Where have all the chips gone? Scott Willsey joins John to look at how the shifting automotive use of ICs, a drought, a pandemic and poor planning amongst other things has led to an unprecedented global chip shortage with no end in sight.
Welcome to Pragmatic. Pragmatic is a show about technology and contemplating the finer details in their practical application. By exploring the real-world trade-offs, we dive into how great ideas can be transformed into products and services that impact our lives. lives. This episode is brought to you by ManyTricks, makers of helpful apps for the Mac. Visit ManyTricks.com/Pragmatic for more information about their amazingly useful apps. We'll talk more about them during the show. Pragmatic is supported by you, our listeners. If you'd like to support the show, you can by supporting our sponsor or by becoming a Premium Supporter. Premium Support is available via Patreon and through the Apple Podcasts channel subscription. Premium Supporters have access to early release, high quality ad-free episodes, as well as bonus material from all of our shows not available anywhere else. We're edging closer to our monthly goal to go advertising-free across the network, but we can only do that with your help. Pragmatic is also a Podcasting 2.0 enhanced show, and with the right podcast player you can stream Satoshis and boost with a message as you listen. Visit engineer.network/pragmatic to learn how you can help this show to continue to be made. Thank you. I'm your host, John Chidjie, and today I'm joined by Scott Willsie. How's it going, Scott? Good. How are you John? I'm doing very, very well. Thank you. I appreciate you making the time to come on the show. One of the things that I've been hearing a lot about, and I think it's not just me, everyone's complaining about a global microchip shortage. I wanted to have a bit of a dive into a little bit about microchip manufacturing, but more about also what's led to this situation. And I know there's been some different deep dives, but I wasn't really happy with them. So, And I thought, who do I know who knows a little bit about like microchips and stuff? And I thought about you. So I appreciate you making the time to come on the show. Thank you. Yep. No problem. And one thing that I should say is like any other industry that has different specializations, my particular area of specialization is semiconductor test. So just like the people who are in the wafer fabrication end of things probably don't know anything about what I do. I know some of what they do, but not as much as a lot of people. So just, yeah, it's a specialized thing and we all get our tunnel vision going, but yeah, I definitely am in that industry. - Well, that's okay. I mean, it's kind of like, so when I worked for Nortel going back now 20 years, because wow, time flies and you're having fun and all that. But yeah, when I worked at Nortel, I wasn't working in circuit test or functional test, but I'd go to the Westwinds facility and I'd have a look at the test benches and so on and so forth. So I was kind of peripherally aware of what it was, how it worked, and I2C and JTAG and all the different test components and some of which I actually did design into, but never had a chance to particularly test my design, unfortunately, because, hey, I got laid off, but that's another story which I'm sick of telling. But never mind. The point is that, Yeah, I mean, I was working in RF design and reliability analysis and prediction. So again, yeah, you're right. It's different speciality within the company. And Nortel did have an ASIC design division, but we weren't a fab. So, you know, I'm 20 years plus removed from a different part of the business that didn't have anything to do with ChipFab directly anyway. So I still say you're slightly closer to it than me. So irrespective of all of those, you know, statements, we're all good. And anyway, it's fine. Right. So funny thing. Well, with complicated products these days, there used to be the case you might have one or two integrated circuits, but there's now getting to the stage where we've got thousands of integrated circuits. And I know that like systems on a chip, they want to try and integrate all that into one massive chip. And I mean, obviously, that's a thing and that's fine. But lots of manufacturers don't want to pay that money to do an application specific system on a chip. They don't want to do all that extra expense and developing that. So what they'll do is they'll simply say, right, well, I'll just take you and I'll take you and I'll take you this microcontroller and that bit of memory and whatever, whatever, whack it on a board. Hey, presto. So all of those ICs that are going into all these different products, you know, that is still the way the majority of manufacturers will manufacture their stuff. And what's happened in the last little while is we've had this thing called COVID-19. I'm not sure if you've heard of it. Never. You were. No. No. Oh, well, let me fill you in on that then. There was this thing that made people really sick and different companies, tried different different countries, try different strategies, like closing down borders and stuff. And anyway, then people said, hey, there's a vaccine. And we don't feel like having it. And I think they are up to date. There you go. That's why nobody's been at work for the last year and a half. I think, yeah, that's it. Yeah, that's it. Yeah. Got it. Yeah. Okay. Everyone knows what's going on. So COVID-19, what happens? All the manufacturers say, well, we can't go to work, so we're going to stop production. Well, obviously. And some of these industries, you just can't really do much from home. And one of those is the automotive industry. So it's hard to assemble and paint a car and a car body from home. So do what you had to do. General Motors, they closed a whole plant in Kansas City. In fact, I think it's been multiple ones. That one particularly was closed in February, still hasn't reopened. And a lot of the other manufacturers have done similar things. And Porsche, for example, said that there was a three month wait because for some of the vehicles, because they are waiting on a tire pressure sensor, but the rather the IC that monitors the pressure in the tires because of a lack of supply availability for that chip. Ford, for example, had a stockpile of their pickup trucks that covered such a massive area. They actually took a photo of it, I think it was from the space station. You could see it from space, this massive collection of pickup trucks in a massive backlog, you know, just because they couldn't get ICs to finish them. It's like it was 99% done, but you couldn't actually drive the car without the missing, there were missing chips in them. So the whole situation has come about because of that. But when you start peeling that away, it gets a little bit more interesting. So one of the things that I sort of I did realize this, but not to the depth of it was that a lot of the chips that weren't being prioritized, that are creating the issues are actually the dodgy little cheap ones. They're not even the complicated ones. Yeah. And it comes down to the, I think that because the automotive manufacturers decided that they would cancel a bunch of their orders. 'Cause we can live without those for three months, we'll probably shut down for three months and COVID will just blow over, it'll all be fine, right? Two thumbs up. Not really how it panned out. So anyway, so the commodity chips sort of like lost out to the more expensive ones or the ones that have got better margins. And that's basic economics, it kind of makes sense to me. - Yeah, and I think one of the things that is probably less obvious is that The infrastructure to support legacy items is not as robust as the infrastructure, well, in my opinion, from what I've seen, is not as robust as the infrastructure to support the manufacture and test of the newer items. And so, if you, like even when I started working at the company I work at many, many, many years ago, we had things like brake controllers and then some chips that were being sold to the government, and they were all commodity stuff, even at that time. we were using rickety old equip rickety old test equipment and software that hadn't been updated in forever to test them because there was no need to now Even just a couple years into my stint there some of that stuff was becoming hard to get hard to repair hard to find parts for and There were like third-party companies that would go into business just to support some of that legacy stuff for smaller test facilities that couldn't afford to upgrade their equipment or Change their process to newer equipment and so I have to imagine that Once some of the focus went away during the pandemic on some of these commodity items Some of that