Chinese AI startup DeepSeek has made bold claims about the profitability of its AI models—though with some significant caveats.
In a recent post on X, DeepSeek said that its online services have a staggering 545% “cost profit margin.” However, this figure is based on “theoretical income” rather than actual revenue.
The company elaborated on this in a longer GitHub post, breaking down its calculations based on a 24-hour usage period for its V3 and R1 models. The organization believes if all usage had been billed at R1 pricing, its estimated daily revenue would reach $562,027, while the cost of leasing the necessary GPUs would be just $87,072.
However, the company also admitted that its actual revenue is “substantially lower” due to various reasons, including:
- Nighttime discounts
- Lower pricing for V3 models
- The fact that many services remain free (such as web and app access)
If the app and website weren’t free, and other discounts weren’t available, usage would presumably be much lower. So it looks like these calculations are highly speculative, and they are more of a gesture towards potential future profit margins than a real snapshot of DeepSeek’s bottom line right now.
The Chinese startup’s claims came among the ever-growing concerns about AI’s profitability. Deepseek made waves in January 2025, when it launched a model that matched OpenAI’s GPT-4o in some parameters, despite being built to a cost, and under trade restrictions blocking Chinese companies from accessing powerful AI chips.
Once launched Deepseek sent shockwaves across the industry, which led AI stocks to tumble and spark new debates on AI spending and efficiency.
DeepSeek’s disruption didn’t stop at Wall Street—its app eventually surpassed OpenAI’s ChatGPT at the top of Apple’s App Store rankings for a brief period. However, since then it has slipped and is now ranked #6 in the productivity category, trailing ChatGPT, Grok, and Google Gemini.
Discussion about this post