DeepSeek claims ‘theoretical’ profit margins of 545%
The Chinese AI Startup Deepseek recently stated that his AI models can be very profitable -with some asterisks.
In A message on XFochte Deepseek had that its online services have a “cost gain margin” of 545%. However, that margin is calculated on the basis of ‘theoretical income’.
It discussed these figures at the end of more details A longer github -post The approach to achieving “higher transit and lower latency”. The company wrote that when it is used to use its V3 and R1 models for a period of 24 hours, when That use was all invoiced with the help of R1 prices, Deepseek would already have $ 562,027 in daily income.
In the meantime, the costs of leasing the required GPUs (graphic processing units) would have only been $ 87,072.
The company admitted that its actual income is “considerably lower” for various reasons, such as nocturnal discounts, lower prices for V3, and the fact that “only a subset of services is monitored”, whereby the web and app access remains free.
Of course, if the app and the website were not free, and if other discounts were not available, the use would probably be much lower. So these calculations seem to be very speculative – more a gesture to potential future profit margins than a real snapshot of Deepseek’s Bottom Line at the moment.
But the company shares these figures in the midst of broader debates about the costs and potential profitability of AI. Deepseek jumped in the spotlight in January, with a new model that supposedly corresponded the O1 from OpenAi to certain benchmarks, even though they were developed at much lower costs, and in the light of American commercial restrictions that prevent Chinese companies from having access to the most powerful chips. Tech stocks tumbled and analysts praised about AI expenditure.
Deepseek’s technology not only rattled Wall Street. The app briefly moved the Chatgpt of OpenAi at the top of Apple’s App Store – although it subsequently fell from the general ranking and currently #6 is ranked in productivity, behind Chatgpt, Grok and Google Gemini.
