Toy factories to TSMC: Taiwan’s transformation

Taiwan’s journey from low-cost manufacturing to high-tech leadership is nothing short of remarkable. In this episode, Catriona Macnair shares first-hand stories from her recent trip, exploring how the island became essential to AI, semiconductors, and the future of global innovation.

Transcript

Nick: Hello, this is Nick Robinson from Aberdeen and you're listening to the Emerging Markets Equity Podcast. Together with our expert guests, we'll dive into the driving forces behind emerging markets and uncover the opportunities that are shaping the future of this exciting region. So back in the 1980s, made in Taiwan meant cheap toys, bicycles and electronics. Fast forward to today, and that same label powers the digital age. Taiwan has become the world's high-tech workshop, home to Taiwan Semiconductor and the only place in the world that's capable of producing the most advanced chips that drive artificial intelligence. Taiwan's transformation over the last 30 years has been incredible, and my colleague, Catriona Macnair, just returned from an investment trip there. So today I thought we'd get an update from her about the trip. With all the AI hype and concerns about a bubble, this trip was particularly timely, so I'm delighted Cat has joined to share what she learned from visiting those companies that are closest to the action. Cat, welcome back to the podcast and thanks for joining. I think the last time you were on, we were talking India, which was which was a great chat, particularly as I had an India trip just coming up at the time. So it was very useful for me. So hopefully today can be just as useful.

Catriona: I hope so. Well, thank you very much for having me back on again, Nick.

Nick: Now, well, it's great to have you on. Yeah, how about we just kick off with a bit of an overview about the trip in terms of the feel in Taiwan and the companies you met and what you got up to there?

Catriona: Sure. I spent a week there with one of our colleagues from the UK-based emerging market equities team here at Aberdeen. He leads on emerging market utilities, and we work together on our impactful emerging market equity fund. So we're both non-tech sector experts. And the idea was that we'd spend a week seeing companies in Taiwan to further our knowledge and understanding of global semiconductors, memory and AI supply chain. So very topical, as you say, given China's deep-seek moment earlier in the year, and then the mind-blowingly large CapEx numbers we're seeing announced out of US-based companies of late. Taiwan today makes up about 20% of both the emerging market and Asia-Pacific ex-Japan indices. And more than half of that relates to Taiwan Semiconductor or TSMC, which alone is about, I think, 11% of both benchmarks. It's the only name actually that's above 10% in either benchmark. So the largest company in emerging markets and in Asia. We saw more than 20 companies over the course of the week, and I'd say at least 15 of those operate within the tech hardware space or are very correlated with TSMC and heavily reliant on tech hardware thematics. On the first day, we were in Taipei and we ventured southwest a little bit towards the west coast of the island for a day of bespoke meetings. And although, as I say, TSMC is the big index heavyweight, it actually wasn't on our formal schedule because Aberdeen had a call with the company a matter of weeks before. But we managed A 10-minute detour and we saw their big sprawling campus, the fabs, the sort of Heathrow-sized R&D centre, which is about as exciting as it gets, I think, on holdings bingo for emerging market equities fund managers. Not every day you see the headquarters of the biggest company that's in our index. The rest of the week was spent mostly in and around Taipei, combining more bespoke meetings with a small scale market access event. And then we finished with a group tour of a data center on the Friday. And it was punctuated a little bit by the weather. So the early part of the week, the inevitable jet lag battle that you need to get through, but also we had some fairly extraordinary storms. There was a super typhoon in the Philippines. that had been slowly making its way up towards Taiwan and towards Taipei. So Monday, the wind was sort of howling around some of the meetings that we had. On the Tuesday, I think we grossly underestimated the rain. And then actually on about the Wednesday or thereabouts, the storm changed course. So I think a few companies we met or were speaking to were hoping there might be a typhoon day where they got to just have the day off or day at home. But it didn't play out like that. The storm changed course and Thursday, Friday all went ahead as planned. But a bit of a reminder in terms of some of the physical climate risk that we see within the region.

