Zai's domestic inference cluster hits 100k+ chips; GLM-5.3 runs on custom interconnect
zephyr_z9 · x · 2026-08-27
LatePost reports Zai's domestic inference cluster now exceeds 100,000 chips, utilizing Ascend, Moore Thread, and Hygon hardware. Minimax also confirmed building large-scale domestic compute clusters in H1 earnings. Zai previously aimed to build AIDC in Inner Mongolia and will need million-chip data centers to serve trillions of tokens daily.
GLM-5.3 Flash's trial ran entirely on domestic accelerator clusters connected via a self-developed high-speed interconnect, using a dedicated inference engine on SGLang accelerated by a GLM-5.3-driven Infra Agent.
More from Infra
- Optical connectivity standards: CPO has rules, but NPO is chaos — jwt0625 · 2026-08-27
- ABF Substrates & PCB Identified as Key Constraints for 2027 AI Hardware — zephyr_z9 · 2026-08-27
- First startup enriches uranium for nuclear-powered data centers — Polymarket · 2026-08-27
- G2 spent $1.27M on 970B tokens, shifting focus from adoption to efficiency — prasanna_says · 2026-08-27
- Zhipu's GLM 3.5 Flash Served 42T Tokens in 6 Days Free on Chinese Chips — bindureddy · 2026-08-27
- TokenVisor supports Nvidia, AMD, and Intel GPUs in a single cluster — AccBalanced · 2026-08-27