Microsoft Pulled The Plug On 100,000 Claude Code Licenses Due To Rising Costs

Uber burned its entire 2026 AI budget by April. Microsoft killed 100,000 AI licences. And NVIDIA says compute now costs more than employees. Something is seriously off.

Every line of AI-generated code carries a price tag.

Source : ABPLIVE AI

Show Quick Read

Key points generated by AI, verified by newsroom

• Microsoft suspends Claude Code for engineers due to escalating costs.
• AI coding tools become prohibitively expensive for large companies.
• Token-based pricing drives significant, unforeseen AI operational expenses.

Microsoft has pulled Claude Code access for around 100,000 of its engineers, with the deadline set for June 30. The company had rolled out the AI coding tool in December 2025, and adoption grew quickly among its engineering workforce. But as usage climbed, so did the bills. The core issue is token-based pricing, where every line of code generated carries a cost. 

At the scale Microsoft operates, those costs became impossible to justify, forcing the company to cancel all Claude Code licenses despite having invested $5 billion in Anthropic, the company that makes it.

Why Are AI Coding Tools Becoming Too Expensive to Run?

The cost problem is not unique to Microsoft. Uber's CTO Praveen Neppalli Naga said the company burned through its entire 2026 AI budget by April, stating: "The budget I thought I would need is blown away already." 

Heavy users of these tools are spending between $500 and $2,000 per month each. At Uber, engineers even built internal leaderboards to track token usage, which ended up accelerating spending rather than controlling it.

NVIDIA's VP added to the conversation with a blunt observation: "For my team, the cost of compute is far beyond the costs of the employees." This came from a VP at the company that manufactures the very chips powering these AI systems.

What the Numbers Say About AI Spending Going Forward

Goldman Sachs forecasts token consumption will grow 24 times by 2030. Gartner projects token prices will fall by 90%, but that does not mean total costs go down. AI agents consume 5 to 30 times more tokens than standard chatbots, meaning lower unit prices get offset by far greater usage.

Meta tracked 60 trillion tokens in 30 days through an internal dashboard called "Claudeonomics" before shutting it down. Amazon employees began "tokenmaxxing" to game internal AI leaderboards.

The four largest tech companies are collectively spending $725 billion on AI infrastructure this year. Yet the same companies deploying these tools are discovering that, in practice, AI is costing more than the human workers it was supposed to assist.

Microsoft Wants Older Employees Out, Offering Voluntary Retirement As AI Takes Over