DeepSeek has triggered a dramatic rethink on artificial intelligence spending around the world, except perhaps in China. The startup's impressive low-cost model showcases the country's innovation prowess. Yet it also spotlights how U.S. curbs on high-end chips hold back the world's second-largest economy in the global technology race. For Beijing, the smartest way forward is to keep pouring money into breaking the West's chokehold on advanced semiconductors.
The Hangzhou-based upstart stunned the world with its latest R1 model that is just as powerful as offerings from the likes of OpenAI and Meta Platforms (META.O) but is cheaper to run. As investors digested the implications of inexpensive AI, and wiped out nearly $600 billion in market value from chipmaker Nvidia (NVDA.O) alone, U.S. officials questioned the efficacy of tech sanctions unleashed by Washington to slow China's progress in the industry and called on Donald Trump's new administration to tighten export curbs.
How the Republican president, who on Monday called DeepSeek's model a "wake-up call" for his country, responds will now be key. On one hand, DeepSeek's breakthrough suggest sanctions accelerated creativity in the People's Republic. Indeed, Chinese firms from Huawei to Alibaba (9988.HK), have devised clever workarounds; tech analyst Ben Thompson of Stratechery argues that "all of the decisions DeepSeek made in the design of this model only make sense" because it could not use Nvidia's top-of-the-line offering.
Even so, the chips DeepSeek used have been banned in China since 2023. That means compatriots hoping to replicate its success face even tougher constraints today. In a rare interview last year, the company's founder Liang Wenfeng said that while money has never been a problem for the startup, bans on advanced chips are. In short, China has the talent and resources, but the biggest constraint is in its access to high-end processors.
It is here that Beijing is funnelling billions of dollars to help homegrown champions like Huawei and the $55 billion Semiconductor Manufacturing International (0981.HK) build domestic alternatives. The pair have racked up some important wins, including mass-producing a 7 nanometre smartphone chip back in 2023 that many thought was out of their reach. DeepSeek shows there are no shortcuts.
Follow @mak_robyn on X
CONTEXT NEWS
OpenAI CEO Sam Altman called Chinese startup DeepSeek's R1 artificial intelligence model "impressive" on Jan. 27, but emphasized that OpenAI believes greater computing power was key to their own success.
"DeepSeek's R1 is an impressive model, particularly around what they're able to deliver for the price," Altman said on X. "But mostly we are excited to continue to execute on our research roadmap and believe more compute is more important now than ever before to succeed at our mission," Altman added.
DeepSeek-R1, launched last week, is 20 to 50 times more affordable to use than OpenAI's o1 model, depending on the task, according to a post on DeepSeek's official WeChat account.