infrastructure went away too And then that is really hard to bring back even if you want to even if it seems cost-effective, which it really isn't anyway Okay, so I'm guessing that that also has to play a role I mean it comes down to a matter of floor space as well I suppose I mean if that legacy stuff is simply not as profitable then why wouldn't you replace it with newer equipment and then the older technology is no longer easy to actually retrofit. That's kind of the point you're making, I think, isn't it? Because no one makes the legacy test equipment as well anymore. Yeah, it's that in order to support the manufacturing and testing of that stuff, you have to have extremely low overhead to begin with. And if you've been through a pandemic and to you, stop testing some of that and you're going to use your floor space for something else. There's no point ever going back to that. And then only the only the smallest companies can really afford to do commodity items in bulk and still make money at it anyway. And so, yeah, it feels like a chicken and egg situation where it just gets worse the more you look at it. And I think this would have happened eventually anyway. I just think the pandemic revealed some facts about certain cheap ICs that were always going to happen, but it just accelerated the process. Because I mean, I guess thinking about it from the point of view of if I was a chip fab and I had a certain amount of floor space and I had a lot of mounting amount of profitability from personal electronics, which more and more people are going to still use, whereas the automotive ones, which were never good margin in the first place and probably mostly legacy stuff. And I'm looking at that floor space, they've just canceled a bunch of orders. What's the natural thing to do is to simply extend and then use that floor space for newer ICs that are higher profit margin. - Yeah, definitely, definitely. - Yeah, that makes a lot of sense. Appreciate you bringing that one up because that was one I hadn't come across with the legacy test, for example. And I guess when I was digging into this, it's generally the newer complex ICs that have been the problem. And I guess there was the whole GPU with crypto mining though. That's another interesting side, side 'cause that kind of happened around the same time. Well, I don't know. I mean, so I've been sort of loosely following cryptocurrency more or less since podcasting 2.0 sort of became something I was getting interested in. So probably about November last year. So prior to that, there's a whole history of Bitcoin that I never really followed much of only really retrospectively. But my understanding was that things like ASIC miners have kind of taken over a lot of the mining space, but GPUs are still very useful if you want to do, let's say, mine Ethereum and then you get paid out in Bitcoin, for example, which is something that a lot of people do using intermediary services like NiceHash, for example. But that's all based on GPU. Right. So, yeah. So there is a lot of demand for the GPUs for crypto because the value of Bitcoin goes up, well, is up at the moment, I suppose. And that all happened around about the pandemic as well. Yeah. And I think the reason why people suddenly noticed the GPU thing wasn't for the same reasons they suddenly noticed they couldn't get other products because these commodity chips. But the pandemic accelerated people's desire to buy hardware or upgrade hardware. and now GPUs aren't as available as they used to be. And they're bumping into it more because they're trying to order computers and graphics cards and they're just not there. - Kind of, it's really a fascinating confluence of different- - Yeah, absolutely. - It's crazy. I think what might be useful just to pause a little bit on the causes and just, let's just talk a little bit about integrated circuits and wafers particularly. 'Cause I don't think I've ever talked about it before. And since I've got your, you know, I figured it would be a good time to sort of pick your brain about it. Because it's funny, you know, when I first heard about silicon wafers, the first thing I thought about was, you know, like a wafer biscuit. I don't know why, I was probably hungry at the time, but, you know, I mean, as you look at these things, they are wafer thin, yes. But you also hear them referred to as substrates. And another one that I've heard less commonly is a slice. But in any case, just give us a little bit of a background or your thoughts on how would you describe wafers, fabricating them and so on and so forth. Yeah, so basically, the way that integrated circuits are made is they're basically printed onto what are called wafers, which are basically silicon ingots. So if you think of a long tube, not a tube because it's not hollow, but a long cylinder of silicon, and then that's sliced up super super thin into wafers. And of course, they're polished within an inch of their life so that they're basically extremely fragile and beautiful little mirrors and then they, and with wonderful edges that aren't sharp or anything, they basically really polish these puppies up. If you saw the silicon ingots, you'd wonder how they could make the wafers so uniform in shape, but they do. And then they basically take those and basically print circuits on top of them with chemical deposition and, well, the whole process that John will get into. But yeah, it's a really fascinating thing. And when I started, the company I started with, we had four-inch wafers and we were just... I don't know how long they were into six-inch wafers, but it was four and six-inch wafers, which the six-inch are 200 millimeter is that right? Yes. Yeah. So just on the wafer sizes and quickly so the the first ones I was doing research on when I was looking at this the one inch or 25 mil was around 1960 and then two inch went up or about 51 mil went up to 1960 it was from 1969 and then three inch about 1972 and then what you're talking about now is the 100 mil or four inch ones it was 1976 that's 45 years ago which is how old I am and then we've got 125mm which is 4.9 inch, 981mm and then that wasn't around for long they just went to 6 inch which was 150mm. Yeah 150mm. That was 98mm. Right so then and then not long after I started we went into 8 inch which is 200mm. Yes. And now of course everybody's on 12 inch which is 300mm. Well I say everybody that's not necessarily true. I only see 300mm wafers now but a lot of people are still using 200mm. Well one of the other things I when I was researching this is a lot of the shortages have come from the 200 millimeter sort of fab plants and one of the other thing that's interesting is that people are investing more in the 300 mil and people start looking to 450 mil but you know it's like they're not investing in more 200 capacity or at least they weren't that's right but I think the situation is now sort of like may have changed that slightly so maybe there are some more that are considering it But it's funny because as they get bigger in size, it's like the economics gets better. The economics gets a lot better. And I also wonder from the vendor standpoint, like even though we're only dealing with 300 millimeter now, we still have cases where vendors are like, we don't want to support that equipment, that old equipment anymore. You guys are going to, you know, eventually, even though we try to use stuff forever, they don't want to support even old 300 millimeter equipment anymore. So I wonder what the situation is like for people that are still fabricating and testing 200 millimeter wafers, what their vendor support situation is like. It could be that that's also a huge part of their problem. For sure. I think that, cause the other thing I was thinking about is the larger the wafer also, the more fragile it becomes, but you've also got some issues with making sure you've got consistency from the center out to the edge. So I'm not entirely sure, like, because everyone's pushing, well, not everyone, but I know that the market is moving towards, you know, better, better economics, which means you want to go to 450 mil. I do wonder at what point, because the substrate's got to be thicker and thicker to handle, well, essentially not breaking the larger it gets. Or not bowing in the cassette even. Yeah, exactly. So it's like at some point it's the physical limits of the material we're using as a substrate will make it prohibitive to go any bigger. So you're going to hit some kind of a wall. I don't, I don't think we've hit that wall necessarily yet, but no, I don't know. And I think with some thicknesses of wafers, I think they were running into those issues when 450 was being researched because we did have some minor amounts of 450 research going on that I saw mostly with regard to test equipment and what that would look like. But