 And we're live. Hi. Hi, Nicholas. This is Nathan Brookwood. I'm the founder, principal analyst, and research fellow. A lot of big, long title. At Insight 64, where for the last 22 years, I've been following the semiconductor industry with a little bit of a focus on processors, but I actually look at platforms and systems as well. So I have a pretty broad range of things that I'm trying to follow because I get bored easily. But there's so much stuff happening in this industry. And there's so many huge stories, I think, like Huawei, and that's a huge thing. And the US is trying to get the sovereignty back with making the chipsets in the US. Do you think that that's not like a piece of cake? How are they going to even try to do that? I think that's a lost cause, actually. Intel, of course, still maintains pretty close to state-of-the-art fabs in the US. Global Foundries has a large fab in the US, but they gave up on trying to pursue the really state-of-the-art 5 nanometer, 3 nanometer, and so forth. They're really focusing on niche markets. And so there isn't much of an infrastructure other than for Intel supporting semiconductor manufacturing in the US. And these fabs are now costing close to $10 billion to put up. So when somebody says, here, take $20 billion and give me an infrastructure, they're not going to get much of an infrastructure with that. And I think at this point it really makes more sense for us to make sure that the supply channels, the supply chains are clear and that we have the right relationships, particularly with TSMC and Samsung, because I don't see an easy way for the US to catch up. TSMC is like, as a monster, is huge, right? It's huge. And they grow up a new fab in about 12 months from groundbreaking to production. Intel takes 24 to 30 months to do that. And so for TSMC, they have focused strictly on semiconductor manufacturing. They've got a good name for that, right? And they are, you know, in semiconductors, volume is everything. If you have the highest volume production, that means you're going to have the lowest costs and you're going to be able to improve and go to the next process technology faster than your competitors. And if you go back 10 years ago before smartphones were really a big thing, Intel had the largest manufacturing volume and that gave them the ability to lead in terms of process technology. Nowadays, TSMC, because of their relationship with companies like Apple and Qualcomm, has much greater volumes than Intel, and that gives them the ability to bring up a new process every year. You know, the iPhone, the chips that TSMC makes for iPhones and tablets are on five anatomy now. And that's awesome technology. It's going to be pretty hard for anybody to catch up with them. It's especially hard right now because the industry is undergoing a transition in terms of the lithography technology to use for the smallest features, going to something we call extreme ultraviolet, which helps in getting five nanometer transistors placed on the chips. And they've now been doing that for over a year with some of their processes. Samsung is using EUV for some of its most precise imaging. And Intel has a bunch of these devices that are in a development lab in Oregon but aren't being used for high volume production. And until you get into high volume production, it's really hard to make sure that what you're doing works. So given that TSMC is so far ahead of Intel, I find it problematic or challenging, I guess is a better word, that Intel is going to be able to catch up. There's even a story that I think is interesting is that the EU is in talks and also the EU is panicking a little bit and saying, hey, everybody is like going into all these trade wars of the last four or five years and blocking who gets chips and who gets what. You know, like Taiwan is not really supposed to be part of China, right? But I would think that China has some kind of influence in terms of telling them what to do also. They shouldn't just supply the Americans when the Americans are blocking the Chinese. Something like that. Well, I think that all of these geopolitical considerations are becoming more and more worrisome. You know, when China took over Hong Kong back in 97, they were very proudly proclaimed one country, two systems to cover the fact that they were going to keep the democracy aspects in Hong Kong going. But we've seen how they've been clamping down a little bit on that over the last five years and that's worrisome. And they also, mainland PRC still thinks that Taiwan is part of China and at some point there may be some sort of clash. On the other hand, the Chinese are very practical and they would not, I think, if they did intervene in Taiwan or want to kill the goose that's laying the golden eggs or the golden chips. And so that would probably continue on pretty much undisturbed. When I look here, it even shows that Intel is panicking that they're trying to get TSMC and Samsung to make some of their chips because maybe even Intel has some issues delivering to capacity. They've been struggling with the most advanced semiconductor technologies. And this is no secret. They were supposed to be on their 10 nanometer about three years ago. They're still not really there in volume. And in fact they have taken some of the chips that they had designed using their 10 nanometer design rules and back ported them to work on 14 nanometer processes. And when you do that, you give up the advantages of smaller denser transistors. And so that makes Intel less competitive. And I think we're going to see that more and more over the next year or two because AMD who threw in their lot with TSMC three years ago now has a real process technology advantage. Over Intel. And, you know, there's, you know, Pat Gelsinger, you know, Intel made a change in CEOs. They brought back Pat Gelsinger, who is clearly a very smart technical guy and an astute business person at this point after his seasoning at VMware. And