DeepSeek released a free, open-source large language model in late December, claiming it was created in only two months with a budget of under $6 million. Last week, the lab introduced R1, a reasoning model that outperformed OpenAI's latest o1 in several third-party tests.
"There will be a lot more pain to follow today as we follow the U.S. down," said Andrew Jackson, head of equity strategy at ORTUS Advisors.
"The big questions is whether the U.S. will U-turn on their approach and deregulate chip and SPE restrictions seeing as they are (currently) ineffectual or try and ramp things up even more," Jackson wrote in an email. DeepSeek had to navigate strict semiconductor restrictions imposed by the U.S. government on China, which limited access to advanced chips.