Chinese AI startup DeepSeek has made a bold claim—its AI models could generate massive profits. The company stated that its online services have a “cost profit margin” of 545%, but there’s a catch. These figures are based on “theoretical income,” not actual earnings.
In a detailed post on GitHub, DeepSeek shared how it aims to improve AI performance with higher speed and lower delays. It calculated that if its AI model usage over 24 hours had been billed at R1 pricing, the revenue would be $562,027, while GPU leasing costs would be just $87,072. However, DeepSeek admitted that its actual revenue is much lower due to discounts, lower pricing for some models, and free services on its app and website.
While the numbers suggest potential for high profit, they don’t reflect the company’s current financial reality. If DeepSeek were to remove free access and discounts, usage could decline significantly.
The company’s claims come amid growing debates about AI costs and profitability. DeepSeek gained attention earlier this year with an AI model that reportedly matched OpenAI’s performance despite being developed at a lower cost. This stirred discussions about AI spending and even impacted tech stocks.
DeepSeek’s AI app briefly overtook ChatGPT in Apple’s App Store rankings but has since dropped, currently sitting at #6 in the productivity category, behind ChatGPT, Grok, and Google Gemini.