Chinese AI startup DeepSeek has disclosed that its V3 and R1 models could achieve a theoretical daily profit margin of up to 545%, assuming all users switch to paid plans. Currently, many of DeepSeek’s models, including V3 and R1, run largely for free, incurring a daily cost of $87,072 for Nvidia chips. Daily revenue, if billed at R1’s pricing, would be $562,027, projecting annual revenue over $200 million. However, these figures assume universal paid usage, which is currently unachieved by competitors.
DeepSeek’s low-cost and high-profit AI model could herald the arrival of a more efficient, cost-effective, and highly competitive AI era, challenging some of the business models of foreign rivals like OpenAI.
Despite these promising figures, DeepSeek’s actual revenue is substantially lower due to factors such as offering free web and app access, lower pricing for the V3 model, and reduced developer fees during off-peak hours.
DeepSeek’s rise in the AI industry has been disruptive, with the company first turning heads earlier this year when it revealed that it had spent less than $6 million on chips used to train its models. This is a stark contrast to the billions of dollars that U.S. rivals like OpenAI have invested in cutting-edge hardware. Moreover, DeepSeek relies on Nvidia’s H800 chips, which are significantly less powerful than the hardware deployed by American AI firms.
However, DeepSeek faces a significant barrier to international expansion—censorship. Unlike Western AI models, which are often built on open data sets and trained with a focus on free expression, DeepSeek’s models are required to adhere to strict Chinese censorship laws. For instance, its chatbots and AI tools are programmed to filter out politically sensitive topics, making the models less attractive to international developers and global enterprises that value unrestricted access to information.
