News Summary

Chinese AI startup DeepSeek has recently claimed a theoretical profit margin of 545% for its AI models, V3 and R1, despite the actual revenue being substantially lower due to various factors such as free services and discounts. The company outlined these figures in a GitHub post, discussing the operational costs and revenue potential. If usage during a 24-hour period had been billed at R1 pricing, the revenue could have reached $562,027, with GPU leasing costs at $87,072. However, these numbers are speculative, aimed at showcasing potential future profit margins rather than current financial performance. The reports come amid broader debates on AI’s profitability and cost efficiency, as DeepSeek’s models previously challenged market leaders like OpenAI, contributing to fluctuations in tech stock prices. This disclosure marks the first time DeepSeek has shared insights into the profit margins of its inference tasks, which play a crucial role in real-world AI applications.