AI Data Center Electricity Demand May Be Wrong Because Of DeepSeek

Andrew Neel Pexels

Just a few weeks ago,  the largest AI development and product companies led by Microsoft, OpenAI, Google, Facebook and Amazon assumed that the sum required to build huge AI server centers would rise into the hundreds of billions of dollars. The new China AI product DeepSeek may make those estimates much too high.

*AI Has Winners And Losers

*How Does AI Change Global Warming

*AI And Tesla May Be Industry Leaders

DeepSeek was developed for under $10 million, its owner says. It is hard to confirm that number, but so far most experts say it was in that neighborhood. For OpenAI to create an equivalent product cost well over $100 million. Additionally, the OpenAI work needs expensive, electricity hungry chips, most of which are made by Nvidia. The DeepSeek software was built with much cheaper chips which use significantly less power. 

If US AI companies can adopt some of the development approaches used by the DeepSeek creators, the costs of data centers could collapse below current forecasts. This would save the industry a huge amount of cost and capex. It could badly damage Nvidia, by eliminating costs for some of the largest tech companies in the world.

Major Expenditures Could Drop

Among the other “losers” if the electricity needed by AI data centers drops will be utilities which have entered the business of providing this energy. One of the utilities which had already made commitments to these data centers was Constellation Energy. Its stock collapsed on the DeepSeek news.

DeepSeek may have completely changed the energy needs of AI data centers, and with it the energy needs for the US.

More from ClimateCrisis 247

Similar Posts