Nowadays there is discussion everywhere that Artificial Intelligence (AI) will replace humans and save a lot of money for companies. But, the reality seems to be completely opposite. The big companies of the technology world, which were the first to adopt AI, are now surprised to see its huge bills. Data from big companies like Microsoft, Uber and Nvidia show that in many cases the expense of AI has exceeded the salaries of their human employees.
Microsoft pulled back its steps
Just six months ago, Microsoft gave thousands of its developers, project managers and designers access to an AI tool called ‘Cloud Code’. The company wanted its employees to use it extensively for coding. The employees embraced it and started using it extensively. According to the report of The Verge, this use has increased so much that now Microsoft is having to withdraw access to this tool from its own engineers. The company is now canceling most cloud code licenses and moving employees to GitHub Copilot CLI. However, Microsoft’s business agreement with Anthropic Company, which includes an investment of $5 billion, will continue as before.
Uber’s entire budget spent in just a few months
This problem is not limited to Microsoft only. Praveen Nepalli Naga, Chief Technology Officer (CTO) of ride-hailing company Uber, has recently revealed that the company’s entire AI coding budget fixed for the year 2026 was exhausted in just four months. This situation came when Uber itself inspired its employees to use AI tools. Leaderboards were created to compete among teams to use more AI.
After all, why is the bill getting so expensive?
The root of this huge expense lies in the pricing policy of AI. Large Language Models (LLM) charge money per ‘token’. Token is actually the smallest unit of text that AI processes and generates. In such a situation, if the work is being done faster or more, then in both the cases the consumption of tokens increases and there is a direct increase in the total expenditure. Companies like Amazon encouraged their employees to use more tokens. Whereas in Meta, it was being tracked which employee was using AI the most.
Goldman Sachs estimates that by 2030, token consumption will increase 24 times with the advent of autonomous AI systems. Gartner’s research shows that token prices may fall by 90% in the future, but the number of tokens used will increase so much that the total bill of companies will not reduce under any circumstances.
Computer becomes more expensive than human salary
The most shocking statement in this entire matter has come from Brian Catanzaro, Vice President of Applied Deep Learning of Nvidia. He has admitted that the cost of computers (compute power) for his team has exceeded the total cost of employees. Nvidia is the same company that makes AI-powered chips around the world.
Leave a Reply