Caroline Hyde

They've also turned down this idea that they're going to buy NVIDIA's H20 chips.

They don't think that those are good enough for the market.

So as Mike Shepard was talking about earlier, there may be a discussion about H200s.

But there's no putting this genie back in the bottle.

China has decided they are going to make their own chips domestically, and they're making a lot of progress.

They're making a lot of progress in designing them, but then they've got to get them spat out by SMIC.

And it feels as though ultimately the production, what you're really garnering, isn't very many.

Yeah, that's a very good point.

They are way, way behind the global standards.

There is some desperation here that if you have to produce at lower standards, you will.

But if you look at some of the details, it's very, very interesting.

So SMIC is still producing these chips at 7 nanometers, so well behind the 3 nanometer that we see from TSMC at this point.

Also, the yields, we've got some insights into the yields.

Only about 20% of the chips they're making right now are actually usable.

So that's only one out of five chips that they produce they're actually able to use.

You compare that with the TSMC, which is up at 80 or 90 percent yields.

Their economics are just totally different.

But it's a sign of how badly Beijing wants to be able to catch up in this market that they're just going to eat those costs.

They're going to eat four times the cost of what it would cost some other globally competitive manufacturers to make these chips because they feel like from a strategic standpoint, they need to make this progress.

Peter Ahlstrom, brilliant reporting.