AI Labs DeepSeek and Zhipu AI Develop Custom Chips, Shifting Industry Focus
Chinese AI companies DeepSeek and Zhipu AI are now developing their own custom inference chips. This move mirrors efforts by OpenAI and Anthropic, which are also pursuing in-house silicon development. The trend indicates a significant structural shift within the leading AI research and development organizations. By creating their own chips, these companies aim to reduce their substantial reliance on third-party GPUs, primarily from NVIDIA. This strategy is driven by the desire to cut escalating operational costs associated with AI model inference. The ultimate goal appears to extend beyond merely developing advanced AI models to controlling the underlying hardware infrastructure. This strategic pivot suggests a long-term vision for greater autonomy and cost-efficiency in AI deployment.
AI development is increasingly characterized by vertical integration, with leading labs like DeepSeek and Zhipu AI moving from pure software to in-house hardware design. This strategic shift is driven by the economic realities of AI inference, where GPU costs represent a significant portion of operational expenditure. By developing custom chips, these organizations seek to optimize performance and reduce dependency on a limited number of hardware suppliers, thereby gaining greater control over their supply chain and cost structure. This trend reflects a broader industry movement towards specialized hardware tailored for AI workloads, potentially leading to greater efficiency but also concentrating significant technological power within a few entities. The long-term implications involve potential shifts in the semiconductor market and the competitive landscape for AI development.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.