BEIJING: Chinese AI developer DeepSeek disclosed that it spent $294,000 to train its R1 model, far lower than costs reported by U.S. rivals. The announcement comes amid global debate over China’s role in AI development.
The Hangzhou-based company revealed the figure in a peer-reviewed paper published Wednesday in the journal Nature. This is the first official estimate from DeepSeek regarding R1’s training costs. The article lists company founder Liang Wenfeng as a co-author.
DeepSeek’s low-cost AI systems, first reported in January, had previously stirred global investor concerns. Analysts worried that these models could challenge the dominance of AI leaders such as Nvidia and OpenAI. Since then, DeepSeek and its founder have mostly stayed out of public view, aside from occasional product updates.
The paper explained that the reasoning-focused R1 model used 512 Nvidia H800 chips and required 80 hours of training. DeepSeek also acknowledged using Nvidia A100 chips in early preparatory stages, giving it access to powerful computing resources for model development.
U.S. AI firms have questioned some of DeepSeek’s statements about its development costs and technology. The H800 chips were specifically designed for the Chinese market after U.S. export controls banned Nvidia’s more advanced H100 and A100 chips from being shipped to China. DeepSeek maintains it acquired and used these chips legally.
Training large-language models is typically expensive, requiring powerful chip clusters running for weeks or months to process extensive text and code. By comparison, OpenAI CEO Sam Altman stated in 2023 that foundational model training costs “much more” than $100 million.
Experts say DeepSeek’s ability to operate an A100 cluster helped attract top talent in China. Analysts also note that the company’s low training costs may reshape global discussions about AI competitiveness, highlighting China’s growing capabilities in the sector.
DeepSeek’s disclosure comes at a critical time as nations assess AI development leadership and investment in artificial intelligence technology.
In other news read more about: Baidu Unveils Low-Cost AI Models Amid Intensifying Competition in China
