Microsoft is pulling licenses for Anthropic's Claude Code across its Experiences and Devices division—the team behind Windows, Office, Teams, and Surface. The reason? Token costs spiraled beyond control. Engineers will migrate to GitHub Copilot CLI, Microsoft's cheaper internal alternative, by June 30. This isn't a retreat from AI. Claude models remain accessible through Copilot, and Microsoft's broader $5 billion investment in Anthropic stays intact. The move targets runaway operational expenses, not the technology itself.

This pattern is spreading. Uber's CTO recently disclosed that the company burned through its entire 2026 budget for Claude Code and Cursor in just four months. Nvidia's VP of applied deep learning research noted that compute costs now exceed employee salaries on his team. The economics of token-based pricing break at scale. Heavy usage, once celebrated as adoption success, becomes a budget crisis.

Microsoft's decision coincides with broader workforce adjustments. In April, the company offered voluntary retirement packages to roughly 8,700 U.S. employees—about 7% of its domestic workforce—under a 'Rule of 70' formula combining age and service years. CFO Amy Hood confirmed headcount declined year-over-year in Q3 and expects the trend to continue. Meanwhile, capital expenditures will exceed $40 billion this quarter to expand AI infrastructure. The company's AI business hit a $37 billion annual revenue run rate, up 123%.

The tension is clear: AI tools deliver genuine productivity gains, but at enterprise scale, the math gets ugly. When thousands of developers query models dozens of times daily, token charges pile into millions. The promise of replacing people with models collides with the invoice. For now, companies ration access and favor internal solutions. Costs will likely fall as inference improves and enterprise licensing matures. But this recalibration matters. If the tools supposed to replace workers cost more than the workers themselves, the timeline for displacement stretches.

Engineers inside Microsoft will keep using AI—just through a different interface with tighter cost controls. That compromise captures enterprise AI today: useful, expensive, and still evolving.

Source: Webpronews