The first major casualty of the success of Chinese artificial intelligence (AI) start-up DeepSeek has turned out to be Nvidia, whose stock slumped 17 per cent on Monday amid fears that innovations from the Hangzhou-based firm could reduce the industry's reliance on the US firm's advanced chips. While Nvidia rebounded in pre-trading on Tuesday, analysts noted a shift in the perception of the company's role in costly AI development driven by graphics processing units (GPU), threatening one of the world's most valuable technology titans.
What has DeepSeek achieved?
DeepSeek claims to have pre-trained its V3 model on only 2,048 Nvidia H800 GPUs over a two-month period, with each chip costing about US$2 per hour to run. The total training cost for the V3 model was US$5.5 million, with 2.8 million GPU hours, far less than rival models. Meanwhile, its open-source reasoning model, R1, released earlier this month, has demonstrated capabilities comparable to those of more advanced models from OpenAI, Anthropic, and Google, but with significantly lower training costs.
Does DeepSeek prove Nvidia chips are not indispensable?
Not yet. In a 2023 interview with Chinese media outlet Latepost, DeepSeek founder Liang Wenfeng said the company had gradually built up a stockpile of more than 10,000 Nvidia GPUs, making it one of the top owners of computing resources among Chinese AI start-ups. In an interview in July 2024, Liang said that the company's main problem was not money, but access to advanced US chips that are restricted from export to China, highlighting the importance of the hardware.
While DeepSeek models have delivered impressive results, the company's lack of access to leading AI accelerators, such as Nvidia's best Hopper and Blackwell products, could challenge its long-term ability to keep up in large-language model performance with US peers.
Nvidia on Monday responded to the hype generated by the Chinese AI firm, saying its advances show the usefulness of its GPUs for the Chinese market and that more of its chips would be needed in the future to meet demand for DeepSeek's services.
Is DeepSeek using local alternatives to train its models?
One of the most guarded areas of DeepSeek's AI breakthrough is whether it is using any China-made semiconductors in its training. There have been signs that part of its hardware set-up includes Huawei Technologies' Ascend AI chips, a top alternative to Nvidia chips in China. What are the signs?
Can American AI firms learn from DeepSeek to cut costs?
If US Big Tech companies started to learn from DeepSeek and opt for cheaper AI solutions, it could put pressure on Nvidia. Wei Sun, principal analyst for AI at Counterpoint Research, said the Nvidia sell-off reflects shifting perceptions in AI development.
"DeepSeek's success challenges the belief that larger models and more computing power drive better performance, posing a threat to Nvidia's GPU-driven growth strategy," she said.
