← Software Engineering Daily
Software Engineering Daily
TSMC and the Geopolitics of the Chip Industry with Tim Culpan
00:47:37
Taiwan Semiconductor Manufacturing Company, or TSMC, produces a significant portion of the global supply of advanced semiconductors. Its cutting-edge technology powers everything from smartphones to high-performance computing,
Loading summary
Transcript68 lines
- [00:01]
Tim Culpin
Taiwan Semiconductor Manufacturing Company, or tsmc, produces a significant portion of the global supply of advanced semiconductors. Its cutting edge technology powers everything from smartphones to high performance computing and its Customers include Apple, Nvidia and Arm. TSMC's dominance in chip production has made Taiwan a critical player in the global tech supply chain, drawing attention from major economies like the US And China. This has escalated global tensions with concerns over the stability of Taiwan and the potential risks to the global tech industries if chip production were disrupted due to regional conflicts. Tim Culpin is an independent technology journalist and the author of the forthcoming book the World's Smallest Inside the Rise of tsmc, Foxconn and A Nation of Taiwan Technology Titans. He's based in Taipei and has been covering the semiconductor and electronics hardware industry for 25 years, including 18 years as a journalist and column at Bloomberg. Most recently, his work can be found@timkolpin.substack.com Tim has written extensively about TSMC and recently broke news on the developments at TSMC's Arizona factory. He joins us today to discuss what's happening at TSMC and what that means for the U.S. taiwan and China's chip industries. This episode is hosted by Sean Falconer. Check the show notes for more information on Sean's work and where to find him.
- [01:39]
Sean Falconer
Tim, welcome to the show.
- [01:40]
Tim Culpin
Thanks Sean. It's great to be here.
- [01:42]
Sean Falconer
Yeah, thanks for being here. I'm excited to chat about, you know, I think chip manufacturing, geopolitics and so forth with you today. But let's start with an introduction, like who are you and what do you do?
- [01:53]
Tim Culpin
I've been living in Taiwan for 25 years. I'm now an independent journalist. I was at Bloomberg for 18 years, the technology correspondent and columnist. And now I'm writing a book which will be published a year and a half from now called the World's Smallest Superpower. It's about the rise of Taiwan and why it's so important in the global sphere of technology. So that's really what I'm up to right now. I'm substacking like so many other people. Timkolpen.substack.com is where you can find me. That's the plugin. And as we're talking about today, the chip industry. The global chip industry is a huge focus of what I'm looking at right now because it's such an important topic for everyone.
- [02:27]
Sean Falconer
Yeah, absolutely. And I think it's something that's been more on my radar in the last couple years, hearing things in the news, also listening to other podcasts and people talking about this. But in terms of tsmc, Taiwan semiconductor manufacturing company, it plays this really essential role in the global chip production, particularly for companies like Apple. And I think if you're not involved in that world or you haven't really given any thought, it's kind of, it's easy to overlook like the position, sort of dominant position that it has. And like, how did that happen? And like, why is it so crucial to global tech? You know, supply chain essentially.
- [03:04]
Tim Culpin
It was never inevitable that TSMC would be the world's most dominant or most important chip company. It was founded in 1987 and intel was huge back then. AMD were huge back then. Even IBM was big in chips back then. And the idea at the time was to do something called Pure Play Foundry, purely make chips for somebody else based on their designs. It was a relatively new idea and they decided to do that in Taiwan through the Taiwan government. And then TSMC spun off later. And it's done very well because it has been solely focused on customers. The remit and the goal of TSMC from its founding has been to never, ever compete with their own customers. And that early on didn't seem like a big deal. But today it really does matter. It means that TSMC can make chips for Apple and AMD and Qualcomm and Nvidia. And actually, ironically enough, it makes Intel's leading edge chips because it only cares about its customers. It's not competing with its own customers. And as it's done so, they've worked very, very hard. The Taiwanese work ethic is crazy. The Fabs run 24. 7 as they do in most parts of the world. The R and D also works 24 7. You have R&D engineers working at 3am trying to work out the recipes of how to make the next leading edge chip. It's a very, very tough work ethic and it's very, very focused on keeping track of data. Chip manufacturing is very data intensive. A lot of AI and machine learning involved in chip manufacturing. And so they've done that by being very, very, very consistent to the point where they overtook amd. They've overtaken intel, they've overtaken everybody else. And all the other chip manufacturers around the world have just fallen behind because TSM just keep advancing beyond even the work ethic.
- [04:46]
Sean Falconer
I would think a company that's been solely focused on manufacturing chips for such a long period of time, surely they've also built up certain even process expertise that is hard to recreate and duplicate because you don't even know necessarily what you're copying because all you're seeing is the end result of a chip or something like that. From an outsider's perspective, that's really true.
- [05:07]
Tim Culpin
Every new node, which comes approximately every two years thanks to this concept known as Moore's Law, is built on the previous node. So you can't really just start halfway through. You need to know how the previous node was done, how the equipment was used, what the recipes and formula and parameters were for the equipment for the previous node. Because when the new one comes along, that previous knowledge has to be used so you can move forward and keep moving the baton forward. That's why it's very, very difficult for anybody to catch up to someone like tsmc. And if you're someone like intel or AMD and even Samsung now, if you fall behind a little bit, you're really going to struggle to catch up because it's a very fast moving industry. Software is probably the only other industry that moves as quickly as semiconductor manufacturing in that it really iterates so quickly.
