US tech stocks tumble as DeepSeek's AI breakthroughs challenge America's edge in advanced computing
Nvidia and other US technology firms experienced a sharp decline on Monday, part of a global sell-off triggered by advancements from Chinese startup DeepSeek.
CNBC reported that Nvidia, a major player in the AI sector, saw its shares fall 16.9 percent, marking the steepest drop since March 2020.
Nvidia’s stock reached lows not seen since October, contributing to broader declines in AI-focused companies and US markets.
Other US-based chipmakers also suffered losses. Micron and Arm Holdings dropped more than 11 percent and 10 percent, respectively, while Broadcom fell over 17 percent, and Advanced Micro Devices slid by more than 6 percent.
Companies tied to AI power infrastructure, such as Constellation Energy and Vistra, saw even steeper declines, with their shares plunging over 20 percent and 28 percent, respectively.
The sell-off extended to international markets.
In Europe, Netherlands-based chip companies ASML and ASM International reported significant losses. In Asia, Japanese chip-related stocks, including Advantest and Tokyo Electron, also declined sharply.
DeepSeek recently unveiled a free open-source large language model, claiming it was developed in just two months for under US$6m, a cost far lower than those of comparable US projects.
Last week, the company launched a reasoning model that reportedly outperformed OpenAI’s latest version in third-party tests.
“DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling,” said an Nvidia spokesperson. They noted that DeepSeek’s models were developed using widely available technology and remain compliant with export controls.
“Inference requires significant numbers of NVIDIA GPUs and high-performance networking,” they added, highlighting the essential role of Nvidia’s technology in AI development.
Marc Andreessen, co-founder of Andreessen Horowitz, praised DeepSeek’s model as “one of the most amazing and impressive breakthroughs I’ve ever seen” in a social media post.
He described it as a “profound gift to the world.” Andreessen, who has publicly supported US President Donald Trump, drew significant attention with his comments.
The rapid advancements by DeepSeek have raised questions about the heavy investments by US technology firms in AI models and data centres.
Some experts are concerned that the US is no longer as dominant in the sector as previously believed.
“DeepSeek clearly doesn’t have access to as much compute as US hyperscalers and somehow managed to develop a model that appears highly competitive,” said Srini Pajjuri, semiconductor analyst at Raymond James.
Pajjuri suggested that this development could push US hyperscalers like Amazon and Microsoft to act quickly and further capitalise on their access to advanced GPUs.
GPUs are critical for training large-scale AI models. Nvidia remains the market leader in this area. Citi analysts pointed out that while DeepSeek’s innovations highlight the growing competition, access to advanced chips remains a significant advantage for US-based firms.
They also noted the announcement of the US$500bn Stargate AI project as a reflection of the ongoing demand for cutting-edge technology in the AI industry.
However, Bernstein analysts questioned whether DeepSeek’s reported US$6m development cost included all related expenses, such as prior research and experimentation.
While they acknowledged that the company’s models are impressive, they downplayed concerns about the long-term impact on the US AI sector, stating that fears of a “death-knell of the AI infrastructure complex” are overblown.
