DeepSeek claims 545% cost-profit ratio for its AI models
Chinese startup DeepSeek reports 545% profit ratio from V3 and R1 models
Chinese AI startup DeepSeek has disclosed striking cost and revenue data for its V3 and R1 models, revealing a theoretical daily cost-profit ratio of up to 545%.
In a GitHub post published on Saturday, the Hangzhou-based company provided rare insight into its financial performance from the "inference" stage — the phase after model training where AI systems make predictions or execute tasks, like powering chatbots, reports Reuters.
The company estimates its total daily inference cost for both models at $87,072, while their theoretical daily revenue reaches $562,027. Over a year, this would amount to just over $200 million in revenue.
Rattling Global AI Markets
DeepSeek’s disclosures arrive at a time of heightened volatility in global AI stocks. The firm’s earlier claims that it spent less than $6 million on chips to train its models — a fraction of the billions invested by U.S. competitors like OpenAI — have already unsettled markets.
The use of Nvidia’s H800 chips, which are less powerful than those available to U.S. firms, has further fueled doubts about the massive expenditures on cutting-edge hardware in the American AI sector.
This cost-efficiency and rapid popularity of DeepSeek’s models have led some investors to question the sustainability of high spending by Western AI firms. AI stocks outside China saw a notable sell-off in January, driven by growing recognition of DeepSeek’s capacity to achieve competitive performance at a fraction of the cost.
Caution Around Revenue Figures
Despite its impressive theoretical margin, DeepSeek emphasized that actual revenue is lower due to multiple factors. Access to some of its services remains free, only specific functions are monetized, and discounted rates apply during off-peak hours. Moreover, the V3 model, which costs less to operate than the R1, contributes to the lower realized revenue.
As AI competition intensifies globally, DeepSeek’s approach highlights the potential for cost-effective model deployment, raising questions about spending strategies across the industry. The firm’s continued transparency around costs and revenue could further influence market perceptions in the months ahead.
Additionally, U.S. President Donald Trump and Nvidia CEO Jensen Huang met at the White House to discuss the rise of China's AI company, DeepSeek, and the tightening of artificial intelligence (AI) chip exports.