yeah, I think they were definitely running into 300 millimeters, pretty handleable like it's not really an issue. That's a solved problem. They're not considered to be that fragile, but I think 450 was a different story. Yeah, exactly. So, um, just, uh, circling back then to the, uh, to the whole manufacturing process of the wafer. So after you're saying you, they, uh, they squeeze it out like toothpaste. I mean, I say like toothpaste, not really, but you know what I mean? Um, extruded, I think is the correct word. And, um, slice and die, slice and polish. So you got your cool wafer. And then there's the different methods of actually making transistors and circuit pathways on them. And not wishing to go into too much detail about P and N silicon, but, you know, P for positive, N for negative, I guess. But one allows holes to flow and one allows electrons to flow. Not that holes flow. They don't. But never mind that. But the point is that you create, take the silicon, you put impurities in it. And of course you put a PN junction next to each other and then you can create a diode. And then you put three next to each other in a sandwich, you get an NPN or a PNP. And that's if you're talking about BJTs, but these days it's normally done with gates and a channel in a FET or field effect transistor. But irrespective of all the, just trust me on that. because that's like rudimentary silicon NPN junctions. But the whole idea is we start off with silicon that has no personality, and then we dope it to give it a personality. Put it that way. Yes. Sounds like a crazy party. Yeah. It sounds good, doesn't it? It sounds like a lot of fun. See, this is why I feel like I've been missing out all these years. Because you've been planning the space in a manner of speaking, and I haven't. So it's kind of unfair. Well, to me anyway. What I hadn't come across until I did this research was ion implantation, which sounds really cool, actually. And it makes me think a little bit about targeting, like a targeting scanner for a, for like a cancer treatment machine, you know, like a Therac 25, although not a Therac 25 because they kill people, like one that works and doesn't kill you, one of them, whereby you have a source and that source, you know, between 10 to 500 kiloelectron volts and through a series of separation magnets and beam forming magnets, It just applies those ions to the target on the wafer. And I hadn't come across that one before. That sounds really cool, but it doesn't sound like it's something that's done at a mass production scale. So etching photolithography, I guess. So I actually did the whole photolithography etching thing at a circuit board scale and where you put on the photoresist, and then you put on the positive mask, you expose the ultraviolet light, You take that off, then you etch away your copper, and then you take off your mask, and then you put on your protection layer. Now, wafers and silicon wafer tech is more or less the same kind of process, but it's evolved a little bit from putting in hydrochloric acid, which is good, 'cause I think that would be very painful. The wafers might scream if we did that to them. So the wet and dry one, there's two different ways of doing that. The dry plasma one is very cool. if you come across that one? - I am aware of its existence, but how does it function? I don't know. - That's okay. So my understanding of how it works is that you basically take the wafer, you've put your resist layer on, and at that point, then you expose it, you put it into a chamber, and they reduce the pressure in the chamber to about 133 pascal. I'm trying to remember, it's not a vacuum, but obviously to create a plasma, you need to increase the temperature, and it needs to be a near vacuum. Otherwise that all of the plasma particles or high charged particles will just get absorbed by the gas that's still within the chamber. So same kind of way a nuclear fusion reactor kind of works. I say kind of works 'cause it does work, but it's still not useful yet. So anyway, nevermind. Another story for another episode maybe. But anyway, so on the VLSI, very large scale integration, a lot of the dry plasma is becoming more popular. And so those ions, those basically just bombard the non-resist protected parts of the silicon to create the pattern that they're looking for. So that's kind of cool. - Yeah, it's very cool. Some of the physics in the manufacturing process are pretty mind blowing. And that's why it's also getting super expensive is as these processes get smaller and smaller, You're hitting physical limits so that things like photolithography that you were mentioning while relatively easy for PCB size Components for what they're doing in some of these tiny Processes for for ICS it gets really complicated and then the investment in new equipment to support newer processes Gets very expensive really fast really really fast So it's kind of weird because you're although you're right that It's not the more complex chips that are the bottleneck. They're also super expensive But because the investment ramps so heavily the smaller your process goes That's absolutely why you can't afford to not make money on the devices that you're selling. Oh for sure You you just have to make devices that that cost hundreds of dollars. Yeah, that's where the money is as I say, right? I kind of It's a good point too, is that we talk about the size of like 7 nanometer, 10 nanometer, 13 nanometer, whatever. And that size, as you say, if you're trying to essentially etch and then maintain a conductive pathway between two places very, very closely packed together and so on and so forth to create the layer that you're looking for, you are very rapidly reaching, you know, not quite atom layer, but where the technique necessary to actually do the the erosion of the bits you don't want, for example, or the doping of the bits you'd need to dope, for example, it's like that that is getting more and more complex. So maybe the whole ion, for example, ion implantation might become more the way that it is done if it can be done more accurately. It's just that it's going to be a lot slower and require bigger equipment and be more expensive. But if we want to shrink the dye further down, we have to do use technologies like that. - Mm-hmm. - That's my understanding anyway. - Yeah. - All right, so got to mention thin film deposition because, you know, some of the cool, some of the really cool technology with that is, can go do a single atom layer at a time. That's kind of neat. - Yeah, that's crazy. - But you know, you can split that into like chemical or physical just broadly. So chemical deposition or physical deposition. So you've got chemical uses a fluid precursor and it tends to lead to a conformal coat, with rather than a direct coat. And within that subgroup, there's plating with a dissolved mineral, for example, that you want to deposit. And it works, you know, just like electroplating, for example, which is something that I've done on a macro scale, but I'm aware that it is done on that scale as well. But anyhow, then there's organometallic powders. So that whole organic thing, apparently they call it Solgel for short. That's a cool name. There's actually a lot more chemical ones and different ones. I was going to stop there, but if there's any others that you think are worth mentioning. No. Nope. All right. That's okay. So on the physical side for thin film deposition, you know, you've got mechanical, electromechanical thermodynamic effects that they use for depositing the film. It's generally classified as physical vapor deposition, which, or PVD for short, which, you know, if I remember correctly was one of the techniques that they used actually for one of the iPhones that was the jet black iPhone. I'm trying to remember if that was physical vapor or if it was a CVD, but irrespective anyway, nevermind. Moving on. And then finally, when you've done all the different processes to etch out what you need, then you slice and dice that wafer and put it in a package and seal it. And then you sell it and they solder it onto a board or solder it onto a board, depending upon what country you're from. And that's it. - So super simple, right? - Very simple. Nothing could, what could go wrong? - What could be simpler, right? Oh man, anyway. So we talked a bit about wafer thickness and so on. So now I just want to do move and talk a little bit about what was leading up to the COVID situation in terms of semiconductor manufacture. So a whole bunch of different, you know, like you got people like Deloitte, of course, that do studies on anything and everything that people pay them to, the whole bunch of different financial groups. And one other one, Susquehanna Financial Group, and they've been tracking semiconductor distribution since like 2017, which is not as far