we'll see what he can do. But it's not just the guy at the top. It's the whole infrastructure inside the company in terms of the technical talent in both design and manufacturing. Lots of which has slowed out of Intel over the last few years. And so Gelsinger is going to have to do a lot of rebuilding in order to get Intel back into shape. So here's, it says Intel's new CEO says trouble chip project is recovering. There's, there's a lot of articles about this new CEO. A lot of things to say. Do you mind trying to click next to the mute button? The little arrow and then turn off auto adjust and make volume auto adjust. Yeah, turn it off. Yeah, because I think it's, it's going up and down on your volume. How's that? Yeah. Okay. That's good. Now I think you're in a fixed volume and it's not going to mess around with it. What a deco cancellation. You can keep that on. That's good. Okay. Yeah. So what do you think about this news Intel CEO? Gelsing, excuse me. Gelsinger is the right guy at the right time. It was really a shame when he left Intel, but having been out of the company, I think he has a lot of regard inside the company, the employees like him. And having been outside, he now understands a little bit more about the challenges Intel faces competitively. You know, when you've been in the company as Pat had was for, for 20 years before he left, you sort of, you know, get into the group think about how Intel is, you know, the only company and nobody else can do anything right. Now that he's been outside for five years, I think he sees that there are a lot of other people who are doing some very clever things. And so he can now, when he repositions the company, I think it will become more competitive, but it's going to take a very long time. And as I said, I'm not sure that Intel trying to rebuild its manufacturing chops is the right thing. If I was in charge, that would be scary. If I was in charge, I would seriously try to find a way to unload the manufacturing capacity to some other third party or sell it off to people who need those tools. And rely entirely on TSMC, maybe a little bit on Samsung, rather than trying to do it in house. Because they've had issues getting down to the 10 nanometer, they've delayed it for three, four, five years. And then it's hopeless for them to get to seven nanometer. Well, I don't know that it's hopeless. But it's certainly challenging. And meanwhile, every year, TSMC moves the bar a little higher. And that's going to make it harder and harder for Intel to catch up. And that's why I just don't think it's a good idea to try. I am reminded of what happened to Intel in the mid 1980s, when they still had a lot of their business relying on the sale of memory chips. And the microprocessor had started to take off, but it still wasn't really making, couldn't support the company by itself. And Andy Grove and Bob Noyce looked at their strategic situation. They concluded that they couldn't compete with the Japanese in terms of memory production. And they decided to get out and put all of their efforts into microprocessors. Now, looking back, that was a very wise decision. Although at the time it seemed kind of gutsy because there was a whole division that was making memory chips who wasn't to be needed any longer. And I think channeling Andy Grove, if he came back today, he would look at Intel's situation and say, hey, why are we knocking our heads on this manufacturing wall? The game is no longer about manufacturing. The game is about designing and being innovative in terms of design. And we know how to do that pretty well. So why don't we just get rid of this boat anchor called manufacturing, go to a TSMC and become more competitive in terms of our products. That was something AMD did. And I think it's a path that Intel could follow. I would think that you quoted as an analyst source in this article talking about the M1 Mac. Maybe the issue is that they keep trying to do AX86. They should just give it up. Maybe it's easier to do a 10, 7, 5, 9-meter when it's ARM because it's a more simple architecture to fab. I don't buy that at all. An ARM chip today, if you look at what Apple is including in the M1 or any of the larger ARM chips, even some of the larger ARM server chips that are coming out from people like Ampere, those are very complicated chips. And the fact that it's an X86 versus an ARM has very little impact in terms of the ability to manufacture it. And that's why, for example, CSMC can manufacture, and the same factory probably, ARM based chips for Apple and X86 based chips for AMD. So it's not so much the instruction set architecture that makes it hard. It's all the other things that you need to do in a modern semiconductor foundry, being able to handle the analog components of a chip, the digital components, large integrated memories, and so forth. That's what makes these chips hard. And the fact that it's X86 or ARM is really in detail at that level. So the best in the world are TSMC and Samsung. How about Global Foundries? What's up with them? Well, Global Foundries decided about three or four years ago that it was too expensive to try and maintain their position on the leading edge with regard to semiconductor process nodes. And they were playing with EUV. They were working on smaller transistors and so forth. But they just didn't have the volume that a TSMC or Samsung would have. And therefore they decided to abandon the pursuit for leading edge technology in terms of geometries and focus instead on things where they felt there was a good market and where somebody who could deliver consistent quality for customers who didn't need that leading edge would be able to succeed commercially. And they therefore pursued their silicon on insulator technologies, SOI, and some of the other niche markets, communications being a big one, where the process geometries were not so crucial. And they've been able to make a business out of that and compete effectively