Chinese artificial intelligence startup DeepSeek has made bold claims regarding the profitability of its AI models, revealing a theoretical cost-profit ratio of up to 545% per day. The Hangzhou-based company disclosed this information on Saturday, marking the first time it has provided insights into its revenue and costs.
DeepSeek stated that it spent under $6 million on Nvidia H800 chips to train its V3 and R1 models, a figure significantly lower than what U.S. AI firms such as OpenAI reportedly invest in similar high-performance computing infrastructure. While these figures suggest an impressive return on investment, the company cautioned that actual revenue would be substantially lower than the theoretical maximum.
A key aspect of DeepSeek’s profitability comes from “inference” tasks—the stage in AI processing where trained models generate outputs, such as chatbot responses or data analysis. Compared to the costly training phase, inference is relatively less resource-intensive, allowing companies to recoup investments faster. DeepSeek’s disclosure highlights its efficient use of computing power, which could position it as a competitive player in the global AI landscape.
While U.S. AI giants have poured billions into developing and running advanced models, DeepSeek’s approach suggests that Chinese firms may be finding ways to achieve cost-effective AI deployment. The company’s announcement comes amid increasing scrutiny over AI investments and profitability, as many AI firms struggle to turn massive training expenses into sustainable revenue streams.
However, despite the promising figures, questions remain about the sustainability of DeepSeek’s model. Industry experts note that theoretical profitability calculations may not always translate into real-world success, especially given fluctuations in customer demand, operational costs, and competitive pressures. Additionally, regulatory challenges and geopolitical factors could impact DeepSeek’s growth trajectory.
DeepSeek’s transparency in sharing these figures signals a growing confidence in the Chinese AI sector, as domestic companies seek to assert their capabilities on the global stage. If the company’s cost-efficient approach proves viable, it could set a new benchmark for AI profitability and challenge the dominance of Western AI firms.