back as some, but what I found interesting was that they tracked the average order and delivery time for new chip designs. And it was 15 weeks by February, 2020, that was over 22 weeks. And the next thing that was interesting was the Deloitte study that looked at the amount of semiconductors and electronics in terms of their overall cost to make an average car. And they track this over the decades, I pulled out the ones that I thought were interesting. So in the 2000s, 18% of the cost of an average car. And of course, the average car took into account different kinds of vehicles, different sizes of vehicles, and so on, so forth. But they're talking about, you know, like cars, they're not talking about like trucks or prime movers or any those. They're just talking about cars for, well, I was gonna say for people, I'm not entirely sure what else cars would be for, come to think of it, but never mind. So 18% of the overall costs in the 2000s, in the 2010s, are up to 27%. And now in the 20s, we're 2020s, we're looking at about 40% of the costs of the car are the ICs that go in it, which is kind of crazy. That is actually very crazy. It's a lot because it's like the steel and all of the other bits and pieces in the engine, everything is the mass manufacturer of those. It's like it's a well understood thing. And the engines, for example, for vehicles. You know, all of that is essentially is a limited number of designs and a lot of manufacturers will share the engines and so on and so forth. And it's not it's not as expensive because a lot of that is just like, you know, you get the sand, you pour the molten, you know, in molten metal into your form and blah, blah, blah. It's a well understood, well-known process. Whereas the ICs, they're not. It's all more cutting edge. And of course, as the world's moving towards electric vehicles and more automation. So things like all of your, like tire pressure sensors. Like 20 years ago, what car had tire pressure sensor in it? I can't, yeah, maybe there was some, but I can't think of any. And you go back, you know, 30 years ago And not many cars would have had emergency braking or adaptive cruise control or all these different, you know, high-level features that we've all just like, oh, well, this is now kind of like a standard offering on some of these cars. And all of that needs sensors and it needs microchips to do all of the run, all the algorithms to figure out, well, OK, my ultrasonics, I've got four ultrasonics, I'm going to vote three out of four of them. My distance to the next obstruction is this. therefore I need to maintain this distance, blah, blah, blah, go at this set speed. And it tells the ECU, you know, you're allowed to cruise at this speed or whatever. Sure. And that's and that's just one thing. Yeah. And I'm guessing the I'm guessing that the cost difference is going up exponentially, not just because of the numbers of chips, but also the complexity of the devices. And they're probably because they're newer devices, they're probably being made on newer processes. They're probably they probably just cost more to make to begin with, regardless of what their function is. And then there's more of them. So my guess is it's a steep upward curve for the price of for the cost of what these automotive automotive makers are putting into their cars now. Yeah. So effectively, those two things are compounding. And that's what what's driving that percentage up in terms of cost. So the quantity as well as the complexity. Yeah. And it's all driven by the new features and so on and so forth. I mean, just thinking about, you know, just Tesla for one, for example, I mean, that car is in essence a computer on wheels. So the amount of computing power it's got on it, it's kind of ridiculous. Just one example, but that's an extreme. And obviously they're not an average car. If you look at it proportionally, I don't even know if Tesla's account for 1% of all the cars in the world, probably less than that, even they're not a big player. So all of the other players like GM and Ford, and a lot of the European manufacturers like Mercedes and BMW and Japanese ones like Mitsubishi and Toyota and Honda. So they're in putting more and more smarts in their cars as well. And it becomes an escalation. So it's like if one manufacturer has now has adaptive cruise control, everyone flocks, oh, this is a really good feature. And then all the other manufacturers then have to pile on and they get on board and then they have to have that feature. Otherwise their car will be left behind and it won't be as attractive to buyers. So I think the more smarts go in these cars, that's gonna get worse. When I say worse, I mean, more and more of the car's cost will be the smarts in it. - The good news is the changes in the automotive industry with the more stuff they put in might alleviate some of their shortage problems because the automotive industry tends to lag way behind in terms of their electronics that they use and the technology in their cars. But now that seems to be changing a little bit. And if that's true, that means that although more expensive to implement, they'll have more modern devices and maybe they won't be running into as many of these shortages, who knows, but things like tire pressure sensors and that kind of stuff are always going to be cheap items that they don't want to spend a lot of money on anyway. Yeah. And that's the funny thing is that if, um, so the next angle to look at actually is, um, the cost of the ICs that they use in terms of like overall, um, semiconductor, um, proportions from a broader market. And, and these numbers came from a Deloitte study as well, And it didn't dig into all of the details as to which of the fabricators this, this came from, it was a mixture of them. So roughly 15% of their sample, were essentially semiconductors that were manufactured specifically for the automotive industry. Right. Which is in, in terms of the, in terms of the global sense, I still think that sounds like quite a lot, but the truth is that it's dwarfed by personal electronics, which is like 50% of the semiconductors that are essentially manufactured, the ICs are manufactured for personal electronics. And then the remaining 35% is a bit of a mixture. So you've got some industrial stuff in there, medical, household appliances, and so on and so forth. Because of course, every household appliance has got to be a smart appliance now. So, you know, my washing machine sings to me when it's done. And when it's done singing, it sends me a push message to say, "Hey, the load of washing's finished. you should put on another one. Is it a good singer? It's actually not a bad tune. Yeah, it's not bad. I kind of like it. Yeah. But anyway, it sounds very positive and happy. It's a happy tune. Yeah, I was just about to. Oh, dear me. And the toaster as well. The toaster gives you a bit of a beep and the sorry, the microwave also sings, but its song is not as nice as the washing machine. So we're going to start rating our appliances based on how nice their music is when they're done. Yeah. Anyway, I get push notifications from our fridge when it wants its filter changed. And I don't know about that. I don't think I'm a fan of my appliances push notifying me of stuff. Yeah, I have enough push notifications in my life without, you know, all these different appliances and so on and so forth with all their different pushiness. But back when I was a boy, our fridge didn't tell us what to do. So, so now circling back to a little bit of the, you know, the, the foundry. So one of the things I think it's, it's interesting to, to look at is, is how they're actually manufactured from the, the idea of a, of a foundry. Cause there used to be a time when, if you wanted to manufacture an integrated circuit, you'd have to design it, build it, manufacture it yourself. Yeah. So I'm thinking like, maybe, I think like Texas instruments, like for example, if I remember correctly. And there's a few others, but the idea of a foundry, which is just a fancy name for a silicon manufacturing, integrated circuit manufacturing place, or a chip fab is what I used to call them, but foundry is what everyone likes to call them nowadays, but whatever. But their whole job is not to do the design. Their job is to simply, here's a design, go build it. So, contract manufacturing effectively. And TSMC is the largest. And in terms of standalone foundries, this is not of all of the companies in the world that can manufacture integrated circuits, but of those that are dedicated foundries, that's all they do. They make up 50% of the global wave of foundry capacity. And that was as of 2020. This episode is brought to you by ManyTricks, makers of helpful apps for the Mac, whose apps do, well, you guessed it, ManyTricks. Their apps include Butler, Keymail, Leech, Desktop