Transcript: The China Semiconductor Industry - A Deep Dive

 

with Steward Randall, Intralink & Nalin Balan, Reality AI

 

· Transcripts - CBD

ART: Welcome everybody to another episode of Ganbei. I'm your host Art Dicker. And today we have two wonderful guests. Stewart Randall is Director of Operations and Global Lead for Electronics and Embedded at Interlink in Shanghai. Interlink is an international business development consultancy specializing in East Asia. Stewart I know is often quoted in articles on the semiconductor industry, especially here in China. Stewart, I've seen you in Reuters, South China morning post and others. Welcome to the show Stewart.  

STEWART: Great to be here.  

ART: And Nalin, we go way back. you're currently Head of Sales and Business Development for Reality AI and you're based in San Francisco and Reality AI does edge AI for advanced sensing and automotive, industrial and consumer IOT products. We were both at Cadence Design Systems. I was Asia Pacific General Counsel. And you were Head of IP Sales for Asia Pacific.  And you were on an earlier episode last year and great to have you back on the show.  

NALIN: I'm in good company with the two of you here.  

ART: So I do want to give you guys a chance to talk a little bit more about your specific expertise and your company's expertise and how people can reach out to you after the show. 

NALIN: So briefly we're in the embedded machine learning space, we focus on enabling all types of sensor applications in commodity hardware. So if you're using, say for example, any kind of microcontroller and you want to deploy intelligent machine learning algorithms into those types of constraint environments give us a call. The automotive industry is probably our biggest segment. About 40% of our customers are in the automotive space.  But we're also in various, typical consumer IOT applications, manufacturing, and industrial sectors as well. Just find us on reality.ai I’m on LinkedIn, so that's a good place. DM me on LinkedIn, shoot me a note. Will be happy to give you some more insights on what you're up to.  

ART: Cool. Stewart?  

STEWART: Yeah, sure. I've been to Shanghai, I think, 13 years. So I'm working at a company called Interlink which essentially is a tech market entry consultancy for China, Japan, and Korea and Taiwan and a British company. We focus on the tech space. So our clients are scale up technology companies mainly in semiconductors, which is what I'm responsible for. Also automotive technologies we have nine automotive technology clients. Software clients , et cetera. So I'm focused on semiconductor. Basically the past decade helping start up and scale up Western  IP EDA  and fabless companies  to the China market, help them get their first licensing deals here. And then also help them move on and stop the full-time operations.  You can find me on LinkedIn or on Twitter. I think it's at Stu Randall.  

ART: I'm fortunate to have a couple of great guests like you guys on to discuss the China semiconductor industry, which is a bit misunderstood, but often in the news these days. And we'll get into that more and more why in a bit first I wanted to give people who may not have as much of the technical background a one minute overview of the value chain and a couple of the big issues that I think are going to come up in our discussion. Basically at a very high level, putting together a chip, a semi-conductor involves a lot of design work on the front end before it finally gets manufactured at a fab. On the very front end, there is electric design automation (EDA), which is the software tools that help you do the layout of the most efficient way of laying out your design for a semiconductor chip. Basically lining up at a very microscopic level, lots of transistors. 

There is in between design work that at what are often called the fabulous design houses, which take these EDA design tools and use them to make the designs. They don't often own their own manufacturing capability. They have those designs of four chips  manufactured at places like TSMC, which is Taiwan Semiconductor Manufacturing Corporation, or SMIC here in Shanghai, these are foundries, which manufacture chips of other companies’ designs, which then eventually go into a device, integrated into a laptop, phone, , uh, and now more and more automobiles. A couple of key points that are going to come up a lot. Probably most people out there have heard of what's called Moore's law. Moore's law is now really  hitting its limit. The physical limits of how many transistors you can squeeze into a finite space. So as our devices become smaller and smaller, this becomes more and more of an issue. 

And getting the chips smaller and smaller in size is requiring more and more capital for the foundries to manufacture.  Another last issue that I want to touch on before we dive into it is these different parts of the value chain – EDA, design houses, manufacturing very much seamlessly work together. They're all trying to optimize size, speed and power consumption. And these different specialists in the value chain, again, need to work very seamlessly on different process nodes. We'll get into that a little bit more as well. So when we're looking at how is China developing these different components of the value chain, it's really a whole ecosystem that they have to develop. OK without having stolen too much of your guys' thunder there let's get into  some of the specific topics today. 

