Latest / International Business with Fexingo: Global Trade, Cross-Border Deals, and Multinational Operations / How Taiwan Semiconductor Manufacturing Outruns Geopolitics
Transcript
- Lucas: So Taiwan Semiconductor, TSMC, is arguably the most geopolitically exposed company on earth that also happens to be utterly indispensable. And this week there's fresh movement on their global expansion that tells you a lot about where cross-border manufacturing is headed. Luna: I saw the EWT, the Taiwan ETF, jumped over five percent in the last five days. That's a big move. Is that tied to TSMC directly? Lucas: Largely, yeah. TSMC is about sixty percent of the EWT weighting. So when the stock moves, the whole country fund moves with it. And the catalyst this week was that their Arizona fab — the first one — is finally expected to start producing chips in early 2027, after multiple delays. Luna: Right, they originally announced Arizona in 2020, then pushed back from 2024 to 2025, then to 2026, and now early 2027. What keeps causing the delays? Lucas: A few things. First, construction labor shortages in the US — especially skilled workers who know how to build out a Class 10 cleanroom. Second, permitting and regulatory complexity that TSMC wasn't used to from Taiwan, where the government basically clears the path. And third, they've been changing the technology node midstream. Originally the Arizona fab was going to do 5-nanometer. Now it's being upgraded to 4-nanometer and eventually 3-nanometer, because customer demand shifted. Luna: So the delays are partly a sign of success — customers want the latest technology. But the cost overruns must be massive. Lucas: Absolutely. TSMC initially budgeted twelve billion dollars for the Arizona complex. That's now ballooned to forty billion for the two fabs they're building. And that's before any operational costs. Compare that to their fab in Kumamoto, Japan, which came online last year almost on schedule and under budget. The Japanese government covered about forty percent of the cost, the local workforce was easier to train, and the entire process was smoother. Luna: Japan seems like a much better fit culturally and operationally for TSMC. The Kumamoto fab is already profitable, right? Lucas: They said they expect it to be profitable this fiscal year. That's fast. The Arizona fab, even once it's running, will take years to reach profitability because of the higher cost base. And then there's the Dresden, Germany fab — a joint venture with Bosch, Infineon, and NXP — that broke ground in 2024 and is targeting production by 2029. That one has its own challenges: energy costs, labor regulations, and the European Union's slower subsidy approval process. Luna: So TSMC is essentially running three parallel international expansions, each with very different economics. What does that mean for their margins? Lucas: CFO Wendell Huang said last quarter that they expect overseas fabs to drag on gross margins by about two to three percentage points over the next few years. But here's the thing — TSMC's gross margins are still around fifty-three percent. So even with a drag, they're wildly profitable. And they're pricing those chips to reflect the higher cost. Customers like Apple, NVIDIA, AMD — they have no alternative. No one else can make 3-nanometer chips at scale. Luna: Right, Intel's foundry business is still struggling, Samsung has yield issues. So TSMC has pricing power. But politically, they can't just stay in Taiwan. The US and Japan and Germany all want domestic supply. Lucas: Exactly. And TSMC CEO C.C. Wei said last month that the company's strategy is to 'