Huawei can supply about 16,000 GPUs now, not enough for the next giant model
teortaxesTex · x · 2026-07-23
The speaker says domestic AI chips face no hardware or ecosystem obstacle, only insufficient production capacity. In practice, the team is working deeply with Huawei’s ecosystem, but Huawei can currently provide only about 16,000 GPUs—far below the 100,000-plus scale that internet giants may need. That makes training the next larger model or a several-hundred-billion-parameter model unrealistic today.
The post argues that the constraint is not technical feasibility but throughput and supply. It suggests there may be a chance to do much larger training runs on domestic hardware next year or the year after, as capacity expands.
Related event: DeepSeek Founder's Investor Call Reveals AGI Roadmap and Compute Plans(37 posts)→
More from Infra
- Two CVSS 9.9 Wasmtime flaws allow sandbox escape, one ARM64-only since 2020 — basedjensen · 2026-10-27
- AI Infra Meme: While Countries Need TW-Scale Power for Datacenters, People Argue About Lightbulbs — mgostIH · 2026-10-27
- Legora runs 2,000 parallel agent evals on Modal Sandboxes for legal AI — blelbach · 2026-10-27
- Cloud release velocity scores: AWS 95, Azure 90, Google Cloud 82 — DavidLinthicum · 2026-10-27
- Selling a 5090 to scale local inference: one business weighs 5 hardware paths — Anarchaotic · 2026-10-27
- AWS tutorial: fleet-wide edge CV model lifecycle with SageMaker SDK v3, ~$0.07 per run — AWS ML Blog · 2026-10-27