Chinese AI startup DeepSeek recently announced that its AI model could be very profitable – with some asterisks.
exist Posts on xDeepSeek boasts about a “cost margin” of 545% for its online service. However, this profit is calculated based on “theoretical income.”
It discusses the numbers in more detail at the end Longer github posts Overview of its approach to achieving “higher throughput and lower latency”. When it checks out its V3 and R1 models in 24 hours, the company wrote. if With R1 pricing, DeepSeek will earn $562,027 per day.
Meanwhile, the cost of renting the necessary GPU (graphics processing unit) is only $87,072.
The company acknowledged that its actual revenue was “significantly lower” for various reasons such as nighttime discounts, lower pricing for V3s, and “only a portion of the services are monetized” and that web and application access remains free.
Of course, if the apps and websites are not free and if other discounts are not available, the usage may be much lower. So, these calculations seem highly speculative – a step towards potential future profit margins than a true snapshot of the current DeepSeek bottom line.
But the company is sharing these numbers in a broader debate on AI costs and potential profitability. DeepSeek Jumps into the Spotlight In January, despite much lower costs and facing U.S. trade restrictions, the new model is said to match some benchmark Openai’s O1, which prevents Chinese companies from getting the most powerful chips. Technology stocks fell Analysts raise questions about AI spending.
DeepSeek’s technology is not just a rattle of Wall Street. Its application Short-term displacement At the top of Apple’s App Store – although it subsequently departed from the general rankings and is currently ranked sixth in productivity, behind Chatgpt, Grok and Google Gemini.
