00:00One area closes the gap with another. You are bound to get gaps anyway.
00:04Mark, they're trying to figure out where the gaps are closing quicker than other places.
00:10What's your take of, you know, just compare and contrast the ecosystems for us, U.S. and China?
00:15Sure. I think when you look at both industries and both countries, I think you're going to find some strengths in each one.
00:21I would say that the U.S. continues to lead in AI, in compute, and even its ability to put some of these things together because of the large, fabulous domains that we find in the U.S.
00:34When you think about China, China is the largest semiconductor industry as a whole.
00:39And today, about 25 to 30 percent of all of the demand in China is being served by the local design houses and suppliers within the supply chain.
00:49So the strengths that we see moving forward is that China, over the past five to seven years, has continued to build a tremendous amount of capacity, specifically in the mainstream and mature notes.
01:02And we think that as we look at the end of this decade, they're going to control about 30 percent of the total capacity around the world.
01:08So any type of design, whether it's a smartphone or a 5G infrastructure piece of equipment or a PC, about 70 percent of all the components that go into these systems is primarily using mature and mainstream technology.
01:23So even though we tend to kind of gravitate very quickly to AI and CPUs and high bandwidth memory, the reality still is that a large amount of the different market segments that consume silicon continue to use some of these older technologies that now China is becoming a very strong player in.
01:42Yeah, so as China increases its capabilities and market share as well of mature nodes, how does that then also erode the power of names like SK Hynix or Samsung and then their ability to reinvest in R&D in return?
02:01Well, I think when you look at the memory space specifically, you know, you do have two big players coming out of China.
02:09Each of these companies control about three to four percent of the market overall.
02:14So there's still quite a bit of time before they can get to a position where they're as competitive as SK Hynix or Micron or even Samsung.
02:22But they are growing. And the more the OEMs in China continue to switch suppliers going from a Samsung to one of these local suppliers,
02:31like CXMT or YMTC, then you're going to see some of their market positions continue to grow.
02:38These companies are growing bits a lot faster. They're investing very aggressively, but they're going to need more time to catch up from a process and technology standpoint.
02:47But they are making gains. And I think the more they're able to scale with some of the Chinese OEMs, the bigger they have a potential to grow.
02:56Chinese chip makers are really limited at seven nanometer capacity, but we are seeing names like Huawei, for instance, building clusters of AI chips instead.
03:08And they're saying that as a result, you are able to get a lot of power and a lot of technology out of those clusters together.
03:15Pursuing that type of architecture, how much does it then overcome the hurdles from lack of access to chips from names like NVIDIA?
03:24I think when you look at the play here, I think that necessity always drives innovation.
03:29And even though Huawei still doesn't have the same access that NVIDIA or AMD has in terms of process technology,
03:36they are trying to find unique ways of designing and building systems.
03:40And I think we cannot underestimate the fact that Huawei has been building systems for decades in 5G infrastructure or 4G infrastructure.
03:50And that know-how has continued to be something that they can leverage.
03:54You know, some of the Chinese companies, whether it's Huawei or Capricorn or Bitmain or Canon Creative,
04:00these are all companies that are very good at designing ASICs and very good at designing SOCs.
04:05And now they're putting their assets to develop these AI chips.
04:10And you're seeing some unique partnerships between some of these companies and some of the cloud service providers within China.
04:16And so I think that what we cannot underestimate is their ability to design.
04:21Now, some of these smaller companies like Capricorn and Canon Creative,
04:25they still have access to other foundries beyond the Chinese foundries.
04:30The only one that's limited has been Huawei, but you can see a resurgence already in Huawei just based on the fact that they're beginning to grow market share again,
04:38both in the equipment side of things, but also even in smartphones.
04:43And that always bodes well for the ability for them to continue to grow.
04:47And I think the most important thing, too, is when you think about AI, you're going to need a tremendous amount of data to feed these large language models.
04:55And a lot of the data that they're leveraging is the data that's being created within smartphones.
05:00And so the smartphone, the smart home, and you think about the automotive areas, these are all markets that China is leading in.
05:07And they're able to use these as hubs to continue to grow the data foundation that you need to feed and train these AI models.
05:15And so I think every single day you'll see a Huawei or anybody else that's in this space continuing to develop their silicon based on all this data that they're able to get access to.
05:26Yeah, and we certainly are seeing the result of that, which is Huawei, as you said, managing to increase its market share domestically.
05:33I think one of the tradeoffs, though, with that architecture they're following is that it's very energy intensive,
05:38maybe not such a big burden in China when you've got more access to cheaper electricity, essentially.
05:46But how does it impact their ability to take this sort of technology abroad when you've got the U.S. and China that are competing for the full stack they want in other markets around the globe?
05:58It does continue to increase the cost burden, right?
06:01So when you think about someone like Huawei, they're already using technology that's likely four generations, if not more, behind the leaders in the space.
06:10And so not only is there a cost adder to the chip design itself, but now you have to add the cost of power.
06:17But like you said, having access to power in China is not probably the concern for someone like Huawei.
06:23But I think the cost structure is something that they're going to have to continue to address.
06:27Now, I do believe at some point in time, SMIC will be able to gain and develop their technology to be able to go down below 7 nanometer.
06:36But still, that cost burden will remain and also the yield concerns, too, because these products do not yield as good as what you would do if you were to build them with TSMC directly.
Be the first to comment