So I wanted to start a bit with both of you guys and I'm going to divide up the value chain a bit and you can add to it as well. I thought we'd start with Stewart to walk us through a little bit about . Some of the foundriesand the fab part of the game here in China, SMIC and others where they're at vis-a-vis, let's say TSMC and some of the other, the others of the world and how much resources the China's putting into developing them. And Nalin, I think I'll look to you for a little bit more of what goes in beforehand, EDA fabless  design and so forth. Let's give a, just a high level assessment of China's capabilities here. So Stewart, if you want to start with the fabs, with the manufacturing, please take it away. 

STEWART Yeah, sure. I mean you could disagree, but I think it's probably the biggest choke point for China is the manufacturing side. I think they understand that they have currently, I think 20 plus fabs on the construction. And 50 or 60 operating right now in China. Obviously they're not all logic, some of memory and some other discrete components and stuff, but that they have that right now. And that sounds quite impressive. And so I don't want to take anything out away from China. They are investing in this but really that, where they struggle is I guess, on the equipment side. So they do have companies at least to have actually an equipment down to seven or five nanometers. But I'm not sure if that's really true. It seems to be more marketing material. 

And as everyone knows in the group, the space, especially UV there's not really anything we have there. So when it comes to in the 20 nanometer or above they’re doing okay, below that they're struggling. I think a lot of this obviously comes down to their dependence and independence, push their self-reliance push. And right now I think we can say it's been a bit of a failure. The last five years  I think from 2015 to 2020, that goal was to be, I think, around 40% self-sufficient which basically meant that the definition of self-sufficient is that it's made in China. So it could be a Samsung fab or a TSMC fab in China. But as long as it in China, they're not counted towards not important. Yeah. And then those five years it's moved from something like 15% made in China to 16%, so way off therefore 40% target. And so I struggle. I can't see how they can meet the 20, 25, 70% goal seems impossible given the current situation. And then obviously we come to the amount of money going into it. 

So traditionally, I don't know if we'll get into this later, but the big funds and government investment has gone into just building capacity mainly. And this probably needs to go more into this mainly because they're trying to get these arbitrary goals. So capacity is with money and it's not really helping. One of the big reasons is because they're importing more and more as well. So that percentage doesn't move much, even though they're expanding capacity , locally and the money as well as A bit too distributed. Even though there's a lot going in, it's not enough to each company , as far as I can see , compared to the amount of capital expenditure, likes of TSMC and Samsung are doing , in R &D spend as well are much lower than its competitors in comparison. I think this is probably the area where China needs to focus the most. It's obviously focusing on building capacity. Although I will say that SMIC in 2021 is actually reducing its capital expenditure. It's one of the only pure play fabs to actually the reducing it's expenditure and everyone else. 

I think other than maybe Micron, I think, everyone else is expanding as TSMC yesterday announcing it's going to be spending a hundred billion over the next three years. I wasn't sure if it was April fool's joke for a second, but I thought 28 billion this year, so yeah, it could make sense. So yeah, a hundred billion compared to SMIC I think is going to be something like 4 billion this year, so it doesn't really compare. And as I mentioned, R&D spend as well, you compare TSMC to SMIC, TSMC’s R&D spend is more or less the same as SMIC's entire revenue per year. So there is still a big gap and that with the latest announcements from TSMC and Samsung, it seems like it's growing. I don't see how, in the short term, at least how China is going to change that, although that they obviously know about it. And I think they maybe need to change how they manage these investments. And also though your point about the raw numbers is there's the raw amount of manufacturing going on and how much , you also have. 

The most advanced nodes, that's really where it's at. And you could be making everything at, 28 and above. But if you're not getting, if you're not continually working your way down and catching up with TSMC who is still getting to five, three, and so forth, then you're perpetually in catch up mode. And you can't help guys like HiSilicon and Huawei and it's a moving target. So if we're going to have seven nanometer capabilities in, in four years time, let's say that others are down to two or something by then.  Eventually we're probably going to hit a natural physical limit. People can't go further. I don't know. Maybe you go into quantum computing realm or something. Yeah, I guess we will hit that point. In which case there'll be time to catch up, but it's not like within this decade.  

ART:  And Nolan as Stewart was saying, you can have all the capacity in the world, but if you're known have the other parts of the chain, if you don't have the design, if you don't have the tools, if you, as we'll get into later, if you suddenly get cut off from access to those things , it doesn't matter how much capacity you have. So how about everything that goes into it before then?  