- [05:53]
Sean Falconer
And the US introduced the chips acts to try to reduce the reliance on foreign semiconductor manufacturing. How does something like that potentially affect TSMC's dominance in chip manufacturing?
- [06:06]
Tim Culpin
I think it won't change a lot. It has basically enabled TSMC to set up in Arizona, expand in Arizona. And TSMC has been very, very, very clear from the start. When the US government started wanting to kind of seduce them to America is, you know, show me the money has been TSMC's line. You want us to set up there, you have to help cover the cost of doing so. And under the CHIPS Act, a lot of money is being spent by US taxpayers as well as the Arizona government at the local level to get TSMC to set up in Arizona. But at the end of the day, the good stuff, the tough stuff will still be done in Taiwan. Because the way the chip industry works is kind of two teams, broadly speaking. There's the team that work out the recipes, kind of like the head chef or the master chef in a kitchen. They work out the recipe of how to make something, how to make the pizza or the souffle. And once they've worked that out, that's the PhDs, that's the people who have specialist technology skills in, in maybe chemistry or in physics. And once they work out those recipes, then they pass it on to the operations team. And the operations teams are the ones who have to put that into practice daily. Running 247 in a factory, you know, a minute of downtime is not acceptable in one of these factories because they're so expensive. And so what we're going to see is more of the operations will be done in Arizona, but it'll still be maybe 5 or 6% of TSMC's global capacity. But working out those secret recipes will still be done in Taiwan because all that previous knowledge, generations of knowledge, sit in Taiwan and that can't be transplanted easily. So we will see more capacity outside of Taiwan, such as in Arizona, but it won't change the general dynamics of Taiwan still being the hub of R and D and future chip development.
- [07:47]
Sean Falconer
Yeah. Also you talked about the work ethic of the Taiwanese and also a part of the world where what you have to pay someone is significantly less than the us so if all manufacturing or a good portion of manufacturing shifts to the US then how would that impact even in the cost of manufacturing the ships themselves?
- [08:06]
Tim Culpin
Well, what's interesting about the chip industry is TSMC doesn't dominate today because of lower wages. Yes, Taiwan wages are lower and engineer is cheaper in Taiwan than they are in the us but the biggest cost of semiconductor manufacturing is the equipment. Every year it's the depreciation of the equipment. Because TSMC spending $30 billion in 2024, they'll spend even more than that next year and it becomes redundant equipment within about five years. They have a very fast depreciation schedule, and so the actual wages of the workers won't be the big factor. There are other costs in the us There's a lot of licensing and red tape and environmental standards and things like that. The process and the bureaucracy in the US is a lot, lot slower than in Taiwan and that slows things down. That makes it very, very difficult for the US to compete, mainly because of red tape. And the other thing is the US doesn't have the clusters of suppliers. There's a lot of chemicals, there's a lot of other parts to the puzzle of making a chip. There's probably 2, 300 suppliers to TSMC right now. Some of them will set up in the US and follow tsmc. A lot of the others won't. They won't be able to afford to. And so the cost will go up as a result of that as well.
- [09:15]
Sean Falconer
With all the suppliers, where are the bulk of those suppliers stationed around the world.
- [09:20]
Tim Culpin
They're mostly in Asia. Japan has a very robust industry of equipment as well as materials. Probably the most important equipment company is a Dutch company called ASML, which used to be part of Philips 30 years ago and it spun off. But the US is a very important part of not just fabrication, but other parts of the chip puzzle. Most of the value of a semiconductor today is still coming from the US. There is design tools. The software that is used to design chips is crazy, leading edge, and very, very complicated and very, very expensive. The US Dominates that. They also have incredibly good materials makers and equipment makers in the us so some of the most important equipment makers are American companies. It's just that the actual process, the final part of making the chip, is not done in the US as much as it used to be. So we do have a supply chain around the world. But because of the 30, 40 years of history in Taiwan and in Asia, Japan and Korea, of course, are very good at making chips. You naturally have both the Asian companies and the Western companies clustered around these few hubs of manufacturing in Asia.
- [10:23]
Sean Falconer
And in terms of China, they've made significant investments into semiconductor sector, trying to catch up to the likes of the us but they still lag behind in advanced chip production. What are the primary barriers for China?
