Investors question whether Big Tech AI spending will pay off

In short: Big Tech is pouring huge sums into AI data centers and chips, and more people are asking whether the payoff will match the cost.

What's going on

Big technology companies are spending hundreds of billions of dollars on the computers, chips, and buildings needed to run today’s popular AI, especially large language models, or LLMs (the kind of AI that writes and chats). Some estimates suggest AI-related spending could reach around $700 billion in 2026 alone.

The worry is not that AI has no use. The worry is that this specific approach, very large spending focused on LLMs, may be a costly first phase with lower returns than expected. Critics point to limits in LLMs, like giving different answers to the same question, and sometimes making things up, which can make them hard to trust for high-stakes work.

Another concern is the gap between spending and confirmed revenue. Some commentary says generative AI revenue at major companies is still in the low single-digit billions, while infrastructure spending is far larger. A separate, high-profile signal came from investor Michael Burry, who disclosed large “put options” (a bet that a stock price will fall) tied to companies seen as big AI winners.

There is also debate about accounting. Some analysts say extending how long companies claim chips and data center gear will last reduces reported costs today, like spreading a car’s price over more years to make each year look cheaper.

What to watch

Watch whether AI revenue grows fast enough to justify the buildout, and whether companies later report higher costs as older hardware is replaced. Also watch whether businesses that cut jobs due to AI actually improve results, or end up rehiring.

Source: NYTimes