Chinese AI startup DeepSeek has claimed that its AI models could generate a "cost profit margin" of 545%, though the figure is based on theoretical income.

In a recent post on X, DeepSeek explained that the 545% margin reflects a calculation using "theoretical income" rather than actual revenue. The company elaborated on these numbers in a GitHub post, stating that if all usage of its V3 and R1 models over a 24-hour period had been billed at R1 pricing, the daily revenue would be $562,027.

Meanwhile, the cost of leasing GPUs graphics processing units required to support these models would amount to $87,072 per day. These calculations form the basis of DeepSeek's 545% profit margin claim.

However, DeepSeek acknowledged that its actual revenue is "substantially lower" due to factors like nighttime discounts, lower pricing for the V3 model, and the fact that not all services are monetized. Access to DeepSeek’s app and website remains free, further reducing its current income.

The company's announcement comes amid broader discussions about AI’s costs and profitability. DeepSeek gained attention in January after launching a new AI model that reportedly matched OpenAI's o1 performance on certain benchmarks. This achievement was notable as DeepSeek developed the model at a lower cost, despite U.S. trade restrictions preventing Chinese firms from accessing advanced AI chips.

The company's app also briefly surpassed OpenAI’s ChatGPT at the top of Apple’s App Store rankings. However, it has since dropped to the sixth spot in the productivity category, behind competitors like ChatGPT, Grok, and Google Gemini.

While DeepSeek's profit margin claim highlights its ambition, the figures remain speculative, serving more as an indicator of future potential than a reflection of current financial performance.

PHOTO: BLOOMBERG/ANDREY RUDAKOV

This article was created with AI assistance.