- [10:37]
Tim Culpin
A lot of it is to do with trying to catch up because Moore's Law works so quickly. The Chinese companies like smic, Semiconductor Manufacturing International, which is probably the most famous and most leading edge at the moment, they have caught up to where TSMC was six or seven years ago, and that's a pretty good feat. They've done very, very well. But because the chip industry isn't sitting still, it's always moving forward. It's not good enough to be where your competitor was five or 10 years ago. You have to be where they are today. And the reason why they're struggling is because they don't have the right engineers. They don't have the focus and attention to detail that the Taiwanese and the Koreans and the Americans have in terms of chip manufacturing. And the problem that we're going to see going forward is that China will keep spending a lot of money on trying to get its chip industry up and running. It's not a new thing. China has been spending a lot of money, government money, for more than 25 years. And that money has in a way gone to waste because they haven't caught up. And if that's Beijing's goal is to catch up to the west, which I include tsmc, then they've kind of failed. But they do have a lot more capacity of older nodes, the stuff that may be 5 to 10 years old, which is still used a lot, a lot of chips actually come from that area. You don't need old chips to be Made at the very leading edge. And so the reason why China is struggling to catch up is they don't have the engineers, they have a government policy right now that really doesn't reward entrepreneurship or innovation or risk taking. And if you want to catch up in something like semiconductors, you have to take risks, you have to try something new. And there's not a lot of incentive for an engineer from China to stay in China. If they've got a really good pedigree, like a good degree and good background and PhDs, they probably want to leave and go to the US and work at a US company rather than stay in China. So that's going to be a real challenge for the Chinese going forward.
- [12:24]
Sean Falconer
Yeah, so it sounds like there's essentially some built in cultural challenges around, like not really fostering a culture of like innovation. Whereas like the US that's certainly not the case. Like the US is very, I think, focuses a lot on innovation. That's why so many startups come out of US based companies and so on. So what is the US Doing in terms of strategic moves to continue to essentially have control over chip manufacturing and continue to work with companies like TSMC to be on the cutting edge and make sure that China is not catching up?
- [13:00]
Tim Culpin
The US has a kind of a carrot and stick approach. The carrot is the Chips and Sciences act, which is to throw a lot of money at companies to set up in the U.S. tSMC is a beneficiary, intel is a beneficiary. Even though it's American, you know, they're getting money to set up. Samsung is also doing that and quite a few other companies. And not just the fabrication companies. There is companies like another company called SAS Global Wafer, which makes the silicon, the actual platters of silicon, which is so important. That's a Taiwanese company. And they've got money from the US Government or will get money from the US Government to set up in Texas, in Sherman, Texas to make those slices of silicon. And so that's bringing a lot of the supply chain to the US it's not going to be large compared to Taiwan or Korea or Japan or even China, but it's helpful. The stick approach is the U.S. commerce Department basically restricting the access that the Chinese companies will have to American technology. And that's not just American technology manufactured and made by Americans, such as American equipment and software, but American technology that might happen to appear in equipment made by a Dutch company or a Taiwanese company or a Japanese company. And the US has the ability to do that, to say all Right. You are not an American company, but your equipment, your product uses American technology. You're not allowed to sell that to the Chinese. That's actually been surprisingly effective. I know that in the US especially in dc, there's a big debate, oh, it hasn't worked, it's been a failure. These kind of things don't work. The reality on the ground is that it has been effective and it'll take a few more years for us to notice the effect, because those restrictions only came in around 2022. And given the way the industry develops, it takes a few years for restrictions to really take hold. It's not going to stop a truck immediately on the line. It's going to slow it down. It's like a massive parachute slowing down development in China. It is effective. Maybe not as quickly as policymakers would like, but it is working. So the carrot and stick approach is working to slow down the Chinese as well as help the Americans catch up.
- [15:06]
Sean Falconer
And then with tsmc, like setting up shop in Arizona and the us has there been any reaction from the Chinese to that?
- [15:14]
Tim Culpin
Not happy. It emboldens the Chinese government's belief that Taiwan and America are in cahoots, that America is just trying to prop up Taiwan. So from a political point of view, it does give Beijing some talking points, but they also know there's not much can be done about it. I know that there has been some propaganda and misinformation and disinformation campaigns out of China to try and sow discord between the American and the Taiwanese workers in Arizona. There's been some issues with labor relations and so forth in Arizona over the last year or two. And I know that part of that has been Chinese disinformation campaigns on social media to try and make it look like the Taiwanese disrespect the Americans using racist terms, trying to make the Taiwanese distrust the Americans and think the Americans are unworthy, unreliable. But I talk to people on the ground in Arizona, and yet there is a clash of cultures in a way between the Taiwanese and the Americans. But they generally get along quite well. They're focused, they're engineers. At the end of the day, anyone who's an engineer of any type, there's a lot of internal office politics, but at the end of the day, you're trying to get the job done and ship a product, and that's what they're focused at in Arizona. So little cultural issues between. The way Taiwanese do business, Americans do business is really quite small compared to the fact that they're all really on the same team, trying to get fabs up and running, operational. And they're really focused on trying to hit the same kinds of levels of manufacturing and yield that you're seeing in Taiwan. And they're doing that. And so that's really boosted morale amongst the Taiwanese and the Americans working in Arizona.
- [16:49]
Sean Falconer
And in terms of tension between China and Taiwan, if that tension continues to increase, how would that potentially impact the global tech industry?