NALIN: True, absolutely true. So obviously there's a big gap to fill when it comes to essentially the EDA tools. Everything from front end to backend that you need to tape out a specific design. And the big three, or I don't know if it's the big two now, but either way… there's a long way to go before that can be replicated, what they’re offering today, in China. I do think it's less of a barrier compared to say the manufacturing side. So the fabs are one aspect, but then you also have semiconductor manufacturing equipment like Applied materials, ASMO, that provide fabs the equipment to make these chips. That's another huge barrier that needs Capex and R&D investment and in that China's got a long way to go as well. I think the EDA tools gap potentially can be filled, but over a long period of time, you can potentially envision folks from Silicon Valley and other areas in the US even returning to China to build up the EDA tools aspect of it. But that's still a while to go. 

I think you're looking at a big, probably a good timeframe of four to five years in my view, before they can get to acceptable levels of having EDA tools that can do these sort of designs. The one area that I think China is fairly strong in is the design aspect itself. So Chinese companies have been designing chips and multinational semiconductor companies have been in China for a long time designing chips. There are fabless design companies like Verisilicon, Allchip, and so on and so forth. So I think that aspect is fairly strong in China, even though they might not have been doing the five nanometer type designs. But I think that experience can go into design aspects as well.  

ART: So the bottom line really is where the integration happens, right? Because the tools have to integrate with everything and the designs are, more about raw experience but the tools have to integrate seamlessly all the way through then. So when we were at Cadence we were intimately cooperating, and this is no secret, with TSMC, right? On a new process node coming out, from the very beginning.  Stewart, you were going to jump in there? 

STEWART Yeah. And may I just say personally, I don't think I've met any fabless company in China ever that doesn't say they're using some Synopsys or Cadence design flow. Never heard of anything else. I've probably met like a hundred different companies and I think I've never heard anything else. So it's probably Nalin, your good work. Sending them Cadence tools or Palladium emulators.  

NALIN It's not it's my former colleagues, but I’ll take credit for that.  

ART I certainly can't take any credit for that. I was the guy they all hated the guy that had to approve all their contracts.  

STEWART So then I'd say market share wise, as far as I know though the big three are even bigger in China than they are globally. I think it'd be like maybe 70% and in China over 90% of , the market is , those three guys. And I think, yeah, I think your point on. Design is right. I think China has capable engineers. I think they're like the HiSilicon proof that chips were on a par of Qualcomm. And they got design chips down to 7 nanometers. And I think if I say the things they're doing, I find that for designs as well. Although they, had a very short lead time at the in the top 10 of semiconductor companies.  I don't know if this is putting you on the spot Nalin, but I just recently I've been looking at a company called X Epic here in China. They’re trying to get into the verification tool space and they seem to have a roadmap to have an entire design flow. And the management is all ex-Synopsys and Cadence. I'm keeping an eye on them. I've noticed this year, they're expanding their marketing presence and stuff. So I don't know if you've heard of them. 

NALIN I haven't heard of them, but I'm not surprised. Just like I mentioned, I do anticipate a lot of the ex-Cadence Synopsis, Silicon Valley folks, potentially turning to China, to bootstrap this type of effort. I think. That's a shorter sort of leap compared to the capex required on the manufacturing side and the equipment expertise that they need to build up. 

ART I know with a lot of ex-colleagues and again, that's not an open secret. Almost every day I hear of an ex-Cadence colleague, who's setting up his own EDA company. Because if you think about at the end of the day it's software, right? So software has lots of bugs. So hat's a lag in the generations of bugs that have been fixed, but that seems like something where you can, in some ways, throw more resources at it and it will make a difference. Whereas other things we're talking about not necessarily. And you have the application engineers, right? Those guys still have all the experience from working at Cadence or Synopsys. And those guys don't forget all that knowledge that they have. And even if it's a different tool set, that's where the real value is we always said right at a Cadence that our tools and Synopsys or whatever, we're pretty much interchangeable. But the real difference was when the bugs would inevitably pop up, you'd have someone immediately there. So we had, Cadence, we had what, 500 engineers in China. So the expertise is here. There's plenty of people who know how to do this. It's just, they need the tools on the very front end and they need the place to manufacture on the backend. 

