DeepSeek, a Chinese AI startup, recently revealed the potential profitability of its AI models, but with some important caveats.
In a post on X, DeepSeek claimed that its online services have a “cost profit margin” of 545%. However, it’s important to note that this margin is based on “theoretical income.”
The company provided further details on these numbers in a longer GitHub post that outlines its approach to achieving higher throughput and lower latency. DeepSeek stated that if it billed the usage of its V3 and R1 models during a 24-hour period using R1 pricing, it would have generated $562,027 in daily revenue. The cost of leasing the necessary GPUs would have been $87,072.
DeepSeek acknowledged that its actual revenue is significantly lower due to various factors, such as nighttime discounts, lower pricing for V3, and the fact that only a subset of services are monetized, while web and app access remain free.
It’s important to recognize that these calculations are speculative and represent potential future profit margins rather than an accurate reflection of DeepSeek’s current financial situation. However, the company’s decision to share these numbers contributes to the ongoing discussions surrounding the cost and profitability of AI.
DeepSeek gained attention in January when it introduced a new model that supposedly matched OpenAI’s o1 on certain benchmarks while being developed at a lower cost and despite facing U.S. trade restrictions. This development led to a decline in tech stocks and raised concerns among analysts about AI spending.
Not only did DeepSeek’s technology impact Wall Street, but its app briefly surpassed OpenAI’s ChatGPT as the top app in Apple’s App Store. However, it has since fallen in the rankings and currently sits at #6 in productivity, behind ChatGPT, Grok, and Google Gemini.
