Moore Threads CEO: AI Computing Clusters Inevitably Scaling to 100K GPUs

At the 2026 World Robot Conference, Moore Threads founder Zhang Jianzhong emphasized that the embodied intelligence industry is approaching its 'ChatGPT moment' similar to the breakthrough seen in large language models, making computing infrastructure a critical foundation. He noted that language models have already crossed the trillion-parameter threshold, and driven by the scaling law, embodied intelligence models are now evolving from 10-billion and 10-billion parameter scales toward 100-billion and trillion-parameter levels.

Zhang highlighted that this growth necessitates a corresponding expansion of AI computing clusters, which are moving from thousands of GPUs to tens of thousands. He concluded that scaling to 100,000 GPUs has become an inevitable trend, underscoring the immense demand for high-performance computing infrastructure in the near future.