DeepSeek reportedly developing in-house AI inference chips

According to multiple reports on July 7, DeepSeek is secretly developing its own AI inference chips. The company has been recruiting chip engineers for months and has approached multiple wafer foundries and memory suppliers. The project aims to reduce long-term dependence on US GPUs, focusing on inference efficiency. Previous US chip sanctions on China indirectly spurred DeepSeek to develop innovative architectures like MLA (Multi-Head Latent Attention).