• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Jukan Reports DeepSeek Huawei Chip Purchase Plans

    Semiconductor analyst shares Bloomberg-sourced details on DeepSeek's chip procurement intentions.

    T(
    RP
    CH
    7 Sources, 26d ago, first seen 26d ago

    TLDR

    Semiconductor commentator Jukan posted that Bloomberg reports DeepSeek plans to purchase more than 160,000 Huawei AI chips. The post states DeepSeek does not plan to use Huawei chips for training. It adds that DeepSeek wants to buy more Huawei chips, but Huawei faces limited production capacity. The post cites shortages of HBM and other memory components as constraints on Huawei's output.

    Combined views

    134.4K

    7 Sources, first seen 26d ago

    Combined views

    134.4K

    7 Sources, first seen 26d ago

    1.2K likes
    1.2K likes
    62 comments
    165 saves
    69 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    62 comments
    165 saves
    69 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    7 Sources

    @jukan05BBG: DeepSeek plans to purchase more than 160,000 Huawei AI chips. For now, DeepSeek does not plan to use Huawei chips for training. DeepSeek wants to purchase more Huawei chips, but Huawei is currently constrained by limited production capacity. Due to shortages of HBM and other memory components, Huawei’s production capacity for the 950DT is expected to be limited to only a few hundred thousand units this year. It could take Huawei more than a year to fully meet DeepSeek’s orders. If completed as planned, Inner Mongolia will become the key hub driving DeepSeek’s AI development.
    @kimmonismusDeepSeek reportedly plans to deploy at least 160,000 Huawei Ascend 950DT chips in Inner Mongolia, potentially creating one of the largest known Huawei AI clusters! Bloomberg reports the chips would run DeepSeek’s models. The company currently has no plans to use them for training, where it has so far relied on Nvidia. How capable are they? Huawei’s announced specs include 144GB of high-bandwidth memory and 4TB/s of memory bandwidth. Nvidia’s H200 offers 141GB and 4.8TB/s. That puts Huawei close on memory capacity, with lower bandwidth. It does hwever not establish equivalent real-world performance. Bloomberg describes the 950DT as broadly comparable to Nvidia’s older Hopper generation. Crazy at which scale china is building their national chip industry.
    @teortaxesTex20 SuperPods to DeepSeek. Might take more than a year. Very sad noises. Of course this is consistent with… everything, from job listings to investor call to the general logic of moving towards domestic silicon. Whale will be taking a while.
    @rohanpaul_aiDeepseek plans the largest known Huawei chip cluster with 160,000 processors in Inner Mongolia Bloomberg says the chips would go into its gigawatt-scale Inner Mongolia data center, creating one of the largest known Huawei AI clusters if completed. Huawei designed the 950DT for decode inference and training, pairing 144GB of memory with 4TB/s memory bandwidth and 2TB/s interconnect bandwidth. Huawei's own roadmap schedules the Huawei 950DT for Q4 2026. The 160,000 chips would fill only part of the gigawatt-scale site, leaving DeepSeek's remaining accelerator mix unresolved. If fulfilled, the project would move a large share of DeepSeek's inference onto Chinese silicon without removing Nvidia from its most demanding training workloads.
    @xlr8harder@teortaxesTex also, whether or not the chips are good enough to train with is a different question if you already have an nvidia cluster.

    7 Sources

    @jukan05BBG: DeepSeek plans to purchase more than 160,000 Huawei AI chips. For now, DeepSeek does not plan to use Huawei chips for training. DeepSeek wants to purchase more Huawei chips, but Huawei is currently constrained by limited production capacity. Due to shortages of HBM and other memory components, Huawei’s production capacity for the 950DT is expected to be limited to only a few hundred thousand units this year. It could take Huawei more than a year to fully meet DeepSeek’s orders. If completed as planned, Inner Mongolia will become the key hub driving DeepSeek’s AI development.
    @kimmonismusDeepSeek reportedly plans to deploy at least 160,000 Huawei Ascend 950DT chips in Inner Mongolia, potentially creating one of the largest known Huawei AI clusters! Bloomberg reports the chips would run DeepSeek’s models. The company currently has no plans to use them for training, where it has so far relied on Nvidia. How capable are they? Huawei’s announced specs include 144GB of high-bandwidth memory and 4TB/s of memory bandwidth. Nvidia’s H200 offers 141GB and 4.8TB/s. That puts Huawei close on memory capacity, with lower bandwidth. It does hwever not establish equivalent real-world performance. Bloomberg describes the 950DT as broadly comparable to Nvidia’s older Hopper generation. Crazy at which scale china is building their national chip industry.
    @teortaxesTex20 SuperPods to DeepSeek. Might take more than a year. Very sad noises. Of course this is consistent with… everything, from job listings to investor call to the general logic of moving towards domestic silicon. Whale will be taking a while.
    @rohanpaul_aiDeepseek plans the largest known Huawei chip cluster with 160,000 processors in Inner Mongolia Bloomberg says the chips would go into its gigawatt-scale Inner Mongolia data center, creating one of the largest known Huawei AI clusters if completed. Huawei designed the 950DT for decode inference and training, pairing 144GB of memory with 4TB/s memory bandwidth and 2TB/s interconnect bandwidth. Huawei's own roadmap schedules the Huawei 950DT for Q4 2026. The 160,000 chips would fill only part of the gigawatt-scale site, leaving DeepSeek's remaining accelerator mix unresolved. If fulfilled, the project would move a large share of DeepSeek's inference onto Chinese silicon without removing Nvidia from its most demanding training workloads.
    @xlr8harder@teortaxesTex also, whether or not the chips are good enough to train with is a different question if you already have an nvidia cluster.