Major tech companies have begun re-evaluating their spending on artificial intelligence following a sharp rise in the costs of maintaining neural networks.
This involves not only purchasing hardware but also payments for computing power, tokens, and employee licenses. Against this backdrop, some companies are already scaling back their use of third-party AI services, El.kz reports .
According to industry media reports, Microsoft has begun limiting the internal use of Claude Code and transitioning employees to its own solution, GitHub Copilot. The reason cited is the high maintenance costs of AI tools.
Costs rising faster than expected
Uber reported that its corporate AI budget was exhausted within just a few months. Management admitted that they do not yet see a direct correlation between the rising costs of neural networks and the emergence of new useful features for customers. At the same time, the use of AI within the company has indeed accelerated some internal workflows.
Some engineers at major firms spend thousands of dollars monthly on working with language models. The main expense item is tokens, which are consumed during code generation, data analysis, and the operation of AI agents. Complex automated tasks, where the volume of computation can increase tenfold, have proven to be particularly expensive.
Companies begin rethinking value in AI economy
NVIDIA stated that some teams are already spending more on computing power than on employee salaries.
Amid growing expenditures, companies are increasingly discussing whether the mass integration of generative AI is justified. Researchers note that increasing token consumption does not always improve model performance, and the cost of performing a single task can vary significantly even with identical prompts.
Despite this, global investment in artificial intelligence continues to grow. According to estimates from analysts and industry representatives, total spending on AI and IT infrastructure is already approaching several trillion dollars a year. Major tech companies continue to invest in data centers, servers, and the development of new models.