Curtain, Timesync, Usher 2, Menuware, Moom, Name Mangler, Witch, and Resolutionator. There's so much to talk about for each app that they make, so we're just going to touch on some highlights. Usher 2, the return of the classic Usher, but now it's a full 64-bit app that works with the latest versions of Mac OS, including Monterey. So what is Usher? It's an amazing, powerful media management and playback app that can see movies that have in TV, Music and Photos apps or any library location you'd prefer on your Mac. It can organise them for you if you like, you can create advanced playlists and sorting criteria, and you can even edit their information all from within Usher. Menuware puts your Mac's menu anywhere, well wherever your mouse pointer is. It's such a simple idea that is so very handy. Instead of having to mouse up to the top of the screen, press a hotkey of your choice and Menuware puts the menu right where your mouse is. You show the current active app's menu or the entire menu, then you pick your app from there if you prefer and as another bonus time saver you can even hide the menus that you rarely use. It's a great time saver for me and I bought it within a minute of trying it because I saw the value in it immediately. Moom makes it easy to move any of your windows to whatever screen positions that you want halves, corners, edges, any fraction you like of the screen. And then you can even save and recall your favourite window arrangements with a special auto-arrange feature when you connect or disconnect an external display. It has full touch bar support and keyboard integration with Adobe's apps and it also works perfectly on an iPad operating in sidecar mode. It's the first app that I load on a new Mac because it's just awesome. Name Mangler. You've got a whole bunch of files to rename quickly, efficiently and in large numbers? Well, then NameMangler can help. It's designed for staged renaming sequences with powerful regex pattern matching. Recent additions include a group by feature. When making a sequence and title case conversions can now keep their existing formatting or convert to lowercase on word length. The best part is it shows you the result as you go and if you mess something up, just revert back to where you started and try again. Which You should think about which is a supercharger for your command tab app switcher. If you've got three or four documents open at once in any one app then Wicha's beautifully simple pop-up quickly lets you pick exactly the one you're looking for. You can switch between tabs as well as apps and app windows with horizontal, vertical or menu bar switching panels with text search for switching, you can show the front most app in the menu bar icon with full touch bar support and much, much more. And finally, Resolutionator. And it's so simple. A drop down menu from the menu bar and you can change the resolution of whatever display you like that's currently connected to your Mac. The best part though? You can even set your resolution to fit more pixels that are actually there. It's very handy when you're stuck on your laptop and you need more screen real estate. Now that's just six of their great apps and that's only about half of them and they all work with the latest version of macOS Monterey. All of these apps have free trials that you can download from manytricks or oneword.com/pragmatic and you can easily try them out before you buy them. They're all available from the website or through the Mac App Store. However, if you visit that URL you can take advantage of a special discount off their very helpful apps exclusively for Engineered Network listeners. Simply use Causality25, that's causality the word and 25 the numbers in the discount code box in the shopping cart to receive 25% off. This offer is only available to Engineered Network listeners for a limited time, so take advantage of it while you can. Thank you once again to ManyTricks for sponsoring the Engineered Network. So one of the other things that's interesting about foundries. And this is something that I didn't realize just how much water they use. It's kind of crazy. So according to when I was looking into TSMC's history, they use 63 megalitres of water every day, which is of course conveniently 63,000 tons. And that equates to about 10% of the local reservoir supply in Taiwan. So that's a lot. And one of the other things that was impacting TSMC's ability to make chips was that in 2021, there was a drought in Taiwan. And, and they, and I said, yeah, but what, what's water got to do with, with making integrated circuits? Well, that's a lot of water. So what they've been trying to do is implement some very, very pure, ultra pure water and ultra pure water. I mean, I'm chemically, when, when I worked at Standalone Power Station years ago, That was when I first came across chemically cleaned water because we need that for the boilers. Otherwise, when you heat the boilers up to make steam, to drive the turbines, to create electricity, you know, obviously if you have different metals dissolved in that water, salts and so on and so forth, they'll come out of solution at high temperature and they'll cause corrosion and they'll just basically destroy your system. So I used to think, well, that's, you know, that's pretty hardcore, But when you're in a foundry, it needs to be a lot cleaner than that. Yeah. Yeah. So they've been trying to reuse as much of their water and recycle it. To strip all of the copper and other residues from that wastewater, from their fabs and recycle it. So when I say 63 megalitres of water a day, well, yes, but a large percentage of that is actually recycled water. So it's not like they're sucking down the entire local reservoir supply. Otherwise, I think all the workers might struggle because they kind of need water to to live? Yes. Yeah. The whole living thing. Yeah. The biology. Food, water, shelter. Yeah. Biology. Like water's kind of important, you know. Darn the biology. Yeah. That's the thing. Semiconductor manufacturing is super water intensive and even with purification and recycling, it still uses a lot of water. And it's kind of intriguing if you just take a look at where some of these things are located and then take a look at global weather trends. And it's hard not to imagine a scenario in which it will start to affect semiconductor manufacturing at some point. Now, how much, I don't know, but it feels to me like it's got to at least be a factor when building new plants, thinking about existing plants, thinking about ways of dealing with, you know, upcoming water shortages, because it's not going to get less, it's only going to happen a lot more. Absolutely right. One of the things that I thought was fascinating is that if you actually look at the distribution of the actual foundries and fabrication capacity, like in 2021, so we look at China, Japan and Taiwan. China is about 18% of global capacity, Japan 16%, Taiwan 16%. That's 50% across those three countries alone. What do they have in common? Well, those foundries are all within about a thousand kilometer radius, so 620 mile radius of each other. Admittedly, the center point of that is the middle of the South China Sea, I think, but the point is that that area there is really not that much of a... They're all very close together geographically from the perspective of natural disasters. So, droughts is just one example, but you've also got cyclones or hurricanes, call them typhoons, they call them typhoons in that part of the world, whatever. You know, but then you've also got tsunami, which, you know, tsunami are not necessary. Earthquakes. Yeah, because I mean, you're so Taiwan and Japan are historically both heavily impacted by the Pacific plate and all of the earthquakes leading to tsunamis. And not to mention problems like Fukushima, which is another story altogether, but tsunamis can be bad and they're not good. So China, perhaps not so much of a history of that, but it's still that they're in that one area. So that's half of the global production in that area. So if you could have a massive typhoon go through there, and as you said, with weather patterns going the way they are, that's going to be more regular, more frequent, and extremes of weather with droughts and such, it's going to be a longer term impact. So when they do build more foundries, if people were saying to think about, "Well, the more we rely on technology, the more we rely on semiconductors and such, we need to start distributing that to different parts of the world. Otherwise, we're putting ourselves at risk of being unable to meet supply when there is a demand." And it's not just the fabs, it's also the manufacturers of the fabrication equipment that we were talking about earlier and the test equipment. A lot of those are Japanese. And I assume in China, a lot of those