Microsoft pulls back from Claude Code: too popular, too expensive

When Microsoft opened access to Anthropic's Claude Code to thousands of its developers, designers, and product managers six months ago, it expected a productivity boost. Instead, it got bill shock. According to The Verge, the company is now cancelling most direct licenses and redirecting engineers to GitHub Copilot CLI — its own tool, which it keeps under control.

Claude Code caught on among Microsoft employees surprisingly fast. Too fast. Mass scaling of its usage exposed a crack in AI economics: the more employees use AI, the higher the bill — and it can grow uncontrollably. The license cancellations do not affect Microsoft's strategic partnership with Anthropic (Foundry, $5 billion investment), but exclusively involve internal deployment among its own employees.

Nvidia: "AI costs more than people"

An even sharper statement came from the company profiting most from the AI boom. Bryan Catanzaro, vice president of applied deep learning at Nvidia, said in an interview with Axios: "For my team, the cost of computing power far exceeds the cost of employees." In other words — even at Nvidia, which manufactures AI chips, running AI works out more expensive than paying people who would do the same work on their own.

This admission resonates with a MIT CSAIL study from 2024 that analyzed the economic viability of replacing human labor with AI. Researchers found that AI automation would be economically viable for only 23% of job positions where vision is the main component of the work. In the remaining 77% of cases, it was cheaper to leave the work to humans.

Tokenmaxxing: when companies push AI too far

Across Silicon Valley in 2026, a phenomenon that Fortune calls "tokenmaxxing" has proliferated. Companies actively incentivize employees to use as many AI tokens (the basic units of AI computation) as possible, and even create internal leaderboards:

• Meta launched a dashboard called "Claudeonomics" that tracks which employees use AI the most
• Amazon directly urges its employees to tokenmaxx — using the maximum amount of AI tokens
• Uber motivated teams through internal leaderboards and CTO Praveen Neppalli Naga admitted in April that the company burned through its entire 2026 AI coding tools budget in just four months

The result? Instead of savings, companies got cost shock. Agentic AI, meaning autonomous systems that independently perform multi-step tasks, consumes an estimated up to 1000× more tokens than standard chatbots. Every "intelligent" decision by an agent means another and another compute cycle — and another line item on the invoice.

The cheap token paradox: unit price drops, bill rises

At first glance, it makes no sense. The price of individual AI tokens is, after all, dropping dramatically. According to Gartner, inference costs (query processing) for a billion-parameter model will fall by more than 90% by 2030 compared to 2025. Yet at the same time, Goldman Sachs predicts that agentic AI will cause a 24-fold increase in token consumption by 2030 — to an incredible 120 quadrillion tokens monthly.

In other words: tokens are getting cheaper, but we're consuming orders of magnitude more of them. Gartner analyst Will Sommer adds: "Product directors should not confuse commodity token deflation with the democratization of advanced reasoning." Even with falling unit prices, total corporate AI bills will rise.

Other giants have already felt this paradox. Keith Lee, professor of AI and finance at the Swiss Institute of Artificial Intelligence, described the current situation to Fortune as a "short-term mismatch". According to him, fixed subscriptions (typically $20–200 per month for developer AI tools) don't cover the actual operating costs for the most active users. The result will be an inevitable shift to a consumption model (pay-per-use), which companies like OpenAI are already testing with GPT-5.5 at double the price of its predecessor.

What this means for the Czech Republic and Europe

While American companies battle rising bills, European businesses are only just beginning to adopt AI tools. According to US Fed data, only about 18% of American companies had adopted AI tools by the end of 2025 (a 68% year-over-year increase). In Europe, penetration is even lower, which in this context can paradoxically be an advantage — companies can learn from the mistakes of American giants.

For Czech companies and developers, several lessons follow:

• Don't blindly follow the trend of deploying AI everywhere. It pays to measure concrete return on investment (ROI) — not just token count or hours spent with an AI tool.
• The consumption model is fairer for smaller teams. Fixed licenses at $25–200 per month per developer look tempting, but if you use the tool intensively, the final bill can be many times higher. Czech companies should negotiate transparent pricing models.
• The EU AI Act adds additional compliance costs. European companies must factor in not only the price of tokens, but also the costs of ensuring regulatory compliance — audits, documentation, AI decision-making transparency.

In the Czech Republic, no major employer has yet publicly announced that it is restricting AI tools for its developers due to costs. Companies like Ecomail, on the contrary, are actively integrating AI into their services. However, as tools like Claude Code, Cursor, or GitHub Copilot become standard equipment for developers, the question of "how much is this actually costing us" will become increasingly pressing for Czech CTOs as well.

The future: cheaper AI, or cheaper people?

Analysts' outlook offers mixed signals. Gartner predicts that inference will be 90% cheaper by 2030. Yet at the same time, agentic models consume orders of magnitude more tokens and AI providers, according to experts, won't pass on the full savings to end customers — they'll keep a portion to cover their own infrastructure.

Jensen Huang, Nvidia CEO, nonetheless believes that every employee will one day have 100 AI agents at their disposal. If this vision materializes, it will mean that the economic equation must change radically — otherwise companies would go bankrupt on token bills long before they even managed to deploy the agents.

The reality of May 2026 is more sobering: AI is an amazing tool, but its operational costs do not yet justify the wholesale replacement of human workers. Microsoft has grasped this. Uber has grasped this. And soon, every company that succumbs to the pressure to "use AI for everything" without calculating the actual costs will grasp it too.