Chinese artificial intelligence developer DeepSeek is engineering its own custom silicon specifically designed for AI inference, according to reports from multiple wire services.
The chip targets the inference phase of AI processing, a critical stage where trained models generate outputs for end users.
This move marks a significant step toward vertical integration for the startup, which has previously relied on third-party hardware to power its large language models.
The development underscores the intensifying competition in the AI hardware landscape, where major tech firms and startups alike are seeking to optimize performance and reduce dependency on external suppliers.
By designing chips tailored to its specific inference needs, DeepSeek aims to enhance efficiency and potentially lower operational costs.
This strategy mirrors trends seen among larger players in the industry, who are increasingly investing in proprietary silicon to gain a competitive edge.