Tech giants are shifting strategies due to AI costs
Major tech companies have begun re-evaluating their AI implementation strategies. Instead of the expected efficiency, they face a new problem: the more actively employees use AI tools, the faster costs rise, making the economic model difficult to sustain. Specifically, Microsoft has stopped providing Claude Code licenses to its engineers, migrating them to the GitHub Copilot CLI platform. This is reported by Ixbt.com reports .
This decision was made just six months after Anthropic tools began to be widely used. Although the strategic partnership and multi-billion dollar cloud agreements between Microsoft and Anthropic remain in effect, the internal operating model is shifting to a cost-saving mode. A similar situation was observed at Uber: the company had already spent its 2026 AI budget within the first four months of this year.
At the current stage, the main contradiction is that while AI tools increase productivity for individual tasks, they sharply increase the need for computing power and overall expenses. According to Goldman Sachs analysts, by 2030, the widespread adoption of agentic AI systems could push monthly token consumption into the hundreds of quadrillions, placing exponential pressure on infrastructure.
Gartner research predicts that by 2030, the cost of processing requests in large language models will drop by 90 percent. However, this does not mean corporate expenses will decrease. Complex agentic systems consume significantly more tokens per task, offsetting the drop in the cost per computing unit. As a result, a paradoxical dynamic is emerging where total costs rise as the technology becomes cheaper.
Tech giant executives continue to promote the idea of an "agentic future" where digital assistants are fully integrated into workflows. However, recent data suggests that scaling this model may be much more expensive than initially projected.
