A Chinese software developer DeepSeek has created and released an Artificial Intelligence (A.I.) “Large Language Model” that is cheaper and more efficient than U.S. models. The DeepSeek program uses less computational capacity and less electricity than the best performing U.S. designs. In addition, all this is accomplished with older models of U.S. produced computer chips. 

One immediate negative impact of the announcement has been a drop in stock prices for U.S. chip producers and A.I. firms. This Monday U.S. stock markets dropped prices for tech stocks on this news. It hit chip maker Nvidia with a 17% price drop representing a $600 billion loss in market value in one day. It’s estimated Nvidia controls 90% of chip sales for A.I. The drop in value on the news of a more efficient competitor may be a reflection that Nvidia’s recent growth may not be limitless. Alternatively, it may also have been timely selling to capture gains in Nvidia’s recent 900% stock price run up since 2023. 

A few analysts have postulated that duplication and proliferation of these Chinese designed programs in the U.S. may lower the previously expected growth in electricity demand for the numerous announced new data centers to be built across the U.S. Or it may not. The use of lower cost chips may decrease the cost of data centers, however, will lower costs also stimulate demand? 

The lower cost may actually bring new investors and developers into the data center market increasing the number of proposed data centers. This view was offered by Bernstein Research analyst Stacy Rasgon as reported by the New York Times “He added that being able to build more A.I. systems efficiently should mean more demand because more companies can afford to invest in them.”  That also seems to be the view of Chevron Corp. which announced Tuesday it would make major investments in natural gas fired power plants to collocate with A.I. data centers. Jeff Gustafson, president of Chevron new Energies was quoted in Reuters (1/28/25) saying “We still see growth in electricity demand.” 

Another scenario may be that the same principles of programming efficiency applied by the Chinese scientists and programmers to the older and cheaper chips could also be applied, this time by American programmers, to Nvidia’s newer higher-powered chips.  

That the Chinese were able to create superior A.I. performance with older chips was a significant development brought about by competition. Its announcement was unexpected, and the new development apparently caught the financial and data center markets by surprise. Once the stock market and new A.I. industry have fully evaluated the impacts and consequences of DeepSeek’s the conclusions could be different than initially impressions. Chevron may be right in that demand for A.I. Services and the ensuing demand for electricity could still rise significantly. Stay tuned.