really with companies like TSMC in those sections of the market. And they've done that with much lower investment requirements for tooling. But I mean, in the beginning of our video, we were talking about Joe Biden and his issues and EU and their issues. But don't Samsung have really, really good fab right now in Texas? Is it where they're making all their five nanometer stuff right now? Actually, Samsung is making most of the five nanometer stuff in Korea. The fab in Texas is focused primarily, I think, on digital logic, which is the SOCs that they make for smartphones and such. But they have, must be more than half a dozen fabs scattered around the Seoul area that are also on leading geometries, both for memories and for a system on chip and microprocessors. So Samsung has diversified a little bit in that regard, but they're still heavily concentrated in Korea. So we, the whole world, the planet, all the humans are at the mercy of Taiwan and Korea. Or is that what's happening? Or like if the China gets angry, they can start blocking what comes from TSMC? I hate to think that that could come to pass, but if it did come to pass, I don't know that we'd have a lot of capability to resist it. I mean, all these talks, all these articles are their hot air, what do you call it? Like it's not going to happen. There's no chance that they're going to make cutting edge fabs in the US and Europe. Yeah, I think that's, that's a bet I'd be willing to take. I don't know if Joe Biden is willing to take it, but I just don't see how the US can rebuild an infrastructure for leading edge process technology. It sounds very defeatist, I'm sorry. I saw, I saw your comment on LinkedIn where you, you, somebody was talking about the chipset supply shortages. And, and you said, hey, but isn't that like just a planned out years ahead? And why are somebody skipping the line or what's happening? Why are they didn't know they needed so many and that they only placed the order recently? That's why they're not getting them, right? That's my view. I know another analyst commented that he still thought it states it was helpful to be on TSMC's good side. But my understanding of the way TSMC works is you go to them and you say, we're going to need so many waiters, so many thousands of waiters per month on a particular process node, starting in June of X year. And TSMC, based on those forecasts, which are backed up in many cases by deposits, then goes and make sure that they have the capacity in place to deliver. And TSMC in many cases will then go and put up another fab and science bar or whatever, so that they can meet those kinds of requirements. And if you don't forecast accurately, if you under forecast your needs, you may very well find yourself not being able to get all the waiters that you want when the man picks up and it's over what you forecasted. That's sort of the way a system like this should work. You know, just as if you, when you had the manufacturing capabilities in house, then you would plan to have the right amount of capacity in your factory. And if for any reason you didn't have the right capacity, then you wouldn't be able to deliver as much product as maybe your customers wanted. And in fact, that's happening with Intel. Intel's had some product shortages over the last year because they didn't have the capacity in place, primarily because they still never got to the tenant enemy. And so they needed to make more stuff on 14. And that created Intel chip shortages, which created up more opportunities for AMD, but AMD couldn't get the waiters either from TSMC. And so people were then fighting over the chips, either for GPUs or CPUs. And you're seeing that the actual price erosion for many of those products has slowed. People are now willing to pay a less price in some cases, whereas before they needed or demanded and often got tremendously steep discounts. So in this, and again, hey, market's supposed to work this way, right? When there are shortages, prices go up or in the case of semiconductors, they don't go down as fast. And then, you know, eventually capacity is added, people change their forecasts, TSMC goes and builds another factory, and all of a sudden we have overflow and then prices come down. So here it says on this article, Anchil Sag is talking about the 37 billion down payment. It's just a down payment and it's not going to be enough. But what I'm thinking is, okay, 37 billion. It's a bunch of money, but it doesn't take years and years and years to get anything done in this. So it's not even going to be done by the first term of Joe Biden. No, definitely not, definitely not. And maybe the thing that maybe will nationalize the semiconductor industry and the government will buy these from Intel, I don't know. I think that's a good idea. That would be nice. Yeah, for Intel, let's assume they wanted to take my advice and get out of manufacturing. Then the question is, who's going to buy those factories? What are they going to do with them? And, you know, it's not easy to convert a fab that Intel built and they all look exactly the same. Every Intel fab on the inside is exactly the same as all the others. That's as a result of a philosophy they've had for decades called copy exactly. But then to change an Intel fab to a TSMC fab is would be a challenge because things are not laid out the same way. The flows are probably different. Some of the individual tools they use are different. And therefore it would be complicated to do. Now, maybe if you knew you were going to be converting the fabs in Israel and the fab in Ireland and the fab in Oregon and the fab in Arizona, then you might come up with a way to do that if say TSMC wanted to take those over and operate them jointly with Intel or whatever. You remember this video we did a few years ago? Yes, that was in the days when we had real meetings with real people. Yeah, that was in a pre-Noah's artwork, what's it called, the thing we're doing right now. But