- [16:58]
Tim Culpin
Well, I think the big issue really is that Taiwan is the world's smallest superpower, as I posit it, and it's beyond just chip manufacturing. If we take just the fabrication of a chip, there's a lot that happens before that process and a lot happens after. When a chip comes out of a factory, whether it's in Arizona or in Shinju or in Dresden, Germany, it has to be tested. Then it's sliced up and it's packaged with a kind of a layer of enamel, and it's got wires put to it, and then it's put on a PCB and all that process. And a lot of that will still be done either in Taiwan or by Taiwanese companies. Foxconn is not as famous, I guess, as TSMC right now, although it has been famous because it makes iPhones. They're as important to the global economy as tsmc. There's other companies like Pegatron and Quanta, which really, most people don't know, but they make a lot of the AI servers, a lot of other smartphones and other devices, and they're all Taiwanese companies. So if anything happened to Taiwan, these companies would struggle to function. And so it's not just semiconductors that would be at risk. It would be the whole hardware supply chain. The AI servers made for Nvidia are done by Taiwanese companies in Taiwan. And so the semiconductor industry would definitely suffer. And the Chips and Sciences act will try and ameliorate that by having more capacity in the us but all the processes before and after making a chip are also at risk. If anything befouls Taiwan is that the.
- [18:27]
Sean Falconer
Main motivation of the chip stack is to try to essentially create a more resiliency to any kind of disruption in the global supply chain. If there was something that happened between Taiwan and China.
- [18:40]
Tim Culpin
Yeah, I think the Chips and Sciences act is it's very much a product developed by politicians. Right. That's reality. With a lot of backing from the US Semiconductor association that's got their hand out wanting money from the government. And yes, the idea is to have more resilience and security in the us but it's for only one part of the puzzle. And I think the issue really is the dependencies. I guess in software terms you could secure totally this one step of the puzzle, but the dependencies on other parts of the hardware stack are still there in Asia. And so it would ameliorate and soften some of the issues if something went wrong in Taiwan and the US would make some chips at TSMC in Arizona. But all the other parts of the puzzle are still out in Asia and the US needs that at the end of the day as well.
- [19:33]
Sean Falconer
In terms of the Chips and Science act, what are the main takeaways in terms of the restrictions or policies put in place by that?
- [19:42]
Tim Culpin
Essentially, the Chips and Science act is giving away money through tax breaks and various other non cash incentives as well as some cash incentives such as cheap loans or low interest loans or zero interest loans to companies that set up in the US and you don't get all of that money at once. You have to hit certain benchmarks and targets. So every company goes to the US Government and says, all right, we want some of this money. Here is our plan, this is what we plan to do. We plan to set up by this date, get production going at this date, get to this production node at this date. And as they hit each of those stages, the company gets a certain amount of that incentive back to them. So it's spread out over a period of time. And so the Chips and Sciences act is really about doing that to incentivize, but it's not enough to make it worthwhile for a company. A company still has to basically spend its own money and find its own clients to make it worthwhile. And that's what TSMC is trying to do.
- [20:35]
Sean Falconer
Are there other Western countries that are trying to follow, like, you know, put similar things in place to try to protect against disruption in the global supply chain?
- [20:44]
Tim Culpin
Yeah, the EU has a very similar policy to the us so the EU broadly as a union, has policies. And then individual countries within Europe are also enacting some of their own policies. Germany has been successful in securing TSMC to set up in Dresden. In Germany, which has already been a hub of semiconductor activity, AMD owned fabs in Dresden before they sold it off when AMD split 15 years ago. So that has been successful. Those factories that TSMC will set up will not be leading leading edge. It's actually focused on the automotive industry, which is slightly older technology, but it's the technology that the auto industry needs. Japan has also been successful in getting TSMC to set up a Factory in a place called Kumamoto. That factory is up and running very, very soon. We'll actually see production out of it in 2025 and for similar reasons, the auto industry as well. Electronics industry is very big in Japan and a lot of the chips that will be made in Japan will be for those sectors. It won't be for AI servers or iPhones. It'll be more for cars and other industrial electronics.
- [21:50]
Sean Falconer
Introducing height, the only autonomous project management tool. Backlog grooming, bug triage, keeping documentation up to date. Those aren't why you got into product building, right? Well, height handles all that grunt work for you. Using a first of us kind AI approach, height proactively takes care of time consuming workflows without you lifting a finger. Height recognizes when you've agreed to trim scope and Halo's mapping the necessary edits back to your product brief. When new tickets are added to your backlog, hike combs through them, adding feature tags, time estimates and more. And it's not just you. Everyone on your team manages projects, tracking updates, scoping work, balancing priorities. But whether or not your product succeeds shouldn't depend on project management. With height, autonomous workflows handle that mundane upkeep so your team can focus on building great products. If you're ready to stop managing projects, it's time for height join the new era of product building, where projects manage themselves. Visit Height app sedaily to get started in terms of like, the bleeding edge of chips, like, it's part of the challenge. Like, even if someone in China got ahold of, you know, the latest chip, just basically, if I understand it correctly, the challenge is that even if they could reverse engineer it, by the time they could put all the things in place to actually manufacture it based on the process of reverse engineering it, they're already two or three years behind whatever that chip is.
