Companies are currently urging their workforce to maximize AI utilization in a bid to harness its productivity benefits. However, this intense push is starting to reveal systemic flaws, potentially leading to irreversible consequences.
According to The Verge, Microsoft has reportedly begun phasing out the majority of its direct Claude Code licenses, instead directing engineers towards GitHub Copilot CLI. This shift occurs merely six months after the company initially granted access to Claude Code, encouraging thousands of its developers, project managers, designers, and other employees to experiment with coding. The technology quickly gained traction—perhaps too much. The sheer scale of its usage by employees is now prompting the firm to reverse course on a tool its own engineers had come to depend on. It’s important to note that canceling Claude Code licenses will not impact Microsoft’s Foundry agreement, which includes an investment of up to $5 billion in Anthropic and granting Foundry customers access to Claude models, alongside Anthropic’s commitment to purchase $30 billion in Azure compute capacity, as per The Verge.
Microsoft isn’t an isolated case in scaling back internal AI usage. Uber’s CTO, Praveen Neppalli Naga, disclosed to The Information in April that the company had depleted its entire 2026 budget for AI coding tools within a mere four months. This follows a period where the company actively encouraged adoption through internal leaderboards that ranked teams based on their AI tool usage.
These recent reports could dampen the significant investments made by leading tech firms in artificial intelligence. While some still envision an AI “renaissance” or “revolution,” the financial burden of adoption is proving to be a persistent hurdle. Furthermore, these developments suggest that the economic implications of replacing or supplementing human labor with AI might be more complex than initial projections indicated. This aligns with recent remarks from Bryan Catanzaro, vice president of applied deep learning at Nvidia, in an interview with Axios.
Catanzaro stated, “For my team, the cost of compute is far beyond the costs of the employees.”
Anthropic did not immediately respond to Fortune’s request for comment, nor did Microsoft provide a statement.
An intriguing paradox is emerging in AI: while individual tokens become cheaper, overall expenses are ballooning.
Uber and Microsoft are not the only companies pushing employees to extensively use AI. Similar to Uber, a Meta employee created a leaderboard, aptly named “Claudeonomics” after Anthropic’s AI model, to monitor which workers are utilizing AI the most. Amazon is encouraging its employees to “tokenmax”—to maximize their use of AI tokens, which are the fundamental units of AI computation.
However, with a token-based pricing structure, increased usage and improved efficiency paradoxically lead to higher costs. Goldman Sachs recently predicted that agentic AI could drive a 24-fold increase in token consumption by 2030 as consumers and businesses adopt AI agents, potentially reaching an astounding 120 quadrillion tokens per month. As businesses leverage AI agents for productivity gains, aggregate costs could escalate sharply, even if the price per individual token declines.