what about the topic of that video and it got 12,000 views was that Intel was going to make 14 nanometer arm processors. It was basically those Altera FPGAs at that time. So I've always thought Intel's main mistake over the last 15, 20 years is that they only play around on the side with ARM. They don't put their logo on the very big performance, the top level. They should put their logo on five nanometer arm processors and provide them as an alternative to the MacBook M1, right? And give all these Windows laptop makers an alternative to Qualcomm. Well, first of all, you'd have to have a functioning five nanometer process and they're a long way from that. Secondly, it's not clear that Intel can be successful selling non x86 processors. I've always said that the criteria for success of a product line at Intel is that it has to be x86 compatible. That's a sort of necessary but not sufficient requirement. And then it also has to be able to compete in the marketplace against all the other available options, including ARM, including AMD, including in some cases, power PC. And every time Intel has straight from the x86 at the base of the product line, it's gotten his hand caught in the ringer. Gitanium back in the 90s and ox, they had strong arm that they acquired from digital equipment and couldn't figure out how to make that product line successful. What else? They played with alternate architectures, the 960s, the 860s, and so forth. None of those achieved anything like the success or even the sustainability in terms of ongoing sales covering the cost of development that their x86 franchise has generated. And that's made it very hard for them to break out of that of that of that category. Yeah, the image we have here in the background is actually in a, I think it was an arm, an arm processor that Intel was doing in partnership with the Rockchip. A few years ago, there was a strange project that had going on. But if I get to your rate here, there's another huge thing happening. There's a giant giant 32 billion or $40 story that the Americans were going to take over the arm processors. I hope you're still still connected. I'm still connected. Oh, that's the idea. Yeah. Yeah. So yeah, your image kind of froze. Maybe can you click on stop video and then start video again. Right there. You see it. Yeah. Hopefully I can get you. I'll just try to do like this. Okay. Hopefully you'll be back. But I can hear you. Right. Okay. Yeah. Maybe maybe you can change the resolution to try to 720p or to the, or down to six 360p. I'm not sure why the camera is not coming through. Potentially 360 360. Oh, there you are. Okay. Cool. Okay. Yeah. You know, I'm a, my website is called arm devices.net. Your, your company is inside 64. So it means 64 bit arm. Right. I'm joking a little bit, but there's the arm is like so huge. It's insane how many billions and billions of arm processors are just shipping daily or not daily, but it's like huge. Yeah. It's like 400 arm based chips a second is the latest statistic I saw in terms, I think from arm arm releases that, that number every once in a while. And I remember three or four years ago, it was 60 arm chips a second, which they referred to as 60 Hertz. Yeah. And that was 400 a second. But there was this thing that happened the week after Brexit, which I thought was hilarious, funny or weird. Right there, they vote Brexit, the British. And then Japan soft bank comes in and buys arm at a discount kind of, because the British pound was hurt in that week or something that I don't know. It's like he kind of got 20% discount. I don't know if that's the reason or not, but I also think that the guys in arm were thinking, Oh my God, the whole thing is crashing. Let's get out of here. Because Brexit is a funny story also. But then now it's going over to the maybe the Americans, the Silicon Valley is trying to buy the arm. This is the Nvidia arm deal. Yes. And the other one where I've taken the position, not that anybody pays attention to my positions, that it's a really bad idea. And I am very, if I was the European Commission or the Federal Trade Commission or a Department of Justice, I would definitely not, or the Chinese people, all of them have to sort of approve these kind of mega deals. Not good. Don't do it. And we'll see whether or not it actually goes through. And my logic there is really very simple minded. Arm provides technology to a number of companies, hundreds, thousands of companies in the information technology arena. And I believe it needs to be like Switzerland, a neutral party who can sell to any of those companies without compromising what they do or what their other customers do. And when they were based in the UK and they really didn't make any hardships and sales, they just sold the IP to third parties. Everybody more or less had access to that technology and nobody had to worry about, hey, are the Arm guys tilting a little bit more toward Qualcomm or toward Samsung or whomever because they were not selling to anybody who competed directly with Arm. They were all customers. And if we have a situation where NVIDIA does acquire Arm, then if I was an Arm customer, I would be very skeptical that NVIDIA, that Arm was not telling NVIDIA what we were planning and allowing NVIDIA perhaps to counter that to their own product developments. And so we talk about Chinese walls, but I don't think those Chinese walls are impermeable. So in my mind, Arm really needs to be owned by a neutral third party. But as a European, I would have hoped that we would own the future, control what's called the destiny of everybody and it would just be in our... Because Europe has so little. We don't have the Googles, we don't have the Amazons, we don't have Facebook, we don't have anything, we're just consumers of this stuff. But we actually invented Linux, we invented the Arm, we invented the Web, all the stuff that everybody uses all the time and it's worth so much, but it's like we're giving