Chinese AI Firm DeepSeek Reportedly Has 50,000 NVIDIA H100 GPUs, Challenging US AI Dominance

Chinese AI firm DeepSeek has 50,000 NVIDIA H100 AI GPUs says CEO, even with US restrictions

TL;DR: DeepSeek, a Chinese AI lab, utilizes tens of thousands of NVIDIA H100 AI GPUs, positioning its R1 model as a top competitor against leading AI models like OpenAI's o1 and Meta's Llama. Despite U.S. export restrictions, DeepSeek reportedly has significant access to advanced chips. The lab focuses on software-driven optimization and open-source collaboration to advance AI technology. Chinese AI lab DeepSeek has access to tens of thousands of NVIDIA H100 AI GPUs for training, according to DeepSeek CEO. The DeepSeek R1 is one of the most advanced AI models on the planet, competing with the likes of OpenAI's new o1 and Meta's Llama AI models. In a new interview with CNBC, Scale AI founder and CEO Alexander Wang said DeepSeek R1 has met, or beats all-top performing AI models in his firm's most challenging AI test. CNBC's Andrew Ross Sorkin interviewed Wang, talking about DeepSeek's new AI test called "Humanity's Last Exam" with the "hardest questions" pumped into it by "math, physics, biology, chemistry professors" that are relevant to the latest research. After testing all of the latest AI models, Wang's team found that DeepSeek's new model was "actually the top performing, or roughly on par with the best American models, which are o1". Wang was asked about the AI competition between the US and China, where he added: "it has been true for a long time that the United States has been ahead". But he did note that DeepSeek's new models do attempt to change that, where he thinks that "is symbolic that the Chinese lab releases, you know, an Earth-shattering model on Christmas Day when you know the rest of us are sort of celebrating a holiday". DeepSeek is using NVIDIA's Hopper AI architecture with H100 and H200 AI GPUs training their AI models, with restrictions put in place by the Biden administration to stop powerful AI GPUs from hitting the shores of China. But it doesn't seem all that hard to get advanced AI chips into China, with Wang telling CNBC: "the reality is yes and no. You know the Chinese labs, they have more H100s than, than people think". Wang said that his "understanding is that DeepSeek has about fifty thousand H100s" and that "they can't talk about obviously because it is against the export controls that United States has put in place", adding "they have more chips than other people expect". In the future, getting access to advanced AI chips to China, Wang said: "But also on a go-forward basis they are going to be limited by the chip controls and the export controls that we have in place".

Chinese AI Lab DeepSeek Has 50,000 NVIDIA H100 AI GPUs, Says AI CEO

This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy. Chinese artificial intelligence lab DeepSeek, whose artificial intelligence model R1 took the industry by storm this month, has access to tens of thousands of NVIDIA's GPUs for training, believes the CEO of an AI company. DeepSeek R1 is one of the most advanced AI models in the world, and it stands shoulder to shoulder with leading American platforms such as OpenAI's o1 and Meta's Llama. NVIDIA's Hopper chips are its current generation GPUs, which are the most widely used in the world as the firm ramps up shipments of the next-generation Blackwell chips. The latest details were shared by Scale AI's founder and CEO Alexandr Wang in an interview given to CNBC earlier this week, with Wang adding that R1 has met or beat all top-performing AI models in his firm's most challenging AI test. Wang started his interview, given to Andrew Ross Sorkin of CNBC, by sharing details of his firm's latest AI test. Called "Humanity's Last Exam," the test uses the "hardest questions" provided by "math, physics, biology, chemistry professors" that are relevant to the latest research. After testing all the latest AI models with this test, Wang's team discovered that DeepSeek's latest model was "actually the top performing, or roughly on par with the best American models, which are o1. . .." When asked about the AI competition between the US and China, Wang commented that "it has been true for a long time that the United States has been ahead." However, he added that DeepSeek's latest models do attempt to change the calculus. According to Wang, he thinks it "is symbolic that the Chinese lab releases, you know, an Earth-shattering model on Christmas Day when you know the rest of us are sort of celebrating a holiday." The conversation then turned to the amount of GPUs that DeepSeek and Chinese AI players might have for their AI models. NVIDIA's GPUs are among the most highly sought products in the world, and due to their potential, their exports are also regulated by the US government. The firm's current-generation Hopper GPUs lineup consists of the H100 and H200 GPUs. The primary difference between the two GPU models is their memory clock speed and capacity. The Biden administration had restricted the sale of these GPUs to China in 2022 through a rule that prevented NVIDIA from selling chips vastly superior to the older A100 GPUs. These restrictions forced NVIDIA to develop alternative chips called H800 and A800, and these were also banned a year later in October 2023. According to Wang, when it comes to the Chinese accessing NVIDIA's advanced GPUs, "the reality is yes and no. You know the Chinese labs, they have more H100s than, than people think." He added and shared that his "understanding is that DeepSeek has about fifty thousand H100s." Wang outlined, "they can't talk about obviously because it is against the export controls that United States has put in place." He also thinks that "they have more chips than other people expect." However, the Chinese might find it difficult to procure additional chips. According to Wang, "But also on a go-forward basis they are going to be limited by the chip controls and the export controls that we have in place."