an AI coding assistant because the bills became untenable. One unnamed company ran up a $500 million Claude bill in a single month after management forgot to set a usage cap, according to Axios.
Nvidia's own vice president of applied deep learning, Bryan Catanzaro, acknowledged that the cost of compute for his team now far exceeds what the company spends on the employees using it. Yet Nvidia Chief Executive Jensen Huang has said a $500,000 engineer should consume at least $250,000 worth of AI tokens annually, and that the company is working toward a $2 billion annual token budget for its engineering force.
The xAI Carve-Out
The most revealing detail in Tesla's policy is what the cap leaves out. By excluding beta versions of xAI products, the company is using an expense policy to funnel employees toward Grok and Composer — tools from Musk's own AI startup — while its own engineers quietly prefer Anthropic's Claude, according to four people. Musk admitted last year that xAI was "not built right," weeks after Tesla invested $2 billion into it.
SpaceX is now set to acquire Cursor's parent company Anysphere for $60 billion in an all-stock deal expected to close this quarter. Tesla engineers became early testers for unreleased versions of Grok and Composer, with xAI product lead Andrew Milich running feedback discussions in internal Teams channels.
The Pricing Reckoning
The prices companies are paying for AI usage are not real prices. OpenAI, Anthropic, Google and Meta are all pricing inference below the cost of serving it, burning venture capital to buy market share. OpenAI spends nearly $2 for every dollar it earns on inference and projects $14 billion in losses this year, with $44 billion in cumulative losses before any profit appears in 2029.
Anthropic moved enterprise customers from flat-rate plans to usage-based billing tied to actual compute in April 2026. GitHub followed weeks later with the same shift for Copilot. Analysts project that when pricing normalizes to reflect real infrastructure costs, enterprise AI bills could rise another 30 to 50 percent above current levels.
The market noticed the divergence between spending and returns in June 2026, when 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. Nvidia, Micron and AMD led the losses.
For investors, the question is whether AI can pay for itself before the money runs out. Tesla's valuation rests on deploying AI at scale across its Robotaxi network and Optimus humanoid robot, not on selling cars — yet the company cannot manage a few thousand dollars of weekly token spend per engineer without imposing caps. If the cost of tokens has already exceeded the cost of the employees they were meant to replace, the gap between promise and economics is widening, not closing.
This article is for informational purposes only and does not constitute investment advice.
