Aug. 21 — Chinese AI chipmaker Moore Threads founder Zhang Jianzhong said at the
2026 World Robot Conference that the embodied-intelligence sector is entering a
ChatGPT-style inflection and that compute capacity is the critical constraint.
Language models have passed the trillion-parameter threshold, and driven by
scaling laws embodied-intelligence models are advancing from billions and tens
of billions to hundreds of billions and trillions of parameters. AI compute
clusters will expand from thousands of accelerator cards to tens of thousands,
and a move toward 100,000-card scale has become inevitable.