- [23:09]
Tim Culpin
Yeah, you've nailed it, Sean. That's it. That's exactly what it is. The US has been ahead in many industries. Aircraft technology and engine technology and even nuclear weapons and all sorts of other things. And other countries caught up because these are not technologies that move that quickly. Like a car today is kind of similar to a car 50 years ago. The combustion engine hasn't changed a lot. The basics are there. The Italians have a Ferrari, it's a better engine, and the Americans have Detroit and so forth. But essentially a car is the same. Chips are very, very different. A chip today made at Leading Edge by TSMC and Shenzhou is quite different, has a lot of technologies, a lot of machinery that were not even seen five years ago. One of which, for example, is extreme ultraviolet lithography technology coming out of Dutch company asml. Nobody had used that machinery five or ten years ago. Nobody had seen it. And so they get this equipment and they've got to work out how to use it. And it doesn't come with an instruction manual of saying, here's the recipe. TSMC has to sit down and work out the recipe. Every single step in making a chip, there's maybe 30 or 40 steps. Every step will have maybe 50 to 100 parameters. The humidity and the temperature and the air pressure and the amount of time that it's turned on. Like we're talking milli. Milliseconds. All of those parameters are crucial to making that step. And then you multiply that out. The multiplier effect means it's very, very difficult. So even if a person gets their hands on an EUV piece of equipment, doesn't mean they know how to use it. It's very, very difficult. So, as you say, if the Chinese were to get their hands on this kind of equipment or work it out or reverse engineer it, everybody else has moved down the road and they've got to try and start that process again. That's why it's very, very difficult, because the industry moves very, very quickly. It's very hard to catch up because reverse engineering is not enough in terms.
- [25:00]
Sean Falconer
Of figuring out those parameters and the process around that, as well as whatever the designs are that are going into the next generation of chips and stuff like that. How much protection essentially is in place to prevent people from getting their hands on that ahead of essentially the implementation of those things already being in existence.
- [25:20]
Tim Culpin
Those parameters are really the secret sauce of how to make it ship. Every single tool has its own set of parameters and a dedicated team of engineers. You could be an engineer at a TSMC fab and know this one tool inside out and have no clue how to operate the next tool in the process. That's how specialized it is. And this information is highly, highly secretive. It's very closely guarded. You, for example, cannot go into a TSMC factory with any electronic equipment. You're not allowed. Thanks to Apple's rigidity with security. The printers, the tsmc, have paper that is embedded with small filaments of metal so that as you exit the TSMC facility and go through a metal detector, if you are taking out any paper that has been printed at a TSMC facility, the metal detector will go off. That's how strict they are on security and TSMC Even has a kill switch. If a FAB was taken over, they can press a button and literally just delete all the information and kill it. So that you could get access to the equipment, but you wouldn't get access to the secret source of how to use that equipment. And that is the key. That's why companies that have access to the same equipment are not able to do the same thing because they haven't worked out the secret formula, haven't been able to get access to the secret formulas of how to use it.
- [26:42]
Sean Falconer
What percentage of people working at TSMC would even, even has like a full view of the entire production line and has a good grasp of how all that stuff works together?
- [26:53]
Tim Culpin
Each fab and there's more than 20 fabs, it's just a word fabrication plant facility. There's more than 20 of these. Each fab has a fab manager or two and they have the big picture macro view of it. But I guess like in software engineering, if you look at a product, even a product manager or the engineering manager probably doesn't know the intricacies of every little line of software code or every little package within the code. They might know how a kind of all fits together because they'll map it out. But it's similar in semiconductors. There'll be a fab manager who has an overall view. There'll be process technology engineers and then line engineers who have an idea of when it comes out of this piece of equipment, the chip needs to be fitting these parameters. Kind of like say an API and software. Once it comes out, you know, we'll do a test at this stage and if it hits these parameters, then we'll move it on to the next stage. If not, then we work out if there's a problem. So there's not a lot of people who have a broad enough picture and a deep picture. You've either got a broad picture or you've got a deep picture. You probably don't have both. And it's not just because of security. It's just pure intellectual capacity. Because the industry moves so quickly, it's not possible for anybody to have that depth of knowledge and breadth of knowledge at the same time.
- [28:05]
Sean Falconer
How many people are working in one of these fabs? What's this look like if you walked into one of them?
- [28:11]
Tim Culpin
Well, if you're actually inside the clean room of a FAB and you see a lot of people, you know there's problems. That's reality. They are kind of very low human occupation facilities because you don't want More humans because they introduce dirt and grime and these are clean rooms. So most of the people in a fab are not actually in the clean room, clean room part of it. They're just kind of in the periphery looking at computers, checking parameters, checking what's going on. It's a lot of just checking parameters and looking at screens and so forth. It's not someone going in and physically picking up a wafer after it comes out of the oven and then moving it to the other part of the process because it's very hands off, because it works very quickly and because of the desire to not introduce any particles of dust. But a fab itself will have a few thousand people working in it. And these people all have their specialized role for a specific piece of equipment or a specific part of the process. And they're running 24 7, so they have shifts. There's generally fewer people working at 3am than 3pm, but they are working 24 7.
- [29:15]
Sean Falconer
What is the typical process for someone to become a specialist in a fab? Like are they going and taking specific training or is this something that TSMC actually provides training to have people always available to have these skills?