are Chinese. And so things that are based in Japan, where all the brain trust is in Japan, a lot of the equipment is made in Japan, or presumably in China as well, then you're gonna be impacted if they have some natural disaster that impacts those companies. And then not only are you potentially affecting your fabrication, but maybe you just can't even get the equipment you need anymore. - Yeah, exactly. So just circling back to TSMC again, just for a second, in terms of their 200 mil wafer capacity and fab count between 2020 and 2024, they are gonna spend something like $4 billion over those years by adding 22 new fab plants with approximate capacity increased globally of 17% as a result for 200 mil wafer. Actually, I said TSMC, I actually didn't mean that's the whole, that was across the whole industry. So they have basically, they were planning to increase because of the demand before COVID hit. And then of course, when COVID hit, the wheel started falling off. So one of the things I did was have a dig into like, okay, so we've talked about countries, we've talked about different wafer sizes, but what does that look like in terms of companies? Because the thing that I keep seeing TSMC coming up, but just because TSMC is a foundry or pure play foundry, they call it a merchant foundry, just because of that. And it's being the largest in the world in terms of what it produces doesn't mean it's actually the largest fabricator. So the largest fabricator is actually Samsung, and they've got 14.7%. Yeah, they're gigantic. I mean, a lot of people think of Samsung for televisions and fridges and so forth, but they do a lot of semiconductor. Yeah. And they do a mixture of their own and they do have, they do, do some contract manufacturing as well, if I remember correctly. So yeah. So it's like, they're massive, right? So they're in South Korea, they're South Korean company. So of course, TSMC 13.1%. That's, they're in Taiwan. And obviously that's, you know, that's, then we talk about them a lot. Next up that's Micron, which is in North America, 9.3%. And then, yeah, which is interesting. And Microns, I'm, yeah, I don't know. I'm much about Micron to be honest. So I'll stop there before I embarrass myself there. But SK hynix, which I'd never heard of before, South Korea at 9%, Kyoxia WD, which I thought, I think that's Western Digital, sorry. I didn't have time to dig into what the WD was. I think it's Western Digital. Anyway, that's Japan, 7.7%. Our old favorite Intel in the United States of America, which I don't have a percentage for, but this is the next five in the top 10, followed by UMC in Taiwan. Then we're back to the USA, Global Foundries and Texas Instruments still number nine and then SMIC in China. So that's the top 10 companies involved in semiconductor manufacturing globally in terms of like the top 10 manufacturers. So- - I actually didn't realize that Texas Instruments was still manufacturing their own. - Apparently so. Yeah, they're still going. So the thing that I also, I did know this sort of, I didn't know some of the terminology like pure play. I thought that was kind of a bit of a funny name, but I've always referred to them as contract foundries or contract chip fabs, but whatever you want to call them, merchant foundries, pure play foundries, contract manufacturing, whatever, doesn't matter. TSMC was actually founded in '87, and that was a spinoff from the government's Industrial Technology Research Institute at that time. And as I said before, they don't actually do their own design. They just manufacture to other people's designs. But of course, then we talk about, you know, we talk about Samsung. So they would fall into the bucket of being an integrated device manufacturer. Another example of an IDM is obviously Intel, Texas Instruments, another example. And then of course you got the other ones that are what some people call them fabulous, which I really don't like that as a term because it sounds like they're somehow lesser. It's like, you know, it doesn't, it's not really, that's not really fair there's some pretty impressive names in that list like Qualcomm's an example. They don't manufacture their own chips. AMD don't. NVIDIA don't either. Right. And it shouldn't fool people into thinking that they're somehow exempt from the manufacturing constraints. They're relying on other people. And so they're every bit as constrained as anyone else when it comes to manufacturing crunches. Exactly right. And that's the thing to remember is that you can be really, really good at manufacturing, you know, Silicon wafers and so on and so forth, doesn't mean you're going to be really, really good at designing a motor basic, for example, like Qualcomm, I might be very good at, or a GPU like NVIDIA. So putting those two together is not always a natural fit. And I think companies like Texas Instruments and Intel, they are the way they are simply because of that's the way that they have been from the beginning, because they had to do it that way, because companies like TSMC didn't exist at that scale. And they've learned to do both. So it's kind of, you know, so they can keep on going and being integrated device manufacturers. So I think it's good that there's a bit of a mixture, because it means that people can focus on what they're best at. And some companies are lucky to have both. Yeah, definitely. Yeah, it's an expensive game. And to be honest, what TSMC has done is super impressive. But if everybody in the world relies on TSMC, you really talk about too big to fail. That's... Yeah, no. You don't want TSMC to go out of business. No, because that's exactly the point is that if everyone is so dependent upon like TSMC to get all this stuff fabricated, then if it goes away for whatever reason, then there's no way the rest of the world can compensate with capacity for a long time. Yeah. Even if they had the money, which will be billions of dollars to build a state of the art fab, then it takes years. So exactly. Takes years to spin that up. And then, yeah, it's, it's, and then you have to have all kinds of expertise that you don't have. Presumably you could suck people in from the no longer existent TSMC or whatever. But yeah, it's just, it's not easy. There's a reason why foundries exist and they specialize in that is because that is a problem that's best offloaded to experts if you need to. Yeah. Yeah, no, it's not easy. So ultimately, sort of trying to sort of wrap this up a little bit is to talk about the problem is quite pure and simple, a supply chain problem. And the way that different companies have handled this in the past has been, and I think Apple's probably the best example that's, I think, well known, certainly the audience of Pragmatic. Certainly, you know, Apple famously bought, I think it was 10,000 or thereabouts, you know, CNC milling machines for their unibody MacBooks. So I think what- Yeah. It's like they were so worried about the supply chain and contract manufacturing, they said, you know what, we're not going to take a chance, we're just going to buy 10,000 CNC milling machines. Like, I mean, you got a couple of billion, trillion, however many dollars in the bank, why not? Let's just go and buy 10,000 CNC machines. That's- To me, that's insane. Like, I mean, I've programmed CNC machines, you know, and, you know, as in like the controller and then the control that the, and the server drives and so on and so forth, the positioning, the cutting head and cutting it selection, all that very cool. Actually one of my favorite jobs, but that's not the point. The point is that 10,000 of those things, that one of them is massive, you know, like it's yeah. Anyway. So, but that's, that's what Apple did because I was so worried about the supply chain. They even, I think they then a few years later, or maybe it was around the same time, they bought the, there's a company that could do laser drilling in the aluminium, the cases to get like 20 micrometer holes. If I remember correctly, that was for the speaker grills. I think I'm trying to remember what it was for, but there's like this one company that could drill holes that small, like you can't even see them. You look at it and it's like, you can tell there's a slightly different color of aluminium there or aluminium, however you want to pronounce it. But they bought the company. They're just like, yep, we'll buy you because they didn't want anyone else to have access to it. So they bolstered their supply chain. And that's all to isolate themselves from supply chain disruption. So Global Foundries, that's one of the American companies that's in the top 10, for example, manufacture. They're now doing something similar insofar as they're enforcing upfront payment. So previously, you'd have a contract, it might have 30, 45, I don't know, 60 day terms, whatever terms and conditions. And