DeepSeek and the Effects of GPU Export Controls

The Chinese do have their home grown GPUs too, although I have the impression that they&#x27;re not super good.<p>Even so, if we look at Groq &#x2F; Cerebras, the fastest LLM inference companies:<p>They&#x27;re both based on architectures that are 7nm+, and so architectures that China can produce locally despite the export restrictions.<p>Ultimately, the export controls are mainly just inconvenience. Not a real blocker.<p>The Chinese don&#x27;t need to achieve state of the art chip manufacturing to achieve SOTA AI outcomes.<p>They just need to make custom silicon specialized for the kinds of AI algorithms they want to scale.<p>Of course, at scale, that&#x27;s going to mean that the US should eventually have both lower production costs, and energy use in consumer use of AI models, and that Chinese products will likely be more dependent on the cloud for at least the near future.<p>The whole strategy seems ultimately meh in a long term sense... Mainly good for building up a sense of mutual enmity and dividing the world ... Which is also going to result in higher cost of living around the world as trade falters.<p>Sad stuff.

It is important to keep in mind that GPUs power per $ is what matters, and not per unit.<p>China can produce much cheaper electronics that can compete even when they aren&#x27;t as powerful as NVIDIA&#x27;s

&gt; Infrastructure algorithm optimization<p>&gt; Novel training frameworks<p>Where can one find more information about these? I keep seeing hand-wavy language like this w.r.t. DeepSeek’s innovation

&gt; DeepSeek isn&#x27;t a typical startup - they&#x27;re backed by High-Flyer, an $8B quant fund. Their CEO Liang Wenfeng built High-Flyer from scratch and seems focused on foundational research over quick profits<p>How is that useful?

Excellent models that need a fraction of compute were obviously going to come from this. OAI is actually encouraged to not to try to make their models because compute is a moat too.

Question: What&#x27;s stopping China from buying GPUs via third party middle-men countries that don&#x27;t have export controls to China?<p>I would assume nothing, similarly to how exports of western tech from western countries somehow magically exploded overnight to Russia&#x27;s neighbors and everyone is pretending not to notice because it makes money.<p><a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;kDCsxbt.jpeg" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;kDCsxbt.jpeg</a>

DeepSeek shows that it is not the size of your computer that matters the most, rather your talent, and the approach you are taking.<p>Should have been obvious but now somehow isn&#x27;t?

With $8B in the bank I have some degree of confidence Deepseek evaded the export controls and used full-fat GPUs in addition to the H800s.

There is also rumor that they in fact have access to 50000 H100 GPU, and not just H800. 50000 H100 is as big as half of Elon Musk&#x27;s Colossus!

This article is assuming that they are being truthful and indeed had access to limited hardware resources, which is doubtful to say the least.

I still don&#x27;t understand the insane investments in LLM with the believe that it will get us to AGI when that is not possible with LLM. The limitation isn&#x27;t compute or model size, it&#x27;s the core concept of LLM.