Nick: Right. I mean, that sounds like a very exciting trip on a number of different levels. What was the, I suppose, the probably top of most investors' minds at the moment is this idea that there's an AI bubble going on, just the amount of CapEx that's going into semiconductor hardware and all the associated parts of building data centres and that kind of infrastructure. what was the feel on the ground regarding that?

Catriona: Regarding the bubble risk?

Nick: Yeah.

Catriona: I think that's probably the $1,000,000 question, or if you're thinking about Nvidia and Microsoft, more like the trillion dollar question. So in most of our meetings, our meetings certainly with companies where we have a long established relationship, we'd start most meetings with that exact question. And the clear takeaway is that there is no indication of a bubble on the ground. So I think they're very used to investors asking that question, but certainly from the tech hardware supply chain side of things, 2026 looks very firm based on our trip findings, also based on rising use cases and habitual use of artificial intelligence applications at a consumer and industry level. But it's harder to have that same degree of confidence into 2027, I think, at this point. TSMC certainly remains very confident in the meetings that we had prior to the trip. And therefore, some of those tech supply chain hardware players will be equally certain of the 2027 future and beyond, but we didn't necessarily get that evidence on the trip. There are no signs of inventory air pockets, stock sitting unused, no inventory burden. But that's not to say there aren't challenges. So one semiconductor advanced packaging company we met actually said they're asking their customers to give more notice so that they can start pre-booking materials like substrates or fiberglass, etc, to book capacity so that they have the materials to be able to deliver on time. So that gives some indication of some of the pressures that we're seeing in terms of the manufacturing side. And then also, if we think about the sort of power and thermal management side of things, order visibility there actually has deteriorated in some pockets fairly significantly. So one company we spoke to gave an example of cloud service providers in the States shipping vast liquid cooling systems, these things weigh over 100 kilograms on a weekly basis on chartered flights. Now, the Taiwanese company doesn't fit the freight bill. It's those US businesses that will. And that perhaps gives you some indication of their degree of confidence around the projected returns that might warrant that spend or that freight cost.

Nick: Right. There's, I mean, it's very interesting because I, you know, there are no signs of any shortage in terms of demand. And certainly all this equipment, when it is bought, it's immediately being deployed. And lit up, so to speak, and being used. So I suppose the next question is, at some point this equipment needs to be paid for and a return needs to be generated on this huge investment. And did you get any sense of that in terms of the kind of next step in terms of the actual monetization and return generation capability that companies are having from this?

Catriona: Yeah, we saw companies across the tech supply chain from equipment operators and testers to niche componentry manufacturers, assembly operators. And during a trip like this, it's quite interesting often to include a few standing items in each meeting to help gauge the range of corporate views on a given issue. And there were two such questions where we got nowhere really. And namely, will hyperscalers returns on their CapEx be sufficiently exciting to justify and sustain the spend. And then also how much of a risk is circular vendor financing. We just didn't get a clear answer from any company really to either of those questions. So bottom line, we don't know yet and nor do they. I mean, from their side of things, exactly as you say, Nick, they get the orders, orders are coming in, they manufacture, manage their inventory risk, deliver on time and execute in the hopes that there'll be more repeat orders. So done from their side. I mean, two things for us probably to keep in mind. The first would be that the current CapEx that we're seeing from these US hyperscalers and Nvidia, etc, is funded from free cash flow. So providing the hyperscalers are still profitable, still throwing off cash, it isn't an issue per se. And then the other thing to keep in mind is perhaps that there's still a degree of wiggle room. So if we look at OpenAI alone, where I think they've announced total deals in excess of 500 billion US dollars, only about 20% of that has actually been implemented. So there's still quite a lot of wiggle room for the remaining 80% to tweak and adapt as we and they get a clearer sense of monetization coming through.