STEWART I guess it's a marketing issue as well. Like they are so dominant these companies that it's hard even if they have a good tool, someone at university already learned the Synopsys or Cadence tools and then their first job, they learned all that. And then you have to teach them a new thing, maybe they can, more or less copy some things, but maybe someone like you, Art might get involved if they did that. But they can make it in a way that it's easy for them to learn this new tool or this new flow. But still I think any kind of Chinese EDA company, maybe they can gain within China. But growing outside of China and having the scale and R&D investment that the big guys do. It's still difficult. So I think maybe it's less of a technical problem and more of a market problem.  

ART So the elephant in the room of course, is the whole export control issue, which we are going to get to in a second. But before it gets to that, China has this clear policy of becoming, like you said, self-sufficient as much as possible. And my question is that really possible because as we can see, we've already mentioned several companies, TSMC in Taiwan, Cadence, Mentor, and Synopsys based out of the US or the Mentor is now part of Siemens group. And you have ASML in the Netherlands. All of these guys are scattered around the world and you have lots of companies in Japan, Samsung in Korea.  Is it even possible for… I mean I don't think even the US could do that. Have its own a completely self-sufficient semiconductor industry. I guess it begs the question. Is it even worth trying or is it as much as possible, still better than not trying at all?  

STEWART Yeah. I think if your goal is just to be self-sufficient to a level, that means you're protected in some way from being caught from stuff, not necessarily we're self-sufficient and being the best in the world. Then I guess that makes sense for a country like China that may feel threatened and wants to make sure that if it's cut off, it can still do some things, but I don't think you're going to be doing leading edge. A company needs both to source the best technology globally and it's going to be very weird situation where all the best possible IP, best possible tools, the best possible processes all are in one country. I don't think that's ever going to happen. If some country gets some new thing, then you're not, you don't have access to that. And then you're not going to be the leading edge. 

So I think it makes sense if it's like a kind of backup situation. We're going to use everything globally, but we just want to make sure we have some, something to fall back on. I think that makes sense. And I think it's to a certain extent, so you can be self-sufficient. I think the traditional IOT type applications are they self sufficient in those, so maybe even mobile over time. But if you're looking at something like a cloud data center, app applications , other relationships, I think that becomes difficult. Yeah. Yeah. I think, Oh, sorry. I have to remember. I think that definition is just manufactured in China rather than we have to be the best that EDA tool or was the best in equipment. 

But I think if, yeah, I think as we've already said, if they want to be truly self-sufficient, the main issue is the equipment side of things, which is probably the most difficult for them. But if we're just talking about falling back on stuff and having something, then they do have something, just not leading edge.  So we touched on it, even jumping all in, if you want to say no, I was just saying, yeah, if you look at it, even the us as a whole, in terms of , fabrication, some are going to cause application, the US accounts for probably, I believe about 12% of semiconductor fabrication today, which is Intel basically. In terms of trying to reboot their Intel Foundry initiative. You probably saw the announcement and such, I think they want to put in US$20 billion in Arizona and so on. But the way for that before I think it was before it didn't work because they just didn't have the key customers. They didn't have the big ones, they didn't have the Apples, the Qualcomms , etc. So I think , TSMC has that as a captive customer base today.  

ART Yeah. So we've touched on it a bit before, but I think it's worth spending a minute on the money part of the equation because, as we know we can see like anecdotally with again, the ex-Cadence folks leaving and easily starting up their own companies. Part of that is subsidies that get from the government. We mentioned, I think Stewart, you mentioned the big fund already, the National Integrated Circuit Industry Investment Fund, which is a massive fund. It's been around for more than a few years now. Putting a lots of money into different parts of the semiconductor industry, but again, a lot into manufacturing. There are local subsidies around someone estimated the 300 billion RMB and total subsidies out there. And we know I'll just throw another stat out there. We know recently it's been , said that the VC, the venture capital world has tripled its investment year over year in the semiconductor space. 

We know the STAR Market, recently launched in China, NASDAQ style. You don't have to be profitable to list, which is great for these startup companies. And so last year 32 chip companies went public on that market. So it's a great source of capital. Some even like a SMIC came back to list in China, because the valuations were so good. And because they were feeling a little unwelcomed in the US. We're actually going to have a separate podcast about that.  Now I think my question is this money well spent, you hit on a little bit already Stewart. Is this money being spent in the right direction? And one more question I'll throw in is, is it actually counterproductive in some ways to throw so much, especially government money into this industry where you're distorting winners and losers. So that's a wide open for you guys.  