you'd say, "Well, I, Apple, will place an order with you." And they're like, "Yeah, no. Give me $2 billion up front or something. I don't know, whatever the number is. I'm sure it's more than $5. But you're going to pay up front now because we've got so much demand." And so these companies are like, "Well, if I can't invest in my own equipment like Apple, because I've got a massive amount of money in the bank, I still have to find some money now to lock in a contract just to get my chips made." It's crazy. There's a whole lot that that goes on in these partnerships, and I have a feeling the benefits of buying equipment and installing them into your partner's facility are multiple. Number one, the partner has incentive to work with you because you're funding a lot of what would have been their cost. You also have knowledge of the equipment and you can make sure that you know how it works and that it can do the job that you want it to do. And then again, like you said, the supply chain problem. not going to run out of those pieces of equipment because you bought them all. Similarly with foundries and I know with test facilities because one of the things that's interesting to me is that like TSMC will send its devices that it fabricates for other companies to yet other companies to be tested and so forth there's a whole lot of interplay going on here and almost I think all of the fabricators, all of the IC fabricators are doing some level of foundry work, some amount, whether it's their entire business or a small part of their business. And then those who have, you know, good testing facilities, they're doing all kinds of stuff. They're doing their own, they're doing other people's stuff, and there's a whole lot of interplay. And the more cooperation you can give in working on designing your stuff in such a way that that vendor can fabricate it and or test it in a way that works for them. And they also have to work with you to, to, you know, do those things in the way that you want it done. There's a lot of interplay and the more cooperation you can give them, the more likely you are to have your products expedited or prioritized over other people. Definitely. Some of that co-located stuff's really interesting too, because I know that so Tesla for some of their battery production, for example, did something similar. Now it's not technically ICs, but that is definitely a model that can work because you get that control. And they even had people from Panasonic, from MemoryServs actually coming in, running that equipment. So it was kind of like a, yeah, yeah, yeah. We'll give you floor space. We'll help set you up, but it's still a part of the plant that's technically not Tesla. So that's, that's, that is another really good way of doing it to ensure that, that your supply chain is not disrupted in situations like this. So, Basically one last comment about about TSMCs component for automakers. And then I just want us to like go and circle back about the whole secrets of events that got us to where we are. So in 2020, only 3% of TSMCs overall production was ICs for automakers. So it's actually not a very big part of their business, for example. And 50% or thereabouts of their sales were for smartphone manufacturers. So that's kind of where the money is. And so when all this chip shortage happened and then, 'cause one of the things that I'm talking about automakers a lot, and there's a reason for that, is that recently I've been, you know, my wife's just got a new car. And one of the things that we came across is we sat in with a dealer and said, "Hey, what's that stack of paper on your desk?" And he said, "Well, this stack of paper is the, it's the unhappy sad, it's the sad list. These are all the people that have ordered a car and still don't have it, that I have to call and tell them that they're still not going to get it until January of 2022. And so I said to him, hey, just out of curiosity, what's the oldest one in that pile? And he pulls the one out on the very bottom and he says, April 2021. Wow. So it's like, that's insane. And all of this has come from a sequence of events that have led to a chip shortage, a genuine chip shortage. and it's not the chips that you would think. And it's kind of been unheard of in the automotive industry because traditionally automotive industry has been not as heavily relied on ICs. So I find that fascinating. And I find it fascinating because if people talk about the context of, well, I can't buy myself a computer right now, or I can't buy myself a GPU right now. So the GPU has been strong demand for GPUs for cryptocurrency, which we mentioned. Yeah, you've also got, if I can't buy a computer, well, a computer is almost entirely silicon. You know, there's very few parts in a computer that isn't. Yeah, okay. I'm not talking about like the plastic case and the, you know, whatever, sorry, the aluminum unibody or anything like that. But it's like the supply chain for that is well and truly understood and it has always been something where people have understood, you know, that if I, like Apple, like I'm creating a product, it is heavily based with these components. Therefore, I will secure my supply chain. I'll have diversity in my supply chain. I'll have stockpiling if I have to, to ensure that I can make products. But when it comes to the automotive industry, a lot of them were completely unprepared for this because they never stopped to think about their supply chain for integrated circuits. And that's why I find the automotive angle really interesting is because, well, first of all, it affected us 'cause that was the first time that we'd bought a car in 20 years where we were told, "Come back and see us next year." Yeah. So that was a bit of an eye opener for me. And that's one of the reasons it spurred me on to dig into this one. But it's not just that. To me, it's like understanding your supply chain and understanding all this other stuff that's happened with COVID-19 has had a ripple effect. And it's fascinating why. So just to go through it from the top. So the pandemic hits, all travel shut down or it's severely curtailed. That meant less people are driving their cars. So there's less demand for new cars. Automotive manufacturers then reduce their production rates and then they start to cancel some of their IC orders for integrated circuits. And now that freed up some capacity at the foundries, which they then redirected elsewhere. Cause they're a business, they got to stay, you know, they got to keep making money at the same time, more people stuck at home. So there's more demand for home electronics and computers work, work from home equipment and so on. So that then actually pushed those foundries to stop making integrated circuits for that market instead. And that took over the capacity that was left for the automakers. And then of course, those same lockdowns were restricting capacity at the existing foundries. So they already had blown out production schedules, and which only really started to recover in the last six months or so, but, you know, they also were impacted from that. And the growth of the demand in the automotive, in cars on average, was increasing as well. So there was already not enough production capacity to meet the demand at that point before COVID happened. So it was already very, very tight. And so whatever capacity that did exist was being then used for high margin items that are not automatic specific integrated circuits Because ultimately they're too low margin and to your point from earlier a lot of them were legacy And so some of the legacy equipment they just that was the probably the final trigger for them to just push it to the side And say right what we're gonna put in more profitable You know fabrication tech in these areas now to make better use that space. Yeah, and it's possible too for all I know some of the people that had the most stability or some of the remaining players in the Legacy nodes as Tim likes to call them Tim Cook It's possible that some of those companies disappeared - It's hard to say. I don't know but I'm just wondering I'm just curious if some of that infrastructure went away and Really the only people who could weather the storm are the people that aren't focused on those products because they've got overhead they've got expensive facilities, they've got a lot of people and they're not into making commodity products because they have to make money. For sure. So ultimately it's supply and demand and with the increased demand and the restricted supply, prices are going up, things are taking longer and it's interesting that there's some of the strategies different automakers have used, for example, like Tesla, for example, now it's in Q1, they actually rewrote their firmware to use different microcontrollers to try and broaden their base of