DeepSeek Develops Custom AI Chips to Reduce Dependence on NVIDIA

Chinese startup DeepSeek, specializing in artificial intelligence model development, has begun creating custom processors to accelerate AI inference. The project aims to reduce computational costs and dependence on foreign suppliers, primarily NVIDIA. No official comments have been released by the company yet, but sources indicate that chip development has been underway for about a year.

Unlike AI model training, which requires short-term peak loads, inference involves continuous processing of user requests. This makes energy efficiency, operational costs, and system reliability critically important. According to Reuters, DeepSeek is actively expanding its engineering team, hiring specialists in chip architecture, verification, and software development.

Developing custom chips has become a strategic priority for many leading AI companies. For DeepSeek, this presents an opportunity to not only cut long-term expenses but also improve model deployment efficiency. However, chip development is a complex and costly process that may take several years. The project is still in its early stages, and its impact on the AI chip market will not be immediate.

Simultaneously, DeepSeek is preparing for its first external funding round. According to media reports, the company plans to raise approximately $7 billion at a valuation of $52–59 billion. The funds may be allocated to infrastructure development, including chip design and other AI technologies.