<![CDATA[DeepSeek to order 160000 Huawei AI chips over Nvidia]]>This post did not contain any content.
Just a moment...

favicon

(www.huaweicentral.com)
]]>https://citiverse.it/topic/d3971acb-f397-461a-bf51-7766f78b9b75/deepseek-to-order-160000-huawei-ai-chips-over-nvidiaRSS for NodeSun, 06 Sep 2026 12:33:11 GMTSat, 05 Sep 2026 23:07:47 GMT60<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 12:12:53 GMT]]>Yes but no. Asics are very inflexible and have a hard time dealing with new functions. GPU's are general purpose and therefore new types of convolutions are much easier to implement and have the driver support for it.

There are frequent claims of random companies making NPU's TPU's etc but they always have software issues for training new models with the latest methods. If not Nvidia wouldn't be having such massive market share anymore.

]]>https://citiverse.it/post/https://lemmy.ml/comment/27661782https://citiverse.it/post/https://lemmy.ml/comment/27661782Sun, 06 Sep 2026 12:12:53 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 12:04:36 GMT]]>Nvidia does not have a monopoly on training.

Link Preview Image
meituan-longcat/LongCat-2.0 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

favicon

(huggingface.co)

Both the full training run and the large-scale deployment are built entirely on AI ASIC superpods. Pretraining spans millions of accelerator-days across more than 35 trillion tokens, with no rollbacks or irrecoverable loss spikes — demonstrating that we have the capability to conduct frontier-scale training on alternative hardware platforms.

LongCat 2.0 is of similar size to DeepSeek V4 Pro and "AI ASIC Superpods" are from Huawei.

]]>https://citiverse.it/post/https://lemmygrad.ml/comment/8673961https://citiverse.it/post/https://lemmygrad.ml/comment/8673961Sun, 06 Sep 2026 12:04:36 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 11:08:25 GMT]]>Training has a lot of extra functionality like calculating how to update the weights of the model during training to make it more performant on the dataset(backpropagation) and much more

Meanwhile inference is mostly running the weights of the model as they are. The model isn't being adjusted in any way. And Nvidia holds a strong grip on training libraries through Cuda

]]>https://citiverse.it/post/https://lemmy.ml/comment/27661183https://citiverse.it/post/https://lemmy.ml/comment/27661183Sun, 06 Sep 2026 11:08:25 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 11:05:03 GMT]]>They run during training too. Also, why would they get 160000 chips only for inference?

]]>https://citiverse.it/post/https://lemmygrad.ml/comment/8673881https://citiverse.it/post/https://lemmygrad.ml/comment/8673881Sun, 06 Sep 2026 11:05:03 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 09:48:17 GMT]]>If they can be used for inference, I would assume they can be used for training

]]>https://citiverse.it/post/https://programming.dev/comment/25846132https://citiverse.it/post/https://programming.dev/comment/25846132Sun, 06 Sep 2026 09:48:17 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 08:34:49 GMT]]>Running refers to inference usually

]]>https://citiverse.it/post/https://lemmy.ml/comment/27660150https://citiverse.it/post/https://lemmy.ml/comment/27660150Sun, 06 Sep 2026 08:34:49 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 07:31:55 GMT]]>The article says "to run AI models" so it probably means both inference and training, not just inference.

]]>https://citiverse.it/post/https://lemmy.ml/comment/27659641https://citiverse.it/post/https://lemmy.ml/comment/27659641Sun, 06 Sep 2026 07:31:55 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 00:22:02 GMT]]>they might not need more training infrastructure at this point

]]>https://citiverse.it/post/https://lemmy.ml/comment/27656193https://citiverse.it/post/https://lemmy.ml/comment/27656193Sun, 06 Sep 2026 00:22:02 GMT<![CDATA[Reply to DeepSeek to order 160000 Huawei AI chips over Nvidia on Sun, 06 Sep 2026 00:16:58 GMT]]>Only for inference like the article says or also training?

]]>https://citiverse.it/post/https://lemmy.ml/comment/27656144https://citiverse.it/post/https://lemmy.ml/comment/27656144Sun, 06 Sep 2026 00:16:58 GMT