Companies Cut AI Spending: What It Means for Investors

For two years, corporate America treated artificial intelligence spending like a badge of honor. Now the badge is getting expensive — and companies are starting to take it off. From Uber burning through its entire 2026 AI budget in four months to Microsoft canceling internal AI tool licenses, a growing roster of household names is rationing, capping, and in some cases scrapping AI programs as the promised returns fail to show up. For investors who rode the AI trade to record highs, the shift marks a critical inflection point: the multi-trillion-dollar AI buildout is colliding with a wall of disappointing return-on-investment data, and the market is beginning to notice.

The 'Tokenmaxxing' Hangover: Why Boards Are Rationing AI

The Economist recently declared that 'the rage for tokenmaxxing is coming to an end.' The term describes the corporate habit of treating AI tokens — the units of computation that power large language models — as a nearly unlimited resource. After two years of pushing employees to adopt AI everywhere, boards have discovered that token-hungry applications and autonomous AI agents generate spiraling, unpredictable bills that rarely tie back to a measurable business outcome.

The Wall Street Journal reported in late May that 'corporate America is starting to ration AI' as the cost of AI computing tokens skyrocketed. A July follow-up was even blunter, concluding that companies had 'suddenly decided to stop blowing' money on AI. Among the rationing tactics the Journal identified: limiting new hires' access to the most expensive frontier models and reserving advanced AI systems for planning work rather than routine tasks.

The New York Times reported in August that a new discipline — 'tokenomics' — has emerged purely to measure whether all that AI spending produces any return at all. The existence of the field is itself a signal: companies now treat AI spend as a cost to be managed, not a badge of innovation.

From Uber to Starbucks: The Companies Slashing AI Budgets

The clearest case study is Uber. The ride-hailing company exhausted its entire 2026 AI coding budget — reported at roughly $3.4 billion — in just four months, with no measurable return to show for it. Uber's response was to cap AI spending at about $1,500 per employee per month and ration access to costly AI coding tools.

Microsoft, one of the biggest backers of the AI boom, canceled its internal Claude Code licenses and is trimming its workforce even as it pours record sums into AI data centers. Starbucks retired an AI inventory system after nine months because it could not reliably handle basic tasks, from counting products to distinguishing one item from another. Amazon reportedly scrapped an internal AI-use program and began capping budgets, while Meta, Walmart and Cisco have all moved to limit internal AI spending, according to reports compiled by industry trackers.

The pullback is not limited to a handful of large names. Ramp, the corporate card and bill-pay platform that observes spending across roughly 70,000 to 85,000 U.S. businesses, found that even the heaviest AI adopters are now trimming. Its data shows the top 1% of AI-spending companies cut their AI spend per employee by nearly 10% in a single month.

The Data: AI Spend Per Employee Just Dropped Nearly 10%

According to the Ramp AI Index, AI spending per employee among the top 1% of U.S. companies fell 9.7% in August — a rare decline after months of uninterrupted growth. TechCrunch, which first surfaced the figure, asked pointedly whether the drop was 'summer doldrums or a warning sign.' The Decoder called it evidence that the heaviest AI adopters are now actively managing their costs down.

The retrenchment aligns with a wave of surveys. DoiT found that 79% of enterprises overspent their AI budgets in 2026, while only 15% said they can actually calculate their AI return on investment. Bain's research found nearly 40% of companies measuring AI cost savings landed below 10%, even though most had targeted savings of 11% to 20%. BCG reported that AI's share of IT budgets has more than tripled from 5% to 17% — a sharp jump that, combined with weak returns, has put chief financial officers on high alert.

Why AI ROI Keeps Disappointing — And What the Numbers Show

The return-on-investment math on AI has repeatedly come in below the hype. A benchmark spanning more than 1,300 CFOs found only 14% report measurable ROI from AI, and that 95% of enterprise AI initiatives fail to deliver expected outcomes. Separate research from the FII Institute found just 25% of AI initiatives have met expected returns, and only 16% have scaled across the enterprise.

Forbes distilled the problem in a July headline: 'AI Costs More Than The People It Replaced.' New research from Gartner goes further, suggesting that companies which cut human jobs to fund AI are backfiring — yielding zero financial gains for major corporations. Accenture, which surveyed 750 senior executives at billion-dollar-plus companies, found that falling AI model prices are not shrinking token bills; they are inflating them as usage spreads faster than efficiency gains.

The disconnect between executives is striking. Separate data suggests 65% of CEOs and their CFOs disagree on AI's value, while Gartner puts total AI spending in 2026 at roughly $2.5 trillion. With so much money in motion and so little measured return, it is no surprise that nearly 70% of companies told one survey they are prepared to scale back AI budgets this year if projects fail to deliver.

Where Things Stand: AI Stocks Wobble as CEOs Warn of a Slowdown

The spending pullback is now colliding with the stock market's AI trade. On September 14, global AI-linked stocks fell after Anthropic CEO Dario Amodei published an essay calling for a slower pace of AI development over safety concerns, with OpenAI's Sam Altman and other industry leaders echoing similar warnings. OpenAI also said it would not pursue an initial public offering this year. The Guardian and Reuters reported that chipmakers and AI infrastructure names tumbled on the news.

Even so, the hyperscalers are still spending enormous sums. Amazon, Google, Microsoft and Meta are collectively guiding to roughly $674 billion to $750 billion in 2026 capital expenditures, much of it for AI data centers. Morgan Stanley forecasts total AI spending will surpass $1.3 trillion by 2027. Nvidia's data center revenue more than doubled to $89 billion in its latest quarter, yet the stock is up only about 11% this year — a sign that investors are no longer rewarding AI growth the way they once did, and that demand may be cooling at the edges.

What Happens Next: What the Pullback Means for Investors

The tension for investors is now two-sided. On one side, enterprise customers are cutting AI spending because the tools are not yet paying for themselves. On the other, the big cloud providers and chipmakers depend on exactly that enterprise demand to justify their data center buildouts. If corporate AI budgets keep shrinking while hyperscaler capex keeps rising, something eventually has to give.

Bloomberg has reported that demand for an Nvidia-backed data center company's IPO plunged, revealing 'fresh cracks in the AI funding boom.' For investors, the key indicators to watch are monthly adoption data such as the Ramp AI Index, enterprise ROI disclosures, and whether hyperscalers begin trimming their own capex guidance. The companies that can prove a measurable return on AI — rather than simply burn tokens — are the most likely winners of the next phase, while the names most exposed to unproven AI spending face the steepest risk.

The Bottom Line: Key Takeaways on the AI Spending Pullback

• Uber exhausted a roughly $3.4 billion annual AI budget in four months and now caps per-employee AI spending at about $1,500.
• Microsoft, Amazon, Meta, Walmart, Cisco and Starbucks have all rationed or scrapped AI programs.
• The top 1% of AI-spending companies cut AI spend per employee by nearly 10% in August, according to Ramp data.
• Only 14% of CFOs report measurable ROI from AI, and 79% of enterprises overspent in 2026.
• AI stocks wobbled in September after industry CEOs called for a slowdown, even as hyperscalers kept guiding to record capex.