Nick: Right. Yeah. And I mean, it is interesting, isn't it I mean, I think for years, if you were a US equity investor, you'd be looking at these US tech companies thinking, gosh, look at all this cash that's being built up. And finally, we're actually seeing it being deployed. How about, I suppose there's a very clear kind of tech supply chain that, is benefiting from all this investment and, it's very clear that Taiwan Semiconductor is, the winner here in terms of being able to make all the chips that go into AI. When you look kind of further down the supply chain, are there kind of any interesting trends or things you picked up from talking to these companies in terms of things we should be looking out for as investors or the next possible opportunity?

Catriona: Yes, I think this is a great point and actually shows the importance of continually doing your homework and the importance of having sector experts, which myself and my colleagues certainly aren't, but I feel much more up to speed after the trip. I mean, the launch of Google's latest AI model called Gemini 3 actually happened two or three days after our trip. So it didn't feature at all in any of our meetings. And of course, if we were there now, that would be front and center in terms of some of the discussions you have with the company. So during our week, front of mind were topics around liquid cooling, certain capacitors, the move to the Rubin series with NVIDIA in 2027, but certainly not Gemini 3 and not really so much around TPUs, which is Google's chip or Google's technology. Thinking around liquid cooling, I mean, that is really critical infrastructure to try and avoid overheating of these AI chips, servers and data centers because connectivity essentially deteriorates when things start to get too hot. We also had some quite interesting discussions around tantalum capacitors, where you can have capacitors that are multi-layer ceramic, film, aluminium, or tantalum. And tantalum actually can store much more power for the same voltage and able to discharge that more quickly. So it lends itself very neatly to AI applications. But I think as we approach the launch of NVIDIA's Rubin series in 2027, there are certain other implications that we need to be keeping on top of. So if we think around that liquid cooling technology, for example, there will be a shift away from what we're seeing now, which is microchannel lids, which are essentially a heat spreader that goes on top and it's able to have a large surface area to help cool things down. That just becomes much less practical when you're looking at Nvidia's Ruben. So we'll need to move towards different thermal management systems like microchannel cold plates, which is MCCP. And likewise, in terms of power requirements, they're very different power densities. So in terms of some of those power solutions providers, we will need to be looking towards high voltage direct current solutions. So that's HVDC. So those two themes came up quite a bit in our discussions. And I think that's where it's really important to have sector leads who are looking at some of these issues on a daily basis, who are able to map the product roadmaps that we see from the likes of NVIDIA, Microsoft and Google, etc, keep us up to speed with some of the evolving tech supply chain requirements. I mean, I think the Google Gemini launched two or three days after we got back is probably quite a good example of how quickly these things can move and evolve. 

Nick: Yeah, no, you're right. I mean, a lot of this stuff goes over my head as well in terms of trying to understand exactly what's going on. But thank God for the analyst. How about the change tack to talk a bit about the data centre tool. That sounds really interesting. And can you just kind of describe what it was like? And I have these visions of kind of data centres that you might see in films or something, but kind of wondering what they're like in real life.