- [29:30]
Tim Culpin
Generally you do need a college education in an aligned kind of engineering degree. It's not necessarily electronic engineering or semiconductor engineering, but it might be chemical engineering, it might be physics, because at the end of the day, semiconductors are a mold of physics and chemistry. You might be a mechanical engineer of a specific type of mechanical engineer and there'd be skill set required. More and more people inside a chip fab are actually machine learning and AI engineers. If you're a data scientist, you've probably got a job waiting for you at TSMC before is a lot of data scientists are needed. But you know, the US is now getting ready to ramp up its education process to feed tsmc. Arizona State is working very, very closely with tsmc. ASU is churning out engineers and right now in the fabs at Arizona are former asu, you know, ASU grads. The Taiwan Education System is two or three major universities in Taiwan that are very much geared towards creating the kind of graduates that companies like TSMC need. You can get a job inside a fab without having an engineering background, but it probably won't be a leading edge engineering role. It might be something a little bit aligned with it, but there's definitely a lot of roles in the sciences and it's not necessarily directly semiconductor engineering.
- [30:45]
Sean Falconer
And in terms of the US with the kind of export controls that they've put in place to restrict semiconductors going into places like China. Are there ripple effects to other countries looking to the US or even those that manufacture chips like Japan or South Korea and other parts of Asia?
- [31:04]
Tim Culpin
There is a ripple effect in terms of the ability of countries and companies to sell equipment to China. Asml, for example, the last couple of years did really, really well in China, ironically enough, because the Chinese could see that the door was closing and the Americans were going to cut them off. And so the Chinese were rushing to buy Dutch equipment. And that kind of peaked just recently. And next year, 2025, 2026, the Chinese will buy a lot less equipment from, from overseas. For that reason, Japan was also impacted where Japanese were not allowed to not able to sell a lot of the equipment that they otherwise might have wanted to sell to the Chinese. So there's definitely been a ripple effect on that side of it. The ripple effect on the other side of it in terms of manufacturing in China is because China will be crimped in its ability to move to the leading edge. It will be stuck kind of in what we call a legacy node or a mature node. And these are nodes that are actually, really, really needed. Not everything, as I said, needs to be leading edge. A lot of chips that we see and use today in our cars or in our computers can use technology that was developed 10 years ago. But if China builds up a large capacity of that kind of technology, then we're going to see a flood of capacity available and they'll be selling off that capacity to anybody at really, really cheap rates. And so we could see a massive gluttony in mature node capacity in China. And if you're a Western or a non Chinese company in that area, you're going to struggle because you're going to struggle to really be price competitive.
- [32:35]
This episode is brought to you by WorkOS. If you're building a B2B SaaS app, at some point your customers will start asking for enterprise features like single sign on skim provisioning, fine grain authorization and audit logs. That's where WorkOS comes in, with easy to use and flexible APIs that help you ship enterprise features on day one without slowing down your core product development. Today, some of the hottest startups in the world are Already powered by WorkOS, including ones you probably know like Perplexity, vercel, Brex and Webflow. WorkOS also provides a generous free tier of up to 1 million monthly active users for user management, making it the perfect authentication and authorization solution for growing companies. It comes standard with rich features like Bot protection, MFA roles and permissions, and more. If you are currently looking to build SSO for your first enterprise customer, you should consider using work OS. The APIs are easy to use in modular, letting you pick exactly what you need to plug into your existing stack, integrate in minutes and start shipping enterprise plans today. Check it out@workos.com that's workos.com in terms.
- [33:41]
Sean Falconer
Of like chip manufacturing that's going on in China. Like as you said earlier, it's not necessarily the latest Apple chip, but there is a lot of chip manufacturing that still happens there. What kind of industries I guess do those chips end up being in?
- [33:55]
Tim Culpin
We forget where we use chips today. Your credit card has a chip in it, right? Smart card reader. It might even have two. One is the NFC and one is the little gold chip that is encrypted. A lot of that stuff is made in China. A lot of the dull, boring chips that go into a microcontroller controls the window in your car going up and down. These are really, really important chips, but they're not leading edge. They're pretty mundane. They're commodity chips. And we kind of forget how often we are interacting with semiconductors on a day to day basis. And so China can churn out a lot of those chips like by the bucket full, don't even need to price it per unit. They price it per 10,000 pieces, a few cents per 10,000 pieces. And that's where we're seeing a lot of Chinese made chips. We're also seeing China getting somewhat advanced in memory technologies like DRAM and nand and they will be probably used in Chinese server farms or Chinese electronics. A lot of it's going straight into electronics made by Chinese companies and staying within China. A lot of Chinese made chips never leaving China.
- [35:01]
Sean Falconer
Is there any concern over China creating or building a lot of these more commodity chips?
- [35:08]
Tim Culpin
There is concern because they could flood the market. We could see if you're a European company that is making smart card chips or NFC chips or even Bluetooth chips which are pretty commoditized. Now Bluetooth has been around a while and when it first came out there was a few companies that could do it. And if you could do Bluetooth, you really, you got a lot of contracts. But Bluetooth, it's an open standard and the process of making a Bluetooth chip is pretty well known. And so if you're a non Chinese company in those areas, you better have something really awesome to offer your clients in addition to that, like some ability to build a product or integrate it into something else. But if you're just purely making that one chip that the Chinese can do, you're definitely going to struggle. And there is concerns outside of China, in Europe, in the U.S. in Asia, that the Chinese could flood the market in a lot of these commodity chips and bring the prices down and companies could go out of business as a result.
