AI News

Where is China's Groq? Why doesn't China make HBM-free chips

ChinaTalkUpdated Nov 16 19:10

AI summary

The ChinaTalk article explores why China has not produced HBM-free chips similar to Groq. The article notes that by the end of 2025, the industry widely believed AI inference must rely on HBM, because inference decoding is an operation constrained by memory bandwidth; but since then the market has matured, and no single architecture dominates everything anymore—different chips are better suited to training, prefill, or decode.

Body

Where is China’s Groq?

Why isn’t China making HBM-free chips?Aqib Zakaria and Hamish LowOct 23, 2026

In late 2025, HBM was all the rage. You need HBM for AI inference; inference decode is a memory-constrained operation, and only HBM was able to reach bandwidth speeds to serve models effectively.

But the market has matured since then, and there is no longer one architecture to rule them all. Different silicon works better for training, prefill, or decode…

Source

ChinaTalk