DeepSeek's AI Training Costs Ignite Global Anticipation
Chinese AI developer DeepSeek spent $294,000 on training its R1 model, stirring debates over China's AI development capabilities. The training used Nvidia H800 chips amid U.S. export restrictions. Despite criticism over transparency, DeepSeek remains significant in China's AI landscape.
Chinese AI developer DeepSeek has reported training costs for its R1 model at a mere $294,000, far below what many U.S. competitors spend. This disclosure is likely to fuel discussions on Beijing's position in the competitive AI development landscape.
The Hangzhou-based company, in its first publicly shared estimate of R1's training expenses, released this data through a paper in the esteemed journal Nature. Earlier announcements of their cost-effective AI systems in January had previously led investors to divest from tech stocks due to fears of market disruption.
In a world where giants like OpenAI's CEO Sam Altman have mentioned that foundational model training exceeds $100 million, DeepSeek's move highlights a significant cost differentiation, especially using Nvidia's China-specific H800 chips due to U.S. export limitations.
- READ MORE ON:
- DeepSeek
- AI
- training
- costs
- R1 model
- Nvidia chips
- U.S. export
- Beijing
- China
- artificial intelligence
Google News