- [36:05]
Sean Falconer
The biggest concern there is around from a competition standpoint of them being able to own that market less than it being any kind of like, I don't know, potential threat to security.
- [36:15]
Tim Culpin
There is a security issue. It's very, very difficult to work out whether chip has secret kind of properties or functionality built into it. In the same way, it's very hard to find bugs or backdoors in code. We know they exist. Code does ship with bugs and backdoors and we do our best to find it. It's a lot harder in a semiconductor than it is in code. Right. First of all, you have to know what you're looking for. And if it's hidden in such a way that you don't have some kind of static or dynamic test to find it, well, good luck. And those who've shipped it know how to access that backdoor in semiconductors. It is not science fiction to believe that there are backdoors in semiconductors that are being shipped out of China. It is happening, it is real. It's very hard to find. They may be happening right now, already shipped. And that's one of the reasons why the US especially the three letter agencies in the US want to have more of their chips made in America so that these backdoor functionality won't be built into them.
- [37:14]
Sean Falconer
Other documented cases of backdoors being found in China built chips there are, that.
- [37:19]
Tim Culpin
I think are not public or not public enough that we could really pinpoint it. I believe that US Security intelligence agencies are aware of them or are concerned enough about them to believe that they exist. And so yes, there's feasibility has been proven by industry and by academia that it's feasibly possible. And if you're a paranoid security agency, then if it's feasible, you know, it's probably been done. And so yeah, there is, there is documented cases among security agencies. I haven't seen those documents. I know that they exist. But there is definitely, you know, cases out in the wild of backdoors built.
- [37:52]
Sean Falconer
Into semiconductors in terms of some of Apple's processors being made in the US out of Arizona. Do you think that that's something that other major companies like AMD and Intel or others will follow suit.
- [38:07]
Tim Culpin
Yeah. So it was inevitable that an American company would be the first client of TSMC Arizona. I didn't know who it would be. My guess was it'd be Nvidia, Apple or amd. And I broke the news that it was Apple. It's the A16 processor, which by the time it comes out will be maybe two generations old. It'll probably go into the next iPhone SE, which is kind of the lowest spec iPhone that Apple brings out every couple of years and might go into some other products. Since breaking that news, I since learned that AMD is next, the next cab off the rank. AMD will make high performance computing processes, which may be AI chips. Not all high performance computing processes are AI chips, but they're in the same general league. I believe that's a bigger deal because, you know, AI is a bigger deal than a smartphone chip in terms of the ability to what, what they can do in the world. And it shows that the US is getting towards a certain amount of AI hardware resiliency and independence. Because once it's then packaged at amcor, which is an American company also setting up in Arizona, once it's packaged, and then it could be sent over the border to Mexico to be put onto a printed circuit board and put into a server. North America, maybe not the US specifically, but North America is getting very, very close to independence in AI server manufacturing. And that's I think, a very big.
- [39:28]
Sean Falconer
Deal for the U.S. you know, looking ahead, I guess over the next five to 10 years, like where do you think this world is going in terms of chip manufacturing? Are we going to send up, you know, I think moving a lot of the manufacturing out of, out of Taiwan to other countries. Like given all the geopolitical tension between China, us, China and Taiwan, how is this going to change over the next few years?
- [39:52]
Tim Culpin
There is definitely more manufacturing made or being built outside of Taiwan. Arizona, Dresden and Kumamoto and Japan are the three main areas that we can think of. But it's worth remembering that as TSMC breaks ground in those three places, it's also building new facilities in Taiwan. It's not sitting still in Taiwan. It's still spending significant money. And Taiwan's a small island and so they're rushing around the island trying to find a few spare hectares to build a factory. And it's not that easy in Taiwan, but they're doing it. And another area where they're spending a lot of time and energy is building leading edge packaging. There's a technology called cowos chip on WAFER on substrate. And that essentially combines old style chip packaging with the semiconductor lithography. And it's very, very important because that allows the core GPU from say an Nvidia Blackwell to be put right next to the high bandwidth memory, which is basically dram. And having that closer is very important for speed. It's like having your warehouse next to your shop, right? If you have to go down 10 miles down the highway, deliveries are going to be slower. So semiconductor engineers are trying to make the chip smaller, combine all the parts of the chip closer together. And so packaging technology is hugely important. So there is going to be more capacity outside of Taiwan, but there's still going to be most of the capacity in Taiwan. Most of the know how and most of the leading edge capacity for packaging will be in Taiwan. So the actual balance of power will not change that much. No matter how much money other governments spend, Taiwan is still the home for tsmc and it's never going to change.
- [41:30]
Sean Falconer
How is the fact that like Taiwan over the last, you know, 35 years or whatever has become this dominant power in chip manufacturing. Like how has that affected like other industries in Taiwan? Because I would think that in order to get the chips out of the country and lots of other things, you know, even just supporting people work, like thousands of people working in one factory, you have to build up a lot of other industries to support the machine.
