
19/09/2025
Chinese AI giant DeepSeek has revealed that training its reasoning-focused R1 model cost just $294,000 (about Rs 2.6 crore), a fraction of what US rivals like ChatGPT maker OpenAI reportedly spent.
The Hangzhou-based company said the R1 model was trained using 512 Nvidia H800 chips. The update marks the first public estimate of DeepSeek’s training costs, which are dramatically lower than figures cited by OpenAI.
CEO Sam Altman has previously said training foundational models cost his company “much more” than $100 million, though OpenAI has not disclosed exact numbers.
DeepSeek made headlines in January when it introduced low-cost AI systems that triggered a selloff in global tech stocks, as investors feared they could disrupt the dominance of companies like Nvidia.
Since then, the firm and its founder, Liang Wenfeng, have kept a low profile, aside from occasional product updates.