The release on Jan. 24 of a “reasoning” version of the Chinese large language (AI) model called DeepSeek, hit the U.S. “semiconductor champion” company Nvidia so hard that its stock dropped by 18% in one day on Monday, Jan. 27, and pulled down the Nasdaq stock exchange, all cryptocurrency stocks and funds, power companies, and more. By the end of the day, the company behind DeepSeek was experiencing cyberattacks hoping to stop the rout.
It seems that the (at least) competitive computing qualities of DeepSeek, using fewer and less advanced semiconductors and far, far less expenditure, may have indicated that a nation doesn’t need tens of billions in “chip fab” subsidies, thousands of new data centers, and repeated rounds of attempted trade embargoes against adversaries, to master supercomputing.
