China’s AI Startup DeepSeek recently declared that the AI model is very profitable with some asterisk.
In X’s post, Deepseek boasted 545%of the “cost profit margin” of the online service. But this margin is calculated according to the “theoretical income”.
This number briefly explained the approach to achieve “higher throughput and lower standby time” at the end of the longer GitHub post. When the company sees the use of the V3 and R1 models for 24 hours, If The usage was all charged using the R1 price set, and DeepSeek has already received $ 562,027 every day.
On the other hand, the cost of renting the necessary GPU (graphics processing device) would have been only $ 87,072.
The company admitted that the actual profits are “actually low” for various reasons, such as night discounts and low prices. He admitted that he could access the web and apps that “only sub -sets of services generate profits.”
Of course, if the app and website are not free and you can’t use other discounts, the usage will be much lower. Therefore, this calculation seems to be very speculative. The current profitability of DeepSeek is a gesture for the future profit margin than the actual snapshot.
However, the company shares these numbers in a wider debate on the cost of AI and potential profitability. DEEPSEEK is attracting attention in January, and despite being developed at a much lower cost, with a new model that faces the US trade restrictions that match the US trade restrictions that match the O1 of O1 and specific benchmarks. Technology stocks have become valuable and analysts have questioned AI spending.
DeepSeek’s technology did not just hold Wall Street. The app briefly replaced Openai’s CHATGPT at the top of Apple’s App Store. But after that, he was out of general ranking and now ranked 6th in productivity behind Chatgpt, Grok and Google Gemini.
