Huawei 950 supernode is described as a GB200/GB300 substitute, but capacity is still tight
teortaxesTex · x · 2026-07-23
A post quoting an industry discussion says DeepSeek-style training has already moved away from NVIDIA’s ecosystem and could be ported to Huawei GPUs next.
Key points
- Huawei’s 950 supernode is described as able to fully substitute for NVIDIA GB200 and GB300 in both performance and price.
- The quoted view claims the main remaining gap is chip-level, not ecosystem-level: Huawei is said to be about 4× behind plus roughly two years.
- Huawei’s current capacity is described as limited: about 16,000 GPUs for some partners, while internet giants may want 100K+.
- The conclusion is that the ecosystem problem is largely solved, but capacity still constrains training ever larger models.
Related event: Leaked DeepSeek Meeting Reveals AGI Roadmap and Compute Strategy(31 posts)→
More from Infra
- MiniMax says MI355X is nearing B200 for serving its 428B multimodal model — salykova_ · 2026-07-23
- Aurora launches an open-source Go gateway for routing and securing LLM traffic — Select-Medicine-9310 · 2026-07-23
- Investors betting against infrastructure spending are missing the current AI cycle — cgarciae88 · 2026-07-23
- TileLang debate says leaving CUDA could cut inference costs with only 1–2% loss — teortaxesTex · 2026-07-23
- Why inference is becoming the bigger AI hardware story than training — Genzinvestor16180339 · 2026-07-23
- Agentic RL setup hosts 135,000 task images and 1,000 concurrent rollouts — willccbb · 2026-07-23