Many people are losing their jobs to AI. However, numbers now show that the use of AI is not necessarily cheaper.
In many companies, there is now an emphasis on having employees utilize AI tools. This sometimes leads to entire teams being laid off if they refuse to use AI.
However, this pressure for increased AI usage is becoming more expensive for large companies like Microsoft. Each individual query costs money, and the more queries are made, the more expensive it becomes.
If employees rely heavily on AI, it can ultimately be more expensive than one employee
Why is that? Basically, it relates to how AI is used and how the costs are ultimately calculated:
- AI cannot read human text, so it converts the text into
tokens
to process queries. - The longer the text, the more tokens are used.
- Each task or request costs a certain amount of tokens, and each token costs money. The more tokens are consumed, the higher the costs for the company.
- Especially agentic AIs, such as chatbots like ChatGPT, consume a lot of tokens at once.
Depending on the workflow within a company, AI costs can quickly rise: many users make many requests, use particularly expensive models, or let complex issues be processed through ChatGPT. In the end, the ongoing AI costs can be higher than expected.
If many people in a company use AI very intensively, the total bill can be higher than the salary costs for the same tasks. Even if individual AI queries become cheaper, as providers like OpenAI or Anthropic reduce the costs of tokens.
And this is exactly what companies like Microsoft have noticed: internally, it appears they have found that AI usage in some cases becomes more expensive than expected, primarily due to high token and computing costs. AI is therefore not automatically cheaper than humans — in intensive use, it can even become financially less attractive. For this reason, Microsoft has likely discontinued the Claude subscription for its employees. This was reported by the magazine TheVerge.
Moreover, the magazine Fortune.com reports that such considerations align with the ideas of other companies. Bryan Catanzaro, Vice President of Applied Deep Learning at Nvidia, recently said in an interview with Axios: “For my team, the computing costs far exceed the personnel costs.”
By the way, the situation is not entirely new: a study already found in 2024 that the use of AI is not necessarily cheaper. This was reported among others by the magazine Euronews.de. The researchers came to the conclusion that only in 23% of the cases studied was the work performed by AI cheaper than a salary paid.
IBM, a large tech company, laid off many employees in 2023 due to artificial intelligence. The surprising thing, however, is that later the company hired back a similar number of people due to AI. AI therefore not only destroys jobs but can also create them, although not the same employees benefit from the new positions: A company laid off 7,800 employees to replace them with AI, but did not expect that they would hire several employees back due to AI
Your opinion is important to us!
Do you like the article? Then let us know!