it away for free. And even the Arm chips have always been on a very affordable little few cents per chip. I'm not saying it's not been profitable, it's been hugely profitable, but it's not like the Silicon Valley-style business model where you become multi-trillion-dollar companies. Well, but when you look at it, if you're looking at it that way, Google gives away most of their stuff for free and they've gotten pretty rich. So it's not an issue of Arm only getting a few pennies per chip, it's how many of those chips and when they're going out at the rate of 400 a second, the pennies kind of start flowing in at a scary rate. Basically, there are lots of different business models and from my perspective, Arm really needs to be an independent entity. If SoftPent doesn't want to own them anymore, oh gee, they need to find somebody who could own them and still be independent. For example, TSMC. TSMC and Arm, I could see being a marriage made in heaven because they both only provide technology to a range of companies and they don't compete with their customers. TSMC is very clear that it does not compete with any of its customers, unlike, say, Samsung who does. Samsung makes chips and they make phones and they make everything that uses the chips and so even people who want to use Samsung as a foundry are a little nervous because Samsung Foundry is not a completely different organization from Samsung, the microprocessor company, or Samsung, the cellphone company, or Samsung, the consumer electronics company. So you're basically saying that the Chinese should get Arm. I'm joking. Well, TSMC, which is not quite Chinese, or maybe the UN can take it over, who knows. I think we were just seeing that the EU is saying they have issues and they want to have some sovereignty and stuff and the EU should jump in and say, we will take it, we will take care of it or something. I don't know how, but because the bureaucracy of the EU is not really super effective, but they could put on the money and say, hey, just keep running in there of where you were. I know it's not going to work. There was Brexit, I'm sorry, I forgot. Yeah, so I'm not totally sure what's the solution. Yeah. But I mean, NVIDIA is awesome, NVIDIA is great, but it's just so weird when the US goes around and blocks all the Chinese over the last two, three years, blocks the biggest Chinese, Huawei, and then they're supposed to just take Arm and then block it even more. I don't think China is going to agree to any of that, not even remotely. There's no way can I defend what the US is doing with Huawei. I think that's a huge strategic mistake. And hopefully the Biden administration will be a little bit more clear headed in that regard, as opposed to the Trump group, which basically really was trying to stick it to China. Yeah, so there's another that you were on your LinkedIn talking about Pat arriving to his building. And right here I saw something that you're talking about, the benchmarks for AI. There's this really interesting chip company called NVIDIA that belongs to Qualcomm now, right? Yes. And it's maybe there's some other arm chipsets that are really focused on accelerating the AI for the cloud, for the supercomputers and everything. Well, I think the entire shift to AI has created new opportunities for companies who are trying to optimize products that are good at either inferencing or training. And what we're doing with artificial intelligence these days is just mind boggling. When I was at MIT in the 60s, I actually took a course on AI from Marvin Minsky, who was the father of all this stuff. And so I've been interested in that for a long time. But you know, until we came up with the deep neural networks and hardware optimized for the convolutions, AI really was not with more of a curiosity than it was meaningful technology for solving real world problems. And that's just completely changed in the last decade. And as that's changed, then that's created opportunities for the new vias and so forth to go ahead and build machines that have specific features that make the AI software run better and do more computations with less power in order to advance the state of the art. And that's going to keep going on. And I think we're just scratching the surface now in terms of what we can do with that. Again, as I say, mind boggling. And if I may put two things together, like there's AMD benchmarks, one of the things you liked on the LinkedIn, and it's just AMD, AMD, AMD, just way in front of Intel. So AMD has been having a good strategy in the last couple of years or? Oh, AMD's had a phenomenal strategy. And you know, it's one of these things where I think necessity is the mother of invention. So when AMD was planning its reentry into the high performance markets back in the 2015 year or 2014 era, they wanted to be able to build chips that could be used in servers as well as desktops and notebooks. And they didn't see an easy way to do that by creating lots of different chips. And so their first generation of the Zen generation, which came out I think in 2017, was designed so that they could use some of the chips that would go into their desktop and mobile processors and can put several of them together inside a single multi-chip package to create a server. And they really pioneered what today we regard as the chip-lip approach. Lots of little chips inside a package and so you lift up the cover and you see that it's not just one big chip, but lots of little chips. And in semiconductor manufacturing, you can make a lot of little chips to cover a given area of silicon a lot less expensively than one big chip that would provide the same number of transistors. And so they did that because they had to, but it turned out to be the right thing to do because it's, you put them on a development arc where they now have up to nine chips I think is the last count in their Rome processors to provide up to 64 processing cores and lots of IO in a single server chip where