ChinaTalk· Aqib Zakaria·· 22 天前35
中国的Groq在哪里？
Where is China’s Groq?
要点速览
HBM 曾在 2025 年底风靡一时，因其能通过高带宽满足 AI 推理需求，但市场已成熟，单一架构不再主导，不同芯片在训练、预填充或解码方面表现各异。
正文
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…
Subscribe for free to continue reading
Enter your email and get every post from ChinaTalk.
来源: ChinaTalk · chinatalk.media