The Chinese beginning of Ai Deepseek recently stated that his models of he could be very lucrative – with some asterisks.
In a post on X, Deepseek boasted that his online services have a 545%”cost profit margin”. However, that difference is calculated based on the “theoretical income”.
He discussed these numbers in more detail at the end of a longer Github post describing his approach to reach “higher turnover and lower delay”. The company wrote that when you look at the use of its V3 and R1 models during a 24-hour period, whether This use was all billed using R1 prices, Deepseek would already have $ 562,027 in daily income.
Meanwhile, the cost of renting the necessary GPU (graphic processing units) would have been only $ 87,072.
The company acknowledged that its current income is “fundamentally lower” for a variety of reasons, such as night discounts, lower prices for V3, and the fact that “only a subset of services are money”, with web and applications left for free.
Of course, if the application and website were not free, and if other deductions were not available, the use would apparently be much lower. So these calculations seem to be very speculative – more of a gesture towards the potential boundaries of profit in the future than a true photograph of the Deepseek end line now.
But the company is dividing these numbers between the broader debates about the cost and the possible benefit. Deepseek was in the spotlight in January, with a new model allegedly consistent with O1 O1 to some standards, despite developing at a much lower cost, and in the face of US trade restrictions that prevent Chinese companies from entering the most powerful chips. Technology shares crashed and analysts raised questions about it.
Deepseek’s technology does not just make Wall Street’s shock. Its app briefly shifted Openai’s Chatgt at the top of Apple’s App Store – though it is then dropped from the overall ranking and is currently ranked to #6 in productivity, after chatgpt, Grok and Google Gemini.