supply just in order to deliver cars to keep their production going, which is, you know, just an example of how far they were prepared to go. Not all automotive manufacturers could or would do that. But even Tesla is running out, like there's a limit to how far you can go before you just need the chip. It's like you can't write your code out of some of these chips, for example, to go somewhere else. So I think that one thing she said earlier to Scott is that it's these facilities cost billions of dollars and lots of planning. They take people with a lot of experience to run. It's not easy being a foundry. It's almost sounds like a song title, but anyway, it's not easy being a foundry. Next thing you know, your pickup will die and your dog will run away. Yeah, pretty much. So anyway, and so that's going to take time to fix. And I honestly think that people saying, well, there is a global chip shortage and it's had a ripple effect across industries that never thought there would be a ripple effect. And I can't see it ending quickly. I think it's going to be probably mid to late next year, maybe even 2023 before we see that capacity finally extend to current demand levels. And so I just can't, you know, as some people saying, oh, it's only going to be for a few more months. And it's like, yeah, but yeah, this is now where are we at 18 months in? I mean, it's not going to get fixed anytime soon. It just you can't just fix this problem quickly, I think is the is the point. Right. Because even if let's say automakers and I have heard that some of the automakers are looking at trying to control their own destiny with regard to semiconductor manufacturing, which in and of itself, You know, what exactly does that mean? How do they do that? That's a huge problem in and of itself, but then even if they Start changing their designs and getting more modern with the devices. They're putting in their vehicles you're still forever going to have things like Some of the stuff qualcomm makes that yeah, they make cpus, but they also make uh wireless radios those are never going to make a lot of money and you know things like that that everybody needs that aren't going to be put in the system on a chip probably people are going to have to make they're going to be cheap devices who's going to make them there's always going to be some stuff that's even if it's a more modern process or a modern device it's still not going to be cost effective for a lot of people to worry about unless they can just make billions and billions and billions of them so i don't know i mean i think that there are still going to be some constraints on things that everybody needs and it's not necessarily tied to the age of the design or the process that it's on. Yeah, it'll alleviate it somewhat, but I think you're right. I don't think this is an easily solved problem. No. And one of the other things I just I think it's worth mentioning as well is that it's kind of, it's not all, it's not all reality. Like, people are saying, Oh, well, we can't give you the fridge next week because of the global chip shortage. And it's now being bandied around as a bit of an excuse. And it's It's like, well, actually no, it's not because of the global chip shortage, it's because you don't, like the chips might be plentiful that you need. And in fact, it's just that there's this one plastic hinge that you guys didn't think to order enough of. And that's just a supply chain, poor planning, prior planning prevents piss poor performance, that kind of thing. And in their case, they didn't manage their supply chain. So now some people are just jumping to that as an excuse. What I find difficult though, is separating fact from fiction when they do that, because in the end, practically everything's got an integrated circuit in it these days. So it's kind of hard to say if you've got a 1960s Ford Mustang, then yeah, sure, you probably don't have any chip shortage there. Mind you, it's 2021. So I don't know why I'd be buying a brand new one of those that's now what, a 50, 60 or 70 year old car? Oh my God, time flies. - Yeah, you've probably got a lot of other parts on that car that you're not gonna be able to get first. But it does bring up a good point though, John, because like, I don't think we're done with pandemics forever. And you were talking earlier about the automotive industry being very hands-on. Well, so is currently the semiconductor industry. And so if, people have to go to work. It's just how it works. These are still, there's no lights out fabs. There's no lights out testing yet. It's a goal that the semiconductor industry keeps talking about, but there is no such thing. And so, yeah, anytime people can't go to work, it also affects, in addition to all the other problems, it affects the ability to manufacture these things as well. - So you're suggesting we need to get better at managing pandemics? - I think so. It might be something we wanna think about a little bit. - Yeah, we should probably think about that some more, hey. But I absolutely agree with that. And there are, whilst I'm sure there will come a time in the future where it's possible for a robot of some kind to do the same job that a human can do at pretty much anything. We are far from that. And, you know, fully automating, you know, a chip fab, you know, like even fully automating a car production line is currently not possible. As far as you can push it, there are still people needed. So yeah, it's not a, yeah, but that's okay. Right. I kind of liked the fact that people are still needed because, you know, that's, That's good. I'm not entirely sure I like a future in a world where people aren't needed to make everything. It's kind of, I don't know, it feels wrong. It's like, what are the humans going to do? Like stare at each other? Right. Like if the planet doesn't need us, I mean, we are part of the human race, so I guess we have a vested interest in the human race continuing. We bring a lot of baggage to the planet, too, though, let's be honest. No argument there. Alrighty. Anything else you want to mention while we're on the topic or should we wrap it up there? I think we should wrap it up. I think it's just a fascinating, complex topic. And like you say, it's not as simple as people think it is. And then it also does point out the, uh, the nature of our reliance on just getting these devices into products, getting them made and how well it shows you one of the unintended and not thought about side effects of shoving chips into everything that we make. Yeah, exactly. I remember I did an episode of this show years ago called "The Internet Makes Everything Better". So to just put a slight twist on that, putting an IC in everything makes it better. Does it really? Does it? Anyway, that's all right. Well, you know what? If you want to talk more about this, you can reach me on the Fediverse at email@example.com, on Twitter @johnchiji, or on Word, or the network @engineered_net. I'd personally like to thank ManyTricks for sponsoring the Engineered Network. If you're looking for some Mac software that can do ManyTricks, remember to specifically visit this URL, manytricksallonword.com/pragmatic, for more information about their amazingly useful apps. If you're enjoying Pragmatic and you want to support the show, you can by supporting our sponsor or by becoming a premium supporter. We're edging closer to our monthly goal to to go advertising free across the network. But we can only do that with your help. You can find details at engineer.network/pragmatic about how you can help this show to continue to be made. A big thank you to all of our supporters, a special thank you to our silver producers, Mitch Bilger, John Whitlow, Kevin Koss, Shane O'Neill, Oliver Steele, Leslie Law Chan, Hafthor, Jared, Bill, and Joel Maher. And an extra special thank you to both of our gold producers, Steven Bridle, and our producer known only as R. Pragmatic is a podcasting 2.0 enhanced show. and with the right podcast player, you'll have episode locations, enhanced chapters, and real-time subtitles on selected episodes. And you can also stream Satoshis and Boost with a message if you like. There's details on how with a Boostergram leaderboard on our website. Now, if you'd like to get in touch with Scott, what's the best way for them to get in touch with you, mate? - Oh, probably on Twitter. Scott A-W is the Twitter handle there, and you can shout at me and I'll stand there and take it. - Awesome, cool, very good. Well, a special thank you to all of our supporters and a big thank you to everyone for listening and thank you very much, Scott, for coming on the show and lending your insights into this complicated one. - Thank you, John. That was a lot of fun. - Thanks, man. (upbeat music) [Music] [MUSIC PLAYING] ♪ ♪ ♪ ♪ (upbeat music) ♪ ♪ ♪ ♪ [Music] (dramatic music)