Andy Halliday

Well, reportedly...

China has just approved purchase of over 400,000 H200 chips.

Now, those are not the latest and greatest chips from NVIDIA, but they're very powerful chips and could accelerate in the near term, accelerate

you know, China's ability, the Chinese companies.

So I think ByteDance, Alibaba, and Tencent are the main ones there.

They'll be able to use these chips in their inference, excuse me, in their training runs to advance their models, which are already near state of the art.

So this is what Dario Amadei is saying is like ill-advised.

Yeah.

But it looks like that's going to proceed to NVIDIA's benefit and to shareholders of NVIDIA's benefit.

Right.

So it's probably going to whipsaw yet again.

I mean, this is just a reported approval by China.

But who knows what's going to happen?

Maybe Trump, you know, reading the news about what Dario Almeida said, is going to reinstitute, reinstitute, reinstate, reinstate the embargo.

Oh, well, I think anytime you create a, uh,

a confined or constrained set of application scope, then you're going to be able to advance the capability and the efficiency of that more directly than you can in sort of a universal, autonomous, embodied AI application that then has to learn the specific nuances of a very narrow scoped effort.

So, yeah, I think it's a good example of where you'll start to see the emergence of that AI application in ways that don't have profound risks with them when either hallucination or mistakes made by attempted reasoning robots, whether cars or a humanoid, are navigating our streets.

I use GenSpark.

I use it not for coding or anything like that.

I'm using the better coding models for that.