STEWART  I'd say some money's better than no money at tool. So it's not entirely terrible. And I think maybe the media talks about the big funds and this VC investment a lot, but I think compared to the amount of money that's being spent by Western firms in R&D , and capital expenditure is actually less, that, so in some way you could say maybe it's not enough. I think, yeah, I think it's not necessary not really put in the right areas. It's spread around a lot. It's often local governments that may decide where it goes and they don't necessarily understand the industry. And maybe it just goes to their friend that stuck semiconductor in his company name. I think RDA, there was like some fish companies and, I think a cement company that just got semiconductor funding, because it said it was going to do that. And obviously there's the big HSMC thing, which is the most famous one recently. 

So no, it's not always spent well, and actually, I was speaking to people. I'm not, obviously I'm not going to name names, they will speak to people here in China. They, that business goal is to get government money and they don't actually, some companies are just newly set up. They know someone in government and they just want to get some money and then run away.  So anecdotally it feels like a lot of it is wasted in that way. I do fear that I don't go to me for it and advice on this type of thing. I'm not a finance major, but I feel like a lot of new VC's in this area. In China, think they're going to get the kind of fast returns they get in other industries. And they don't have an idea of the amount of upfront costs and time and patients that they need in the semiconductor industry. So I think a lot of maybe these new AI startups may find the investors getting quite angry with them. If they're not selling millions of chips all the time. Just list on one of the local markets right in, and then cash out. Yeah. And then, that , then that's another issue. 

I think in China , maybe I'll, more about this, but I think in China, a lot of the investment is just like most, a lot of it is just like the likes of you of me. It's just retail and they don't really know. And then they're not informed enough. And I think if there, if all of that information just comes from reading local news, then they'll think every company is amazing and China's winning all the semiconductor stuff. And then that could be, it could be dangerous. Yeah, absolutely. So those are the retail investors at the end of the day, end up holding the empty bag. I don't think they have much choice as a country. I think they have to invest, I think globally, this is happening globally in the US in Europe. They're all trying to figure out their supply chains, given what happened with COVID and so on and so forth. Biden just , proposed a 2 trillion infrastructure bill, which actually has your mob about 50 billion in semi-conductor investment fab investments. That will be some wastage and startle spending obviously, but I think you'll see, it's not just China, but globally, every country is thinking about how to shore up your supply chains and become more secure on that front. There's going to be naturally some wastage on that.  

ART Yeah, I think it's the government wasting government money is definitely not a uniquely Chinese problem, but but there are some , I think this experience and yet it is, there's a lot of distorting factors. I do want to go back to that question real quick. Do you think it's bad? Do you think in some ways, I guess maybe there's no alternative, right? Do you think too much money gets thrown at companies and they go on a hiring spree taking all the best engineers away?  

STEWART Yeah I'm going to say it's better than doing nothing, but I think you could be done in a more productive way. As I say, I think it's, have a large chunk of money, but it's just diluted too much. I think now there's I have got 10,000 semiconductor companies. And they're all fighting for money. It ever if ever everyone gets a bit, but not enough. And also, I think maybe companies relying too much on kind of cheap funding can become a bit lazy or happy with that little segment of the market. And maybe they waste it on poor investments like maybe unique group has been done. Yeah I think that if I can give the Chinese government advice on how they should do things, but maybe it needs to stay away from the local governments that, and allocate it properly, because I'm not sure they can necessarily tell, which is the legit something that the business or which is strategically important. And as I said, they may sometimes not the case, it trends. So if it was maybe , allocated and from people that maybe an MIIT that, understood who is strategic, who knows what they're doing , then it could, yeah, it could be far more useful for me.  

ART Let's move over a little bit to the talent side of the equation. Because I think we've touched on this a little bit. In my limited experience, there's no shortage of engineers. There's no shortage of double EE majors, but there are as such, maybe a shortage of experience. I don't know if it's, whether it's tapeout experience at advanced nodes or it's guys who have, who've seen it all at the most advanced processes and so forth.  It also probably a finite number of engineers that there that can and will willing to come over here. So is this a bottleneck we see as well on the talent side even if money is not an issue, there's not enough people with the experience to build up these companies fast enough, maybe Nalan, you might be able to speak a little to that?  

NALIN: I think definitely on the fabrication and manufacturing side, yes, there is. There is a dearth of experience in the advanced nodes, right on the design side. There's a huge leap when you're looking at something like 5 nanometer or 7 nanometer type fabrication, and the talent pool and the experience around that. 

