DeepSeek claims ‘theoretical’ profit margins of 545%

Chinese language AI startup DeepSeek lately declared that its AI fashions might be very worthwhile — with some asterisks.

In a submit on X, DeepSeek boasted that its on-line providers have a “value revenue margin” of 545%. Nonetheless, that margin is calculated primarily based on “theoretical revenue.”

It mentioned these numbers in additional element on the finish of an extended GitHub submit outlining its strategy to attaining “increased throughput and decrease latency.” The corporate wrote that when it seems to be at utilization of its V3 and R1 fashions throughout a 24-hour interval, if that utilization had all been billed utilizing R1 pricing, DeepSeek would have already got $562,027 in day by day income. 

In the meantime, the price of leasing the mandatory GPUs (graphics processing items) would have been simply $87,072.

The corporate admitted that its precise income is “considerably decrease” for quite a lot of causes, like nighttime reductions, decrease pricing for V3, and the truth that “solely a subset of providers are monetized,” with net and app accessing remaining free.

After all, if the app and web site weren’t free, and if different reductions weren’t out there, utilization would presumably be a lot decrease. So these calculations appear to be extremely speculative — extra a gesture in the direction of potential future revenue margins than an actual snapshot of DeepSeek’s backside line proper now.

However the firm is sharing these numbers amidst broader debates about AI’s value and potential profitability. DeepSeek leapt into the highlight in January, with a brand new mannequin that supposedly matched OpenAI’s o1 on sure benchmarks, regardless of being developed at a a lot decrease value, and within the face of U.S. commerce restrictions that stop Chinese language corporations from accessing essentially the most highly effective chips. Tech shares tumbled and analysts raised questions on AI spending.

DeepSeek’s tech didn’t simply rattle Wall Avenue. Its app briefly displaced OpenAI’s ChatGPT on the high of Apple’s App Retailer — although it’s subsequently fallen off the final rankings and is presently ranked #6 in productiveness, behind ChatGPT, Grok, and Google Gemini.