Chinese labs likely using Huawei for inference and Nvidia for training, says Lambert
natolambert · x · 2026-09-20
Nathan Lambert's best guess: as RL scaling ramps up, top Chinese AI labs are starting to use Huawei chips heavily for inference while relying on Nvidia for training (possibly not for unusual architectures). As agent swarms and ever-larger post-training runs become the norm, he expects this pattern to accelerate China's domestic chip industry.
More from Infra
- DeepSeek-V4.1-Flash redesigns the Transformer for agents, cutting KV cache to 890 bytes/token — AndLukyane · 2026-09-21
- Jev Engineering gives agents a decision brain, 193x faster and 444x cheaper in tests — agihouse_org · 2026-09-21
- UK's £225m Isambard-AI supercomputer cost about the same as one road bridge — charlieharris01 · 2026-09-21
- Building a multi-model agent gateway: streaming and tool calls are the real hard part — Amendxd · 2026-09-21
- Writing 100 facts into Qwen's n-gram Engram table with zero weight changes: 84% recall — Electronic_Put4530 · 2026-09-21
- Which speculative decoding setup for local Qwen models? Two llama.cpp configs compared — cradlemann · 2026-09-21