ART: I was talking to an ex-Cadence colleague who shall not be named last week and he was telling me that he's worked with companies in China where the engineers are 30, 31, 32 and they've done multiple tape outs under their belt. And he's just blown away by that. So it's not there's no raw talent and it's maybe that'll catch up fast, but yeah, sorry Stewart you were going to jump in there.  

STEWART I agree. Definitely on the design side, there's talent here. I think just one issue that is touched upon earlier with just the amount of companies. Sometimes that talent is maybe spread out too much. The best guys are not all in this let's say 10 companies they're like one or two in many companies. Maybe help them if they had more of a consolidation in that sense. And also, I'm just wondering, if I don't know, then maybe I've been reading recently about AI optimization, EDA tools, especially in the place in our area. I'm wondering if there's a goal to move that into verification and then how that would maybe solve any talent problems where you wouldn't need as many engineers because so much of it is automated. Would that be something that may help China in the coming decade?  

NALIN Yeah, I think that is actually an interesting point. And we're also already seeing machine learning, being used in various aspects of the EDA design tool. And optimizations and such. So yes, I do see it creeping into things like verification as well remains to be seen how fast it happens. But yeah, I think that can put us close the gap as well. 

STEWART On the manufacturing side. Yeah, I think that actually, even on the, in, on the design side, I think China, obviously looking to import more talent. I think recently in , Taiwan, I think there was a couple of companies that got shut down by the government there because , they were approaching cloud engineers, which is illegal to China, which is illegal in Taiwan. I think that was a Bitmain. A lot of that would be on the design side. And on the manufacturing side , yeah, for sure, like SMIC has, it's like Taiwan town, been there for a while now. It was just obviously to, obviously it's been twice engineers I'm from Korea as well, Koreans, but maybe we'll see, that's not me. The HSMC will prove that even top Taiwanese guys, what's his name? Even people like him , can be fooled.  Should I join X, Y, Z semiconductor startup in China, even if that doubling my salary?  I guess that's a question that could be a thing. That's how I wanted to make me ask him, especially when now, as we said, how much money TSMC is spending on expanding then there's obviously there's a drop off to opportunities opening there as well. 

ART We've danced around it a bit but again, the elephant in the room is when, ZTE was sanctioned for the sales to Iran and so forth. And then obviously there were rumors that , Huawei and HiSilicon its captive design house were next to be under the limelight. People doubted it and it happened.  And then finally Huawei was first cut off from buying components, US components in 2019. And last year it was cut off from access to TSMC, which is the really the big deal. And what alternatives are there? Because I don't think this is going to go away. And even if it does go away, even if it gets a little more relaxed under Biden, I think he approved the sale of the finally the delivery of the ASLM equipment to SMIC?   So maybe there's a little bit relaxing and practice of the restrictions, but this there's going to be this guillotine that's just going to be always sitting there waiting to cut off access at any moment. And that's obviously not a secret and that's an intolerable situation to be in from China's perspective. What alternatives do they have? There's rumor or not rumors there's Discussions that other open source options like is at RISC-V are possibilities. AI chips might make me maybe an alternative in some way, shape or form that e-commerce companies are working on.  

NALIN I think Alibaba made a pretty big investment in RISC-V high end RISC-V design, for the cloud computing chips and such. So that's pretty interesting. I think there's about actually, maybe about a dozen plus companies in China now working on RISC-V designs, including Alibaba , TSR, sorry, Alibaba, Huawei and ZTE. So I think that's definitely an avenue. They're all going to be exploring on the fabrication side potentially. For example, working, building up a team that works with Samsung Foundry, it's an option for HiSilicon. 

ART I feel like the US is going to be coming on a knocking on Samsung’s door pretty soon. Remember that defense treaty we have?  

STEWART Especially at a high standard. I think recent report from TSMC shows now the HiSilicon is officially 0% of their revenue. So that means like it's finally what HiSilicon can’t do anything at TSMC. And by that, it must also not be able to do anything at Samsung that it can't do anything in that kind of 7, 5, 3 nanometer spacing. Now on the EDA tool side with them , I not, and maybe you can say some of this, I don't know. But , every time I spoken to him, it's a way it's yeah. Or speaking to Cadence’s people it's not as appealing. Yeah, we can't talk to them. There's no support, but they have licenses. All the licenses, which they can still use, but we're not updating what they have. So how much, if you're in that situation, how much does that affect the design? And I guess if you're not, if your tool hasn't been updated then, which I think happens more on like maybe a weekly basis , then you can't a new process out, then you can't design to that.  