Intel to get that kind of capability needs to have one large chip which is much more expensive to manufacture and doesn't cover the entire range as effectively. So that's given AMD a huge advantage in terms of performance and price performance compared with Intel and Intel is still struggling because of their problems with 10 nanometer but also because of their monolithic large chip approach. And so Intel will I think change to chiplets I mean everybody else is doing it. It's been now pretty proven technology. But in the meantime, AMD is turning in benchmarks where for the most part except for some very narrow areas where one or two Intel architectural features can give Intel an advantage for the broad set of cloud and high performance computing tasks AMD is just leaving Intel on the dust. Nice, but the king of course the winner is going to be in the arm, right? When I look at this, there's the news I'm joking, but a little bit but when I look at the news right now today actually I think it's like in a few hours they're going to turn on the Fugaku the world's biggest supercomputer and it's arm powered and in Japan they're going to use it to calculate tsunamis and stuff, right? So this is also one way and Amazon has been doing some really interesting stuff they're doing a whole bunch of AWS instances on the arm of the last year. No, arm is clearly making head roads in this area the situation with Fujitsu to build the arm that goes into all of the Japanese supercomputers now those used to be powered by Spark and it must have been almost a decade ago that Fujitsu said hey, we're going to do the next generations of these supercomputers on arm they actually added to the arm architecture to embed many supercomputer kind of computational features that weren't in arm before and then they built these really monster chips to do it and so yes, for the last 18 months when you look at the supercomputer rankings there's the arm based chip from Fujitsu in the top of the stack and it's just a huge jump when I look at this thing it's nearly three times more powerful than the PowerPC number two and it's five times faster than the Chinese which was the fastest when I was at supercomputing 2017 in Denver I think that was the fastest one but this one now is five times faster It's amazing, yes and of course one of the key elements here is in order to get these kind of performance numbers you need a lot of cores and if you need a lot of cores it's really helpful if each of those cores uses less power because the challenge in exoscale computing is how to get an exoflop without making the light stem all over town so arm I think has demonstrated some real advantages in that regard compared with other architectures although I don't think that it is only the arm versus x86 that is making the difference there This is my video from two or three years ago at the NANARA Connect 2018 when they were talking about their new ARM A64FX the chip because they were of course very involved with the Linux on ARM to optimize everything for their chip so it's ready and it's kind of like launching today as far as I understand so that's happening but there's so much news there's been a lot of consolidation in this not just Nvidia trying to buy ARM but a lot of giants buying other giants Oh sure, for example HP bought Cray you still can buy a Cray supercomputer but you're getting it from a salesman whose business card says Eula Packard Enterprise So they were competitors? They were competitors, yes and so the world continues and it's unlike the universe which is continually expanding we have the semiconductor and the systems world in which a small number of vendors continue to acquire all of the companies that are smaller than them and so it's always an issue of you are either going to be the acquirer or the acquiree and so I think in many boardrooms looking around saying, hey who can we buy or who do we have to worry about defending ourselves when they want to buy us Alright and there's so many other topics to talk about I have one of your research papers a recent one What do you say in this one? Well there's a trend in enterprise computing these days called hyperconverged infrastructure I don't know if you've come across that in any of your other interviews but it's basically a way of being able to use a small number of computer systems to provide aggregated performance and ease of management the real challenge that a lot of data centers have these days is how do they manage all the machines and distribute workloads and so forth and hyperconverged infrastructures are a good technique to simplify that and as a result it's still a relatively niche market today but if you're talking to IT departments if they're not considering or looking at HCI then they probably should and this is a in partnership with AMD you did this one so are you talking about the epic AMD taking over some of the Intel Z on stuff Well AMD certainly is hoping that the advantages that they bring to the server market in general can be exploited successfully by companies who are selling these hyperconverged appliances companies like Nutanix VMware and such and they've been working with those software suppliers to optimize the features that are in each generation of epic to improve the value proposition for HCI in fact just talking to them recently they're planning to come out with their next generation server processors which are known by the code name Milan I got an email this morning saying they're going to make an announcement about that next week and they are hoping that with Milan they will be able to compete across the board with Intel in terms of delivering superior price performance, superior performance per watt more scalability for things that go into hyperconverged infrastructures so we will see whether or not they can convince their OEM customers the HPs and so forth Dell is very big in hyperconverged environments because Dell Dell owns a big share of VMware and