Microsoft has begun canceling the majority of its internal Claude Code licenses and is directing engineers to migrate to GitHub Copilot CLI by June 30, 2026, according to The Verge via Developer Tech. The reversal comes six months after the company opened Claude Code access to thousands of developers, project managers, and designers. Separately, Uber exhausted its entire 2026 AI coding tools budget within four months after Anthropic’s tool spread across roughly 5,000 engineers, as CTO Praveen Neppalli Naga confirmed to The Information.
Microsoft’s Experiences + Devices Division Leads the Cutoff
The transition is already underway in Microsoft’s Experiences + Devices division, which includes teams working on Windows 11, Microsoft 365, Outlook, Teams, and Surface, according to Developer Tech. Engineers in the division have been told to move workflows to Copilot CLI ahead of the June 30 deadline. Executive VP Rajesh Jha said Claude Code had played a role in the company’s internal learning around AI coding tools but that Copilot CLI gives Microsoft a product it can shape more directly with GitHub, including closer alignment with internal repositories, workflows, and security requirements. The decision does not affect Microsoft’s broader commercial relationship with Anthropic: the Foundry deal (up to $5 billion investment in Anthropic) and Anthropic’s $30 billion commitment to purchase Azure compute capacity both remain intact, as Livemint reported.
Uber’s Budget Collapse in Numbers
Uber rolled out Claude Code to its engineering organization in December 2025, according to Forbes. Adoption climbed from 32% of engineers in February to 84% classified as agentic coding users by March. By spring, 95% of Uber engineers used AI tools monthly, and roughly 70% of committed code originated from those tools. About 11% of live backend updates were written by agents with no human in the loop.
Monthly cost per engineer ranged from $150 to $250 on average, with power users running between $500 and $2,000. Naga himself spent $1,200 in a two-hour session during a personal demo, Forbes reported. Uber’s total R&D spend reached $3.4 billion in 2025, up 9% year over year. The company compounded the cost dynamic by ranking engineers on internal leaderboards based on Claude Code usage, creating a cultural incentive to consume more tokens.
Token Economics: Cheaper Prices, Higher Bills
The structural problem extends beyond these two companies. Token-based consumption pricing does not behave like the per-seat software line items CFOs know how to model. Goldman Sachs has forecast that agentic AI systems could drive a 24x increase in token consumption by 2030, reaching 120 quadrillion tokens per month, according to Livemint. Gartner projects that inference costs on a one-trillion-parameter model will fall nearly 90% by 2030 compared to 2025, but cautioned that consumption growth can outpace falling unit costs. “Chief Product Officers should not confuse the deflation of commodity tokens with the democratisation of frontier reasoning,” said Will Sommer, senior director analyst at Gartner, as quoted by Livemint.
Bryan Catanzaro, Vice President of applied deep learning at NVIDIA, addressed the cost problem directly in a recent interview with Axios: “For my team, the cost of compute is far beyond the costs of the employees,” as reported by Livemint.
The Enterprise Pricing Fork
Microsoft’s own response illustrates one path forward. Microsoft 365 Copilot Enterprise sells at $30 per user per month with an annual commitment, capping vendor upside and giving finance teams a flat line item. Anthropic’s token-based model gives the vendor unlimited upside on heavy users and gives finance teams almost no forward visibility. On May 13, Anthropic announced that paid Claude subscribers would face a separate monthly credit meter for agent tools and third-party harnesses, billed at full API rates starting June 15, adding another layer of consumption pricing to the mix. The early signals from Microsoft, Uber, and NVIDIA suggest that at current pricing and usage patterns, the cost structure of large-scale AI agent deployment remains unresolved.