Catriona: Yes, so this, it caught me a little bit by surprise because I had imagined a sort of low lying, out of town, big sprawling industrial estate and actually not a bit of it. This was in a reasonably central location when a Taipei's more convenient industrial parks. And the idea being that it's close proximity to all essential transport hubs, resources, customers, and keep sort of transport disruption cost to a minimum and reduce latency. So this building was actually 7 stories tall, 5 above ground and two below. And the two underground, the first was to deal with sort of really big chillers for that liquid cooling of thermal management that we spoke about. And the other subterranean floors to deal with seismic isolation and rubber bearings to try and provide earthquake protection, which is very imperative in a place like Taiwan and like Taipei. What I thought was quite an interesting analogy when we were looking around the data center was to think about it like a hotel. So the site owner typically wants to select customers that are willing to pay for that room upgrade, might eat in the restaurant or shop in the gift shop for a souvenir, that sort of thing, rather than last minute, no frills sort of budget holiday makers. And you could liken the cloud service providers perhaps to that higher end customer and perhaps the Bitcoin miners to the no frills budget end of things. So typically you want a data center would prefer to have hotel guests or customers that are more willing to pay for additional bells and whistles. I thought that was quite an interesting takeaway. Unfortunately, the AI server rooms with Amazon's AWS and NVIDIA's GB300 series were off limits during our tours. They're considered red areas, so we weren't allowed to see those. But we did visit a couple of the traditional server rooms, which brought some of the earlier discussions about power supply, rack sliders, liquid cooling infrastructure, etc, all to life. And also the need for actually really reliable, weight-bearing infrastructure in the building. I hadn't really had a sense before that actually a traditional server rack would weigh sort of about 500 kilograms, whereas Nvidia's CEO had talked about AI server racks being more like baby elephants, weighing up to 1,200 kilograms. You've got to cart these things around the building, so get them in a huge lift and then have floors that are sufficiently robust to be able to hold these things sustainably. The top of the building or the roof of the building had various power infrastructure, so uninterrupted power supply in case there was any sort of power outage or power cut, anything to do perhaps with an earthquake or some other event. And then also gas-based power generators, and both are tested every month to make sure that the backup works and the backup to the backup works. The data center we saw is targeting 50% renewable power by 2030 and to be fully renewable based by 2040. So that also has implications that we need to think about given intermittency risks that we see with some renewables and perhaps a need for energy storage solutions. So there are other second order effects that we maybe need to think about around the AI boom and countries perhaps with access to cheap, reliable or innovative power solutions are perhaps more likely to succeed in the new AI landscape than those without.

Nick: Right, I mean, that's an interesting take. And certainly, when I think about the various portfolios we run, we have quite a lot of exposure, not just to the picks and shovels, but also, the supply chain, which is much broader outside of Taiwan in terms of energy infrastructure and even uranium miners, which at some point are going to play quite a big role. What was, I suppose, thinking about Taiwan itself and the impact that all this money that's flooding into the economy. Are you seeing any signs that the broader economy is picking up because of the investment that's coming in at the moment?

Catriona: That's a difficult question to answer. I think we spent, as I say, sort of 75, 80% of our time on the semiconductor supply chain, and that still is very much the key graduate job for many university graduates out of Taiwanese universities. But I think perhaps what came across most pertinently from the trip was resource scarcity. So rather than the impact, if you like, on the broader Taiwanese economy, it's actually more some of the challenges that these businesses face within Taiwan. And I think around power, water and labour, certainly we came away with quite clear concerns around that. So that the state-owned utility company Thai Power is facing pretty pronounced labour constraints. And actually we heard from some privately run industrial power companies that Thai Power is having to leave some products in the factory because they simply don't have the manpower to be able to collect it. They're in the process of hiring talent from neighbouring countries. So I mentioned the Philippines a couple of times. So that should help to alleviate those bottlenecks in early 2026. But it is a bit of an issue for the broader economy outside of semiconductors. We heard repeatedly actually during our meetings, that power needs for Taiwan's semiconductor industry and TSMC in particular are prioritized above all else. So the country's tech sector industry hasn't and won't be impacted, but that could have a lingering impact for domestic Taiwanese businesses or indeed residents. I mentioned that semiconductor manufacturing remains the most sought after industry for top graduates, but there's been really fierce competition for talent. And there was some discussion that staffing costs have been creeping up in pockets, tech hardware names trying to deliver sufficiently attractive competitive packages, long-term share rewards and options as well. There's been a bit of a move towards automation, but actually that doesn't necessarily help in terms of staffing costs. You still need technicians and engineers who don't come cheap. So there is a bit of a sort of staffing issue to be aware of as well across Taiwan. And then lastly, in terms of broader implications for the economy and for resource scarcity would be around water. So we saw our fair share of rain, certainly during the trip in the early part of the week. And I think that probably exacerbated the point that much of Taiwan's rainfall actually comes during typhoons. So you get really heavy bursts of rain that go through short, fast flowing rivers. And that causes some of these rivers to temporarily burst the banks. But actually in terms of storage and reservoirs and things, it's fairly limited within Taiwan. And that's very much at odds with its reliance on the semiconductor industry, which is notoriously water intensive and thirsty process requiring ultra pure water for some advanced processes. I think consuming the equivalent of a small city on an almost daily basis, just looking at TSMC alone. So some mitigation efforts are well underway across the country, investing heavily in water recycling, reclamation and even some desalination to alleviate water stress. But that comes back to that reliance on power and the energy that those sorts of processes will be using. 