VMware is huge in hyperconverged environments so AMD who's always had a very strong relationship with HP did I say that right? AMD has had a strong relationship with HP and AMD also now for the last few years has had a pretty strong relationship with Dell so they're hoping that these new Milan generations can flow into those companies server lines and in particular those companies are hyperconverged infrastructure optimized servers so that could be a big game changer since right now Intel has like 99% of the HCI market it's not an alarm about the end of our interview no I hope it's just the time you have I like to say that I share my office with a clock store and so anybody who's on any kind of calls with me usually here's times going off all the time are there Swiss cuckoo watch? no most of these are German movements as a matter of fact but the Swiss movement is the best right? I'm joking okay no because I'm Swiss so no I'm okay maybe you can talk a little bit about how many different research documents do you have like all kinds of things planned out or you published stuff before about all kinds of different topics recently? well you know I have being an independent consultant I have a lot of options in terms of how I spend my time and at this point in my life I'm actually spending less and less time worrying about semiconductors and more and more time worrying about where I should go on my next trip when we finally get back to some sort of normality and hey I had to both of my vaccinations so according to the CDC this morning I'm actually in pretty good shape I can go and see grandchildren or whatever but I do you know because I find the industry so intellectually stimulating I do keep track of what's going on in terms of actual gigs to monetize my knowledge that's becoming less of a priority for me alright so basically now you can go to the supermarket without wearing a mask hey yes although I just got a new mask I don't have it here that has an embedded LED display that you can program you can have it say I'm vaccinated that's exactly what I'm going to do alright so that would be oh you can sell advertising space the space for sale I wonder what what Nuvia is going to add to Qualcomm because as I understood it was like in the servers but it sounds like it could also be like helping Qualcomm make better chips for laptops and desktops because that's the thing that I've been hoping for now with the M1 it's such a I'm using the MacBook M1 right here and so far I'm not crashing I mean I'm crashing everyday a little bit of course but that's more because of the Mac OS you know the challenge I think for the M1 is being able to run Mac applications that have not yet been ported and optimized for the ARM environment and I think that's going to continue to be a challenge until that conversion is through and similarly Qualcomm has been trying to convince people that an ARM based PC running Windows 10 is a good choice and there I think it's even more problematic that you will be able to run the entire repertoire of Windows applications successfully so I think there's still an emulation problem that is going to exist for a while for example right now I'm in Chrome on my MacBook M1 but I'm experiencing bugs every day pretty much like a bunch of weird stuff like the video doesn't show up or I have to go over to Safari to see something a little bit weird and that's with a Chrome that's supposed to be already optimized for the ARM and then when I use my Windows 10 on ARM Qualcomm laptop of course the only Chrome that I can do is the one that I have to emulate from x86 there's not even a real Chrome yet, there's only Edge or Firefox and I'm not even starting to talk about getting support for all the Adobe apps or something like that, that's not even started on Windows supposedly there's some beta and it's going to eventually maybe happen on this M1 that's the very positive thing is that Apple has a tendency of pushing people around in the way that they want and they're definitely going ARM forever so they're going to push, they have trillions of dollars in the Swiss bank account or Irish bank account so they could encourage people to port all their apps now or the US government could just seize all those trillions of dollars and that would help pay off some of our national debt Yeah, they should just take the Apple money to pay for the 1.9 trillion dollar stimulus H&K, pay your share, yes Yeah, I think that would be fair that would be great Yeah, so there's always this kind of talk also sometimes that well this is just a patent dispute but usually there's all these other things that's again in politics where the different governments are going off the big tech in some kind of way trying to put some slowdown Yeah, but in that Intel situation this was a company that bought some obscure patents that Intel was infringing on and then they brought a suit and they pursued that infringement for a while and the most recent thing coming out of this West Texas district court which is known to be a very patent troll friendly court came out with a ruling what is it, 2.86, 2.6 trillion or some humongous number I've never been a real fan of companies that buy up intellectual property and then pursue people who are violating them and in general when somebody comes to me and says hey Nathan how would you like to be an expert witness for us we're suing Intel to do because we think that they're violating our patents in general I say I'd rather not I don't really like that particular business model but you know some people do and every once in a while they win either in court or they get a settlement out of court so the nice thing about the semiconductor business in general is there are lots of different ways to make money and you can make money by inventing things you can make money by selling your inventions to other people who can then utilize them and hey whatever works for you but again often in the case like this particular