NALIN Yeah. On some, a tools perspective, the constantly updating tools for bugs and other aspects around it. And that's just a natural art of the sort of design flow. And yeah, that's actually is a huge issue, I think. Yeah.  

ART Yeah. A natural part of the business model of EDA companies. 

STEWART And I wondered, I'm thinking of asking questions now, but I'm wondering about the use of unlicensed tools or illegal tools in China. Because often when we've worked with design service clients in foreign designs, I see design service companies and we're talking to the Chinese company and they can't prove that they have an officially licensed Cadence or Synopsys tool. We just to stop the conversation. Yeah. How common is this for you?  

ART Good for you Stewart, Good for you. 

STEWART We have to stop the conversation, but I'm not naming names, but it's something I've come across at times that they just don't have an official license. Is that a common thing?  

ART That was that situation existing pre export control rules, that's been a longstanding issue and also an open secret. But I would say the one thing about that is that it's the support, right? It doesn't matter how many unlicensed seats copies you have, if you don't have the support, it doesn't matter.  Nalin, right?  

NALIN Yeah. If you're doing all these advances and stuff, I think you're pretty much dead in the water if you're not, if you don't have the support that goes along with it. 

ART Yeah. Had our Cadence engineers at the customer sites, right? 24/7, fixing things, how you optimize this and that. You need to do XYZ things to make the tool do this, to do what you want to do. And our guys really understand it because it's their life. 

STEWART Yeah. So I guess the answer is HiSilicon now is in a position where I can't really design as needed. 

ART What does that mean though? For the, what the is, cause that's the thing, does that mean it, from what I understand, it doesn't impact or maybe there's a way they can catch up on the , telecoms equipment going up to 5g. But the, but the handsets and all that, the, that's already impacted. 

STEWART Yeah. see, they seem to be moving some more of a hallway seems to moving and changing its business model to be more software focused. And you can do lots of optimizations on things like base stations for the software, for the hot button. But right now, if they can't design their own chips, then going to have to rely on chips from other companies. 

NALIN I could see a situation where the U S doesn't mind something Qualcomm chips or whatever. Fine. That's great. As long as you can't make your own and maybe that's good, but, yeah, on the chip side, there's also an architectural issue. So if you look at like this five might work out for things like mobile and IOT applications, but then think about applications that require sort of the x86 type processes. Like the Intel AMD does not pay an alternative for that either. Even if you are, even if you do use RISC-V, You still have to, and you're doing, you still need to figure it out a fab that can tape it out in that advanced notice. Yeah. Yeah. So I think that yes, yesterday day before arm were saying that a new B nine potential they could license to highway and Chinese companies, if they want to. 

STEWART It's not, there's no US origin technology. So in that sense that, they could use out EDA tools, make something benign or RISC-V and then take it out somewhere that is willing to work with them. And it won't be anywhere , super advanced.  Just on RISC-V though. I just I think I say a few things because, I think I first started helping , British IP company here. And it was in the RISC-V space maybe five years ago. And at that time when I was meeting that's not this company is that nobody has heard of RISC-V. And the people that had heard of it, I would just refer to it as RISK Vee all the time. That really realized what it was as people, they just read something about it. But today I think basically every fabless company in China is designing something with RISC-V or they're considering it, or they have actually gone completely all in on it and that's all they're doing. So everybody's looking at it and a lot of actually doing something with it. And yeah, you mentioned Alibaba. I think. There's a few other Risk-V core companies we see quite a lot here in China as well really growing here in China. I think it's, yeah, and it's just a threat in some ways. Especially if ARM definitely gets acquired by Nvidia.  I'm not sure if we're going to get into it. 

ART That's the final topic we're going to get into.  

STEWART Okay. Because that does happen. And RISC-V is definitely going to have, even if there's two RISC-V associations in China now.  So China has companies, I think C sky, which was acquired by Alibaba. And they'll say China's qhich I've tried to do that own instructions, that architectures and be like, like the ARM of China. And they basically failed at that. There's no way they could globalize it as an ecosystem.  But RISC-V has given them a platform. It's a global platform globally is building the bit system and it's open. And so people around the world can say, we can't trust you because you're from China, there's this playbook. So it gives them a platform to license their commercial deliverables globally and not just be stuck in this China.  

