Corporate America starts to ration AI as costs soar

Microsoft, the company that bet tens of billions on AI through its OpenAI partnership, canceled most of its direct licenses for Anthropic’s Claude Code in May 2026. The reason was straightforward: internal usage spiked compute costs to levels that made the tool financially untenable. Staff were redirected to GitHub Copilot, Microsoft’s own AI coding assistant.

Uber’s experience was arguably more dramatic. The ride-hailing giant burned through its entire annual AI coding budget in roughly four months. Its COO acknowledged that the costs are becoming “harder to justify.”

Nvidia VP Bryan Catanzaro stated in April 2026 that compute costs now exceed employee costs for his team.

Enterprise AI spend is projected to reach 1.7% of revenue in 2026, more than double 2025 levels. Yet fewer than 1% of executives report significant ROI of 20% or more from their AI investments. Most are seeing returns in the 1-5% range.

Multiple surveys indicate that enterprise AI uptake has stagnated despite the massive financial commitments. Most organizations report cost savings under 10% in the specific areas where AI has been deployed.

Bitcoin miners, facing declining profitability in their core business, are increasingly reallocating resources toward AI and high-performance computing data centers. AI cloud revenue runs between $1,600 and $4,000 per megawatt-hour, compared to just $80 to $151 for Bitcoin mining. AI data center setups require roughly 33 times higher capital expenditure per megawatt compared to mining operations.

No specific crypto tokens have emerged directly tied to escalating AI compute costs.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.