- [41:56]
Tim Culpin
We have a labor shortage in Taiwan. Unemployment rate is very, very low. Taiwan imports a lot of labor and really needs to import more. So the structure of the Taiwanese economy has changed a lot. As you say, TSMC is the largest company in Taiwan. It's the largest member of the Taiwan Stock Exchange. And then anything affiliated to tsmc, whether it's, you know, there is companies who just exist to build the factory shell. That's all they do is build the factory shell. They do it very, very well because TSMC has built a new factory every year and they make their money just from that. So you've got whole engineers, like construction engineers and architects who set up their whole industry and their whole career around building a factory. And they're really good at it. You've got chemical suppliers who are very good at not just, you know, maybe making the chemicals, but delivering it like in specialized trucks and hooking it up and all that kind of stuff. So you've got a whole lot of specializations that are built up in Taiwan as a result of Taiwan semiconductor industry. Not just tsmc, there's another company called umc. There's quite a lot of Other companies in Taiwan that do make semiconductors. And that cluster effect has been hugely important. But as a result, TSMC and its clusters of suppliers have kind of sucked up all of the workforce. And they hire a lot of people. They hire a lot of university graduates, mostly out of engineering degrees, but also a lot of social science and liberal arts graduates are going to these companies as well to do, whether it's marketing or market planning or all sorts of other information sciences work. And as a result, young graduates out of universities, they don't want to. They don't want to go and work in a Starbucks or they don't want to go and work in traditional industries like plastics or automotive. They want to go where the action is, where the money is, because TSMC employees, they get paid quite well, nothing compared to the us, but they also get shares and so forth if you're senior enough. So it can be very lucrative to work for these companies. And it's changed the whole structure of the Taiwan economy, for sure.
- [43:48]
Sean Falconer
It's really interesting to think about the sort of ripple effects within the country and all the side businesses, basically the one that just builds factories.
- [43:58]
Tim Culpin
It's like San Francisco. If you go to a bar in San Francisco, you bump into someone you've never met, you just ask, which startup do you work for? Right. It's like everybody works for one of the big software companies or Internet companies or is working on a startup. It's similar in Taiwan. Everybody is one or two degrees separation from someone who works at TSMC or supplies to tsmc. Yeah.
- [44:18]
Sean Falconer
And I think that's why when those industries suffer a downturn, a lot of other industries that maybe on the surface level seem unrelated also suffer a downturn. Like, if suddenly there's less tech industry in San Francisco Bay Area, then that means, like, less people are spending money at restaurants and bars and so forth.
- [44:36]
Tim Culpin
Yeah. What's interesting in Taiwan, though, is TSMC has never had massive soft cuts. Taiwanese companies don't cut staff. They hire very gingerly, very carefully. There was one period of time when Morris Chang, the founder of tsmc, was not in charge and his replacement did cut staff for the first time ever. And the CEO was cut next and he lost his job and Morris Jung took back the mantle of the company. So it's kind of an unwritten rule that you don't cut staff, be very careful about your hiring, but when there's a downturn, you kind of grin and bear it and you get through it.
- [45:09]
Sean Falconer
Is that a TSMC thing or is that a, like a Taiwanese cultural Thing.
- [45:14]
Tim Culpin
It's broadly a Taiwanese cultural thing. You don't have the massive hiring like you have in Silicon Valley, in the Facebooks and the Googles and so forth, where Suddenly you hire 20,000 people in two years and then you go, oops, we have too many people. Let's cut them. The Taiwanese just don't do that. Taiwanese are overworked. Any Taiwanese will know that their team needs three more people, but the boss is not going to give them that. So you got to get it done yourself. And as a result, when there is a downturn, those people keep their jobs because they still need it. So there's not a lot of extra staffing at a Taiwanese company. Or the upside to that is the downturns don't really hit as hard for the workforce.
- [45:53]
Sean Falconer
I think this is really fascinating, Tim. Is there anything else you'd like to share?
- [45:57]
Tim Culpin
I guess the key thing really is that if the US Wants to build resilience and security in semiconductors, they need to think beyond semiconductors. You need to think of the whole kind of hardware technology stack from the materials all the way through the final product, and understand that the US Is not going to make it all themselves. They shouldn't try and make it all themselves. It's not a dis of the United States economy or workers that you can't do it everything yourself. Global trade is what makes the US Strong. And so the best thing that the US and its policymakers can do is work with allied countries like Korea, Japan and Taiwan and say, well, let's work on this together, make sure we have security and resiliency. But it also means that if Taiwan falls, the US Is in trouble. That's just the reality. It's a big thing to kind of go to defend another country, but the reality is the US Would be in very, very, very bad shape if something bad happened to Taiwan. And that's a reality that we all have to kind of deal with.
- [46:54]
Sean Falconer
Yeah, absolutely. I mean, I think it's not just the US that would suffer. It'd be all Western countries.
- [47:00]
Tim Culpin
It would make Covid look like a hiccup. Right. It would make it look like a small sneeze. What happened with COVID it would be cataclysmic for the global economy.
- [47:09]
Sean Falconer
Well, Tim, thanks so much for being here. I think this was really fascinating. I think you shared a lot of details about sort of the history, what's going on, and, you know, what potentially the future looks like.
- [47:20]
Tim Culpin
Thanks, Sean. If listeners have questions, they're most welcome to reach out. Cheers.