ART So speaking of the China market, that is the last topic I want to hit on. And you touched on it a bit already with the Nvidia arm proposed deal. And yet we don't really know what's finally going to happen with that. But one of the points there of leverage was antitrust review for China, right? China is a huge market. All of these companies, we mentioned Qualcomm Nvidia arm. Et cetera, cannot ignore the Chinese market or they do so at their own peril or certainly to their bottom line. Is that any, is there any leverage there for China to use the size of its market? Which of course is growing. This is where the future is for semiconductor growth. Of course it's growing everywhere, but this is where it is. Growing and there's a lot of foreign semiconductor companies that have R&D centers here. There are joint ventures, ARM, of course, there's a whole lot of drama with that one, but we, can, I'm just gonna throw a bunch of things out there, let you guys take it away. Can they have there are their own industry standards that they can set here or it's just, everything is too much already written in stone. And that's not even feasible. And then be anti-trust. Is there anything that can, they can do there? So what, how can China use the size of its market potentially as leverage here? Not an easy question. Maybe too big of a thing.  

STEWART I think we, we saw recently, even this week that they did that with Applied Materials and Coco say, I don't speak Japanese. I can understand from China's standpoint right now, like why would we want. another equipment company to be American like that, just avoid that just as little as possible becoming American, but then you could get stuck in a situation where no one can acquire anyone because China isn't anything the U S doesn't let Chinese deals go for it. And then US is already not letting any Chinese deals go down and we'll do the same. And then what happens?  

NALIN Yeah. Obviously they let through NVIDIA-Mellanox, which was a bi  surprising at the time. But since then yeah, I don't think there's been anything of note. They also allowed the whole Marvel Infi do you go to right? And I think that's because essentially China needed Marvel's 5G and cloud data center ICs in a sense. 

STEWART Yeah. Yeah, I don't see why Jensen at Nvidia is so confident that this is fine. The UK government has been looking at it. Chinese government looking at it. I think lots of ARM customers are angry about it. Cause it's essentially buying from your competitor for a lot of them. I can see what might make sense for Nvidia, but for everyone else, it doesn't really make sense. Yeah, let's wait for that to fail. And then Nvidia shares will drop and then we're in there.  

ART Maybe it drives everyone to other standard under other standards, right? Like RISC-V. Perversely they have incentive to improve as that way.  

STEWART Yeah, it would help RISC-V for sure. Yeah. Yeah, that's a good point. Yeah. That could also be a way of looking at it. But I think with V nine architecture coming up, I think China wants to make sure that isn't American as well. Think it would help. I think in the press release for that, I think OPO Charmian vivo, and I think if you probably Chinese companies had quotes, so they're obviously all going to buy it , or license it, maybe architectural license , So I think to answer your question, I think, yeah, I think China can use this. I think maybe Nvidia, maybe also has some cards that can play. Cause it's huge market in China, the cloud companies rely on their GPU. I think there's probably some negotiation that Nvidia can have with the government here. But yeah, the market's huge and, as we saw, even this week that. No, they can just block things and, yeah, I'm sure they can get some concessions to that things for leverage 

NALIN No, I agree. And if you look at some of these key sectors, like the automotive sector, China's a key area, even for us suppliers into the automotive, right. Can be very interesting for them. I think at the end of the day, the U S want to figure out a way to do business with China, but just make it a little more on an even standing plan, if possible. 

ART Yeah. Yeah. There's so that, that's traditionally, I I haven't been in China for 14 years. That's traditionally how things get changed, right? No one knows that the U S government is not going to change its stance on national security regulations on MNA and export control. Just because China doesn't like it, they're going to change it because US businesses are lobbying them. To not do it right? Or not implemented to the letter of the law or give a lot of licenses, exemptions and so forth. So that's where the that's where the devil's in the details and the rubber hits the road, so to speak. But and that's traditionally how both sides, that's how that's lobbying.  Very good. Hey, this guys, this was so much fun. Thanks so much for joining. Really appreciate it. Great stuff. And I know the audience is going to get a lot out of this one. Real good deep dive on something that's not covered well, except when Stewart is quoted in the press in various publications.  So thanks for coming on guys. Thanks guys. Sure.

All Posts
×

Almost done…

We just sent you an email. Please click the link in the email to confirm your subscription!

OK