Nick: Okay, so it all sounds like there might be a little bit of inflation to come perhaps at some stage?

Catriona: That's right.

Nick: In terms of some of parts of the economy. And I suppose you can't really talk about Taiwan without asking the big question about China and existential risk that Taiwan faces, given its proximity to China and China's kind of long-term desire to unify. Is that something companies talk about much, or did you get much of an impression of that feeling while you were there?

Catriona: Yes, I mean, as you can imagine, it's a highly sensitive topic to try and discuss on the ground in Taiwan. So we didn't raise it at any of the meetings. We had some discussions sort of on the way out of company meetings or over slightly more relaxed mealtimes where there was an opportunity to discuss the security threat a little bit more informally. What came across really, and I think we've spoken about this before as a team, is certainly that Russia's invasion of Ukraine in 2022 really highlighted that long-held geopolitical tail risks. can become a reality pretty much overnight. And although Taiwan is a veritable minnow compared to China, it's got a population, I think, of 23, 24 million compared to China, well in excess of a billion, but it's strategically very significant. So the one China principle considers Taiwan as a breakaway province from the People's Republic of China, and Xi Jinping has made his ambitions for reunification abundantly clear in recent years. In fact, Beijing's rhetoric has been mirrored by equally provocative military exercises close to Taiwan's coastline. So it's unclear really if that's an effort to normalize Chinese military presence around the Taiwan Strait or show physical strength in the face of America's continued military support of Taiwan or both. Who knows. But certainly Taiwan is accelerating its investments into certain technologies like unmanned drones, that's become increasingly central to Taiwan's defense strategy. And they're looking to try and be very much self-sufficient in that. So I read that sort of 70 to 80% now of their drone componentry is entirely domestic focused. The worry, of course, is that rather than enhancing deterrence, which is the initial idea, or certainly my perception of it, the worry is that this technological arms race inadvertently escalates the crisis on both sides. I think that's the danger or the worry that there's a slip up or a mistake in terms of this deterrent and monitoring, which could inadvertently escalate things. So that's something I think we are very mindful of and the companies and indeed, probably the broader markets will continue to track very closely. 

Nick: Yeah, it's certainly something that's top of mind. So on a lighter note, I mean, the weather was pretty bad, but hopefully you managed to get out for dinner. Were there any meal highlights that you had while you were there?

Nick: Yes, we had a funny evening. We're both Scottish, so we decided it's a little bit drizzly. I'm sure it'll be fine. And by the time we arrived at dinner, it was like we'd be standing in a full shower. We certainly ate very well in Taipei, that's for sure. So we tried spicy beef noodles, which were delicious. That's sort of Taiwanese or Taipei speciality. Also found out firsthand about Szechuan peppers, numbing, they're sort of numbing properties in terms of completely numbing your tongue and lips. And then had some very tasty Japanese food. A bit of a legacy, I think, of Taiwan having been under Japanese rule for 50 years or so. We didn't get to try Taiwan's stinky tofu. We could smell it at various night markets, but neither of us braved the stinky tofu. So maybe next time.

Nick: Yeah, well, hopefully that's the next time for stinky tofu. I mean, it does sound like an odd dish. that's probably a good place to draw the podcast to a close on stinky tofu. So thank you, Cat, for joining. It's been a great pleasure to have you on today.

Catriona: Thank you very much.

Nick: And thanks to everyone who took the time today to listen. If you enjoyed today, then please download our other podcasts from the website or wherever you normally get your podcasts. Watch out for the next episode and tunein.