one with Intel my guess is and I have not looked at it in detail Intel designed their chip using features that were either generally known at the time or that their engineers came up with independently and they accidentally wandered into an area where somebody already had a patent and you know you can't research all of the patents or how a particular design could infringe on everybody's patents because that would slow your design to keep cycles out tremendously and so what you do is you design so that you don't think you're infringing anybody and every once in a while it turns out you did as a matter of fact Intel as a result of that exposure often tries to achieve or enter into what they call cross licensing agreements with anybody who any other manufacturer has an actual property so for example Intel and AMD have a patent cross licensing agreement at various times they've had big battles over what was covered in that but because they do have those cross license agreements that means that for the most part if AMD came up with a really clever circuit trick in one of their products and Intel independently or even independently used the same trick then neither company can sue the other and Intel calls that design freedom because they don't have to worry as much about infringing other intellectual property they've had those agreements with most of their customers and many of their competitors for years and the only problem that people who try to have those cross license agreements to cover them encounter is when somebody who doesn't actually make products and doesn't need any of your patents or doesn't have to worry about infringing your patents comes along and that you discover they're infringing one so that's when the intellectual property owners with who are not manufacturers can in fact challenge Intel and sometimes win in court alright and this was more in the consolidation area the silence 35 billion AMD just has I guess they're a little bit successful the stock grows like crazy over the last three four five years and it's a little bit the same thing that happened with Nvidia they went from being valued 30 or something billion to 300 billion or something like that and that's how they can just start to suggest that they want to buy ARM and now AMD is talking about buying silence I wonder if they will take these FPGAs and put them in some of these future AMD chips oh I think clearly they will I'm not sure if that deal has gone it's been approved yet by all parties but unlike the situation with Nvidia and ARM if AMD were to acquire Xilinx I don't think that any of Xilinx's current customers would worry about AMD tilting these Xilinx roadmaps in ways that would be what's the right word that would disadvantage them regarding with regard to Xilinx's other customers and so from that standpoint that makes sense and there are a couple of things that are going for that acquisition one is that Xilinx is a big player in the 5G infrastructure market FPGAs play a big role in 5G infrastructure in general because you have very critical real-time performance requirements and the standards are always in flux so you want to instead of building an ASIC to do switching say for a 5G base station you throw in an FPGA and that way as the standards evolve you can just download a new code so FPGAs have always played a big role in 3G, 4G and 5G infrastructure AMD had no exposure basically to the 5G market and so buying this gives them access to the 5G providers who are building, deploying 5G infrastructure base stations in particular all right if you don't mind I just opened this interesting funny article it says 2021 will be calmer year for semiconductors except for Intel it's kind of like a funny title I guess they talk about consolidation as much as we want to say boy Intel is really in trouble the fact of the matter is their factories are all running at 100% capacity their customers are begging for more people can say oh yeah they really screwed up on 10 nanometers Intel is doing okay the money is slowing in and an alarming rate every quarter seems to be bigger than the previous quarter and so even though they're struggling with some leading edge technology their customers are still buying and that's really the crucial thing so I think even Intel is going to have a good year in 2021 nice so you're looking forward to a network with all these guys again over there in the Silicon Valley where are you based? I'm in Silicon Valley I'm in a little bedroom town called Saratoga that's just on the western fringe of San Jose Silicon Valley and you know I run into some of these people in the supermarket Do you know all these guys that invented all this stuff? Did you meet them in the last 22 years? Well I would say that I know all of the people who are in the analyst community as a matter of fact I was on a call this morning with a number of analysts and we all recognize each other's voices and we all know well here David's about to talk he's going to give us some sort of really erudite very techy kind of question and so forth Do you publish those analyst calls? No no as a matter of fact these analyst calls are basically done on an NDA basis to alert us when some new important developments are popping around tonight today that I can't tell you about It's exciting every day what happens and this year is going to be a fun year This year is going to be a fun year and hopefully not on wood we will get back to a situation closer to normalcy and it couldn't happen soon enough And meet at the Santa Clara Convention Center for something Yeah well you know what they've done is they've repurposed so many of these convention centers to be massive vaccination centers Yeah alright Okay good talking to you again Thank you so much thanks for an overview so we can understand all the stuff that's happening there's so much stuff happening it's huge very good looking forward to doing it again sometime Thanks a lot and thanks everybody for watching