Companies Spent Millions on AI in 2026. Most Cannot Show It Worked.

Quick Facts

• Gartner projects AI agent software spending will reach $207 billion in 2026, up 139% from $86.4 billion in 2025.
• MIT found that 95% of AI pilots deliver zero measurable P&L impact, while IBM puts the share of initiatives delivering expected ROI at just 25%.
• Uber exhausted its entire 2026 AI coding budget by April after Claude Code spread to roughly 5,000 engineers, with no clear link between usage and consumer-facing output.

Enterprise AI spending is accelerating at a pace that far outstrips the ability of most companies to prove the money was well spent. Hyperscalers are on track to spend $675 billion on AI infrastructure in 2026, up 63% from the prior year. Companies plan to spend an average of 1.7% of revenue on AI this year, more than doubling 2025 levels. Within two years, AI is expected to consume 25% to 50% of total IT budgets at many large enterprises.

The returns are not keeping up. Morgan Stanley found that only 21% of S&P 500 companies could cite a measurable AI benefit at all. S&P Global found that 42% of companies abandoned most of their AI projects in 2025. Approximately 30% of generative AI projects are scrapped after proof of concept, driven by escalating costs, poor data quality, and unclear business value.

The Measurement Gap

The core problem is not adoption. It is accountability. A 2026 survey of 100 senior enterprise AI leaders found that more than two-thirds of enterprises still rely on estimates rather than measured financial results to assess ROI. They track time saved and usage volumes, but not what those inputs translate to in financial terms.

IDC’s 2025 AI Investment Survey found that 61% of enterprises cannot demonstrate measurable ROI from their AI investments because they never established a baseline before deploying. While over 70% of organizations report positive AI ROI, fewer than 1% report significant returns of 20% or more. Most see only 1% to 5% gains, typically measured as productivity improvements rather than bottom-line impact.

The Uber Warning

The clearest illustration of this problem came from Uber. In December 2025, the company gave its engineers access to Anthropic’s Claude Code and set up internal leaderboards tracking token consumption. Adoption climbed from 32% of engineers in February 2026 to 84% classified as agentic coding users by March. By April, the entire AI coding budget for the year was gone.

Monthly cost per engineer ranged from $150 to $250 on average. Power users ran between $500 and $2,000. Uber CTO Praveen Neppalli Naga confirmed the budget overrun to The Information, saying the company was back to the drawing board on its assumptions. He reported spending $1,200 himself during a single two-hour personal demo.

Uber President and COO Andrew Macdonald put the problem plainly in a recent Rapid Response podcast interview. “That link is not there yet,” he said of the connection between Claude Code usage and consumer-facing innovation. “Maybe implicitly there’s more that is getting shipped, but it’s very hard to draw a line between one of those stats and ‘Okay now we’re actually producing like 25% more useful consumer features.'”

Naga gave the episode a name on X: the end of the “tokenmaxxing era.” Uber responded by capping all employees at $1,500 in monthly token spending per AI coding tool.

Microsoft Pulls Back

Uber was not alone. Microsoft cancelled thousands of internal Claude Code licenses by June 30, 2026, moving its Experiences and Devices engineering team to GitHub Copilot CLI. Anthropic’s Claude models remain accessible through Microsoft Foundry and inside Microsoft 365 Copilot, but the interface, agent surface, and cost ownership all changed. Claude Code had become, by internal accounts, perhaps a little too popular among Microsoft engineers.

CEO Satya Nadella has said Microsoft now generates up to 30% of its code using generative AI. The company’s decision to consolidate tooling signals that even the largest AI vendors are tightening governance over how their own teams consume AI resources.

What Executives Face Now

The governance gap is creating pressure at the top. A recent survey found that 54% of C-suite executives admit that adopting AI is tearing their company apart. Nearly all IT decision-makers, 97%, report challenges when implementing new AI initiatives, with systems integration ranking as the top obstacle.

Vendor dependency is emerging as a secondary risk. According to VentureBeat’s reporting, 81% of enterprise leaders are concerned about AI vendor dependency, but only 6% say they could switch providers without material disruption. Nearly half, 47%, say a key business function would stop entirely if their primary AI provider went dark.

Spending will keep climbing. Cumulative AI infrastructure investment is expected to approach $3 trillion to $4 trillion by end of the decade. The companies that pull ahead will not be the ones that spend the most. They will be the ones that build the measurement systems to know what is actually working before the budget runs out.

Read more: Companies are spending millions rewiring how AI gets used. Almost none can prove it’s working.