By 11 August 2031 the main path for Chinese frontier AI does not depend on smuggled NVIDIA boards and Western HBM. The race is the memory wall, packaging, megawatts, and the compiler — not the next gadget roadmap.
Open-weight models are sold as more open than API models. A neural net is billions of opaque floats either way. Open mostly means run it yourself — usually on NVIDIA, sometimes the Ascend bet.
DeepSeek’s new model generation, tuned for Huawei Ascend, is the software half of the NVLink-vs-open-bus fork. How likely is Chinese hardware independence? High for a sovereign inference lane; contested at the deepest manufacturing layers.
Pulling NVLink from RTX 40-class GeForce was not a gamer gift. It reserved multi-GPU fabric for enterprise inference while home labs got PCIe 5.0 ×16. Huawei’s open UnifiedBus 2.0 specs point the other way: link many NPUs as one machine.
Published June 2026 · Infrastructure notes from 3DN This is not a political manifesto and it is not written to shock anyone. It is a note about electricity — the kind of boring constraint that decides whether an AI strategy … Continued
Four RTX 3090s, 256 GB RAM, and GLM 5.2 running scheduled server-admin jobs — inference and operations that stay on premises.
