BREAKING
Zhipu Weighs Custom AI Chip for GLM
0x
usage surge on Vercel
0M
max context tokens
Why Zhipu Wants Its Own Silicon
1US export controls
2Train on Huawei Ascend
3Explore custom ASIC
0B
total params
0B
active per token
Early Stage, Long Road Ahead
NowEarly
Preliminary inquiries
No partner selected
No formal project
Ahead2+ yrs
Chip design is slow
Limited near-term impact
Domestic foundry option
GLM Growth May Shape the Choice
AI NEWS BLITZ
Chinese AI lab Zhipu is exploring a custom chip for its GLM models as compute tightens.