In Reply to: Will data centers kill factories and jobs? posted by mh on July 11, 2026 at 14:26:55
The AI data center boom may be coming to an end.
There are reports suggesting that xAI could slow or halt additional data center construction and instead rent out excess backend computing capacity after potentially overbuilding its infrastructure.
Meta's stock rose about 9% in a single day after the company discussed its AI infrastructure strategy of focusing on renting out backend computing. Some investors interpreted its comments as an indication that the company may have built more computing capacity than it currently needs and has realized that it can better monetize that computing by renting it because of a lack of enough front end applications.
There has also been speculation about OpenAI's financial position. Some analysts have suggested that, at its current rate of spending, the company may have a limited cash runway without additional funding. Reports have also indicated that OpenAI has narrowed its focus by scaling back some side projects to concentrate on its core AI products. It's now begging Trump for a 5% bailout.
If these trends continue, the market for large-scale AI computing could become increasingly concentrated. At present, OpenAI and Anthropic appear to be among the largest external buyers of AI computing capacity with OpenAI slowly dying.
The winners and losers of the AI race are starting to separate themselves and the losers have too many data centers.
Another trend worth watching is how businesses optimize AI costs. Some companies are reportedly experimenting with routing simpler AI queries to lower-cost Chinese models while reserving more complex or sensitive tasks for U.S.-based models. Supporters of this approach argue that Chinese AI providers can offer lower prices because of lower operating costs, cheaper energy, and, according to some critics, the use of intellectual property that was not independently developed. Whether this cost advantage proves sustainable remains to be seen.
At this moment we have a surplus of US AI backend computing with a lack of application driven demand.