Companies Reducing Artificial Intelligence Spending: Why Big Tech Is Pulling Back in 2026

Companies reducing artificial intelligence spending have become one of the biggest business stories of 2026. After two years of unlimited AI budgets and “adopt everything” mandates, some of the world’s largest companies — Uber, Walmart, Cisco, Amazon, and Meta — are now capping usage, demanding ROI proof, and in some cases shutting down internal AI programs entirely. The reason is simple: the bills got enormous, and nobody can quite prove what all that spending bought.

Which Companies Are Cutting AI Spending?

The pullback isn’t a rumour — executives are saying it on the record.

Uber is the most concrete example. President and COO Andrew Macdonald recently said the company is finding it “harder to justify” its AI token outlay, admitting it’s difficult to draw a line between usage stats and “25 per cent more useful consumer features.” Uber blew through its entire planned 2026 AI budget by April, and has now capped individual employees at $1,500 per month in token spending per AI tool.

Walmart has taken a similar step back on its internal AI agent. Global CTO Suresh Kumar said usage of the company’s “Code Puppy” vibe-coding platform “really skyrocketed” — and that Walmart is now deliberately stepping back, asking employees to pick the right tool for each task instead of reaching for the largest model by default.

Cisco is wrestling with the same math at a bigger scale. President and Chief Product Officer Jeetu Patel noted that with AI agents, “for every human you might have 10, 100 or on the aggressive side 1,000 agents… They just keep working and that consumes a chunk of compute.” His engineers want more tokens; the company has to figure out how to fund them.

Amazon warned employees last month to stop using AI “just for the sake of using AI” after engineers began deploying agents partly to climb internal adoption leaderboards. Meta went further — it shut down its internal “Claudeonomics” leaderboard after 85,000 employees burned through more than 60 trillion tokens in 30 days.

The Numbers Behind the Pullback

This isn’t just anecdote. Spending data from payments company Ramp, which tracks activity across roughly 70,000 companies, shows business AI adoption stalling: 56% of its customers paid for AI products in August 2026, up just 0.4% from July — a dramatic slowdown from the growth investors had come to expect.

More striking: AI spending per employee among the top 1% of AI-using companies fell nearly 10%, to $7,205. Part of that reflects cheaper tokens rather than less usage — average token costs dropped to $0.68 per million tokens from a March peak of $1.15 per million as OpenAI and Anthropic competed on price. But increased usage hasn’t fully offset the price cuts, and many companies are quietly shifting to older, cheaper models rather than the newest frontier releases.

A separate 2026 report on AI cost governance found that 62% of organisations had an unexpected AI cost materially alter a business decision in the past year. Among those, 40% required board-level escalation, 33% implemented emergency spending freezes, and 25% delayed or cancelled an AI initiative outright.

Why Are AI Costs So Hard to Control?

Three forces collided to create this moment.

First, token-based pricing is inherently unpredictable. Unlike a software licence with a fixed price, token consumption scales with every prompt, every agent loop, every retry. One developer’s three-person team reportedly spent over $1.3 million in tokens in a single month running agentic AI tools. As one analyst put it, in some cases the cost of tokens now exceeds the cost of the employee within a month or two.

Second, ROI remains elusive. Uber’s COO captured the industry’s frustration: nobody has convincingly connected AI token spend to measurable consumer outcomes. Companies spent two years encouraging experimentation with few guardrails; now finance departments want receipts.

Third, agentic AI multiplied consumption. Individual chatbots were manageable. But AI agents that work autonomously — running hundreds or thousands of sub-tasks — consume tokens at a scale no one budgeted for. Cisco’s “1,000 agents per human” framing shows why the math broke.

What This Means for Businesses Using AI

If you’re running a business (not a tech giant), the takeaway isn’t “abandon AI” — it’s “spend deliberately.” The companies cutting back aren’t leaving AI; they’re optimising it:

• Right-size models: Use cheaper, older models for routine work and reserve frontier models for tasks that genuinely need them.
• Set guardrails early: Usage caps, approval workflows, and chargeback mechanisms prevent the surprise bills Uber and Meta got.
• Measure unit economics: Gartner now recommends tracking “AI unit economics” — linking token consumption to measurable outputs — rather than just total spend.
• Consolidate tools: Microsoft moved developers onto internally built tools partly to escape token-based third-party billing.

The broader market signal is arguably healthy: competition between OpenAI and Anthropic is driving token prices down, making AI more accessible even as the biggest spenders get disciplined.

Is the AI Boom Over?

No — but the “spend first, measure later” phase is ending. What we’re seeing is the market maturing: from land-grab adoption to cost-conscious deployment. The companies reducing artificial intelligence spending today are the same ones that will likely spend more efficiently tomorrow, with better governance and clearer ROI demands.

For investors, the question is whether enterprise AI revenue grows fast enough to justify the infrastructure buildout. For everyone else, the lesson is simpler: AI is a tool with a meter running. The winners won’t be the biggest spenders — they’ll be the smartest ones.

Frequently Asked Questions

Which companies are reducing AI spending in 2026?
Uber, Walmart, Cisco, Amazon, and Meta have all publicly pulled back — through usage caps ($1,500/month per tool at Uber), deliberate step-backs (Walmart), funding reviews (Cisco), usage warnings (Amazon), and program shutdowns (Meta’s internal leaderboard). Microsoft has also consolidated developers onto internal tools to cut token billing.

Why are companies cutting their AI budgets?
Token costs exploded beyond forecasts (Uber spent its full 2026 AI budget by April), ROI is hard to prove, and agentic AI multiplied consumption far beyond what chatbots used. Gartner notes token consumption is becoming a “meaningful component” of enterprise operating costs with unclear connection to business outcomes.

How much are companies spending on AI per employee?
Among the heaviest AI-using companies, spending averaged $7,205 per employee (Ramp data, August 2026) — down nearly 10% as token prices fell and companies shifted to cheaper models. One in four companies has delayed or cancelled an AI project over cost, according to 2026 survey data.

Are AI token costs still rising?
Interestingly, no — average token prices fell to $0.68 per million tokens from a $1.15 peak in March 2026 as OpenAI and Anthropic competed on price. The problem is volume: agentic workloads consume so many tokens that total bills keep climbing even as unit prices drop.

Should small businesses cut AI spending too?
Not necessarily cut — but govern it. Set usage caps, choose cheaper models for routine tasks, consolidate overlapping tools, and track what each AI workflow actually produces. The enterprise lesson of 2026 is that unmanaged AI spending surprises everyone; managed AI spending pays for itself.

Final Thoughts

The wave of companies reducing artificial intelligence spending marks a turning point: AI’s wild-spending era is giving way to its accountability era. Uber’s blown budget, Walmart’s step back, and Meta’s shuttered leaderboard all tell the same story — the technology works, but the economics need management. For businesses watching from the sidelines, that’s actually good news: the tools are getting cheaper, the best practices are getting clearer, and the winners will be those who spend smart, not just big.