Rising AI Costs: Companies Rethink Strategies Due to 24x Surge in Token Consumption

Tech giants like Microsoft and Uber are facing the need to optimize artificial intelligence spending. Recent statements from Uber’s top executives revealed that the company exhausted its annual AI budget in just a few months, with no direct correlation between token usage volume and improvements in consumer-facing features.

Microsoft, meanwhile, has started restricting developer access to the Claude Code tool, migrating them to the internal Copilot CLI. This move coincided with the end of the company’s fiscal year, potentially signaling an effort to reduce costs. Analysts at Goldman Sachs forecast that the adoption of agent-based AI systems will drive a 24-fold increase in token consumption, further exacerbating cost concerns.

The issue is compounded by the fact that many companies do not see a clear link between AI code volume and real business advantages. For instance, Uber notes that while AI-generated code now accounts for 60% of its output, this does not always translate into noticeable product improvements. Similar trends are observed at Airbnb and Google, where AI code adoption is high, but cost efficiency remains questionable.

Experts warn that even significant advancements in hardware, such as Nvidia’s new chips, will not quickly offset the rising costs of tokens. Many companies are already delaying or canceling infrastructure upgrades, which could slow AI project development in the coming years.