Chinese start -up AI Deepseek has recently said that AI models could be very profitable – with some asterisks.
In a post on xDeepseek boast that his online services have a “cost of profit” 545%. However, this margin is calculated on the basis of “theoretical income”.
Discussed these numbers with more details at the end of a larger GitHub post describing his approach to achieving “higher performance and lower latent status”. The company wrote that when considering the use of V3 and R1 models during a 24 -hour period, if This use was invoiced using the R1 pricing, Deepseek would already have $ 562,027 in daily revenue.
Meanwhile, the cost of leasing the necessary GPUS (graphics processing units) would only be $ 87,072.
The company acknowledged that its actual revenue is “essentially lower” for a variety of reasons, such as night discounts, the lower pricing for the V3 and the fact that “only one subset of services is rendered”, with the web and app accessing the balance for free.
Of course, if the application and the site were not free and if no other discounts were available, the use would probably be much lower. Thus, these calculations seem to be extremely speculative – more a gesture towards possible future profit margins from a real -day DEEPSEEK slope at the moment.
But the company shares these numbers in broader discussions about the cost and potential profitability of AI. Deepseek jumped to the forefront in January, with a new model that is supposed to match OPENAI O1 at some benchmarks, despite being developed at a much lower cost and in view of US trade restrictions that prevent Chinese companies from accessing the most powerful brands. Technology stocks collapsed and analysts asked questions about AI expenses.
Deepseek technology is not just ringing Wall Street. Its application has briefly shifted the Openai Chatgpt to the top of the Apple App Store – though it then falls from the general rankings and is currently ranked #6 in productivity, behind Chatgpt, Grok and Google Gemini.
