Chinese artificial intelligence startup DeepSeek unveiled financial figures on Saturday, indicating that its “theoretical” profit margin might exceed costs by more than five times, shedding light on the often opaque business models within the AI sector.
The 20-month-old company, which has disrupted Silicon Valley with its innovative and cost-effective approach to AI model development, stated on X that its V3 and R1 models achieved a profit margin of 545% based on the cost of inferencing relative to sales during a 24-hour period at the end of February. Inferencing encompasses the computing power, electricity, data storage, and other resources required for real-time AI model operation. However, DeepSeek included a disclaimer in the information shared on GitHub, clarifying that its actual revenues are significantly lower due to several factors. These include the limited monetization of its services, discounts offered during off-peak hours, and the exclusion of all research and development (R&D) and training costs associated with building its models. Consequently, while the reported profit margins are striking, they remain hypothetical. This disclosure comes at a time when the profitability of AI startups and their business models has become a pressing topic among technology investors. Companies such as OpenAI Inc. and Anthropic PBC are exploring various revenue models, including subscriptions, usage-based charges, and licensing fees, as they strive to develop increasingly sophisticated AI products. However, investors are scrutinizing these business models and questioning their potential return on investment, raising discussions about the feasibility of achieving profitability in the near future.
DeepSeek announced on X that its online service had a “cost profit margin of 545%” and provided insights into its operations, including its strategy for optimizing computing power through load balancing—effectively distributing work evenly across multiple servers and data centers. The startup also highlighted its innovations in maximizing data processing efficiency within a given timeframe and managing latency, which is the delay between a user submitting a query and receiving a response.
In a series of unprecedented moves earlier this week, DeepSeek, which advocates for open-source AI, surprised many in the industry by sharing key innovations and data related to its models, contrasting sharply with the proprietary approaches of major US competitors like OpenAI.