The AI Spending Hangover: From Tokenmaxxing to a Cost-Cutting Reappraisal

This year was meant to be the one when artificial intelligence finally delivered on its promise to slash corporate costs. Instead, the technology is burning through budgets at an alarming rate. Alphabet’s chief executive recently noted that some of the world’s largest companies have already exhausted their entire annual AI allocations by May. The ride-hailing firm Uber provides a stark case study: its engineering team consumed the full 2026 budget for the coding assistant Claude Code in just four months, with per-engineer monthly costs swinging wildly between $500 and $2,000. Dubbed “tokenmaxxing”, this frenzied consumption of the basic billing units that underpin AI services has prompted a sector-wide reassessment, as the initial thrill of automation gives way to an unsentimental cost-benefit calculus.

Behind the sticker shock is the end of what Kevin Simback of Delphi Labs calls the era of “subsidised intelligence”. To hook customers, AI providers long offered cut-rate access, with venture capital effectively absorbing the true expense. That model is now being retired. Major players such as OpenAI and Anthropic are eyeing initial public offerings later this year and need to demonstrate a path to genuine profitability. Prices are climbing, and enterprises are reacting. Microsoft has revoked Claude Code licences for one division, steering staff towards its own GitHub Copilot, while Meta, which earlier this year urged employees to treat token usage as a productivity metric, is now back-pedalling. Its chief technology officer has cautioned that no one should be using AI merely for the sake of it.

Viewed from Europe, the AI cost dilemma intersects with deeper anxieties about productivity. In Italy, the governor of the Bank of Italy, Fabio Panetta, has framed AI as a possible lever to jolt the country’s sclerotic growth. Yet, as Il Giornale pointedly observes, a crucial element is missing from the eurozone conversation: the market itself. While a recent survey by the business association Cna shows that 35.6% of its member artisans now employ at least one AI tool – a near seven-fold increase in eighteen months – the link between adoption and measurable economic gains remains, for now, more an article of faith than a settled fact.

The sobering picture extends to the bottom line. A Gartner survey of 350 global firms with over a billion dollars in annual revenue found that four in five have reduced headcount since deploying AI. Crucially, however, the layoff rate bore no statistical relation to return on investment. The companies that cut deeply performed no better than those that trimmed lightly. Analysts flag a pattern of “AI washing”: using layoffs to signal technological savviness to investors rather than to unlock genuine value. The message from the front lines is clear: the AI revolution is real, but the path from massive expenditure to sustainable profit is still being charted – and it runs through disciplined, strategic deployment, not indiscriminate cost-cutting.