AI news story

Deepseek plans the largest known Huawei chip cluster with 160,000 processors in Inner Mongolia

Deepseek wants to put 160,000 Huawei Ascend-950DT chips into an Inner Mongolia data center for inference only, not training. It would be the largest known Huawei chip cluster. But production bottlenecks mean Huawei probably can't deliver for over a y

  • Hardware
  • Source: The Decoder
  • Published: 2026-09-04
  • Signal score: 4
  • 9 sources

Editor's take

Deepseek is proposing a substantial AI inference cluster in Inner Mongolia, utilizing 160,000 Huawei Ascend-950DT processors.

This initiative underscores the growing demand for specialized inference hardware, particularly outside the U.S., and highlights Huawei's ambition to compete in the AI chip market against established players like NVIDIA. The sheer scale of the proposed cluster, even for inference-only tasks, signals a significant expansion of AI infrastructure in China and could influence global supply chain dynamics if realized.

The primary obstacle remains Huawei's ability to meet such a large order given current production constraints, potentially delaying deployment for over a year. The ultimate success of this project hinges on Huawei overcoming these manufacturing challenges and demonstrating a consistent supply of its Ascend chips, which will be crucial for Deepseek's ambitious inference plans.

Signal score: 4

This event was corroborated by 9 independent sources. The signal score weighs cross-source corroboration, recency, source weight and topic salience. How we rank stories.

More Hardware stories

  1. Japanese Stocks Advance as Tech, Chip Shares Follow US Peers

    Bloomberg · 2026-09-07

    Japanese stocks rose, driven by tech and chip shares, following a surge in AI and semiconductor-related names in the US on Friday.

  2. Perplexity Details Its GPU Embedding Stack: How Ivy, Tulip and ROSE Serve pplx-embed

    MarkTechPost · 2026-09-06

    Retrieval quality in an AI search product is bounded by two things: how good the embedding model is, and how cheaply you can run it across an index.

  3. Nvidia Partner Hon Hai’s Sales Climb 52% With AI Server Momentum

    Bloomberg · 2026-09-05

    Hon Hai Precision Industry Co. reported a 52% rise in monthly sales, lifted by demand for servers in a global race to build data centers and AI computational capacity.

  4. The KV Cache: AI’s Unseen Database Dominating GPU Memory

    Towards AI · 2026-09-05

    Large language models are memory-bound, not just compute-bound.

  5. NVIDIA Releases Personal AI Router (PAIR): An Open Source Virtual Inference Router that Distributes Local AI Requests Across RTX, DGX Spark, and Mac Nodes

    MarkTechPost · 2026-09-05

    We look at NVIDIA Personal AI Router (PAIR), an open source virtual inference router that spreads local AI requests across the machines already on a home network.

  6. Abu Dhabi’s G42 Weighs US Ownership to Safeguard AI Chip Access

    Bloomberg · 2026-09-04

    Executives at Abu Dhabi-based artificial intelligence firm G42 have held exploratory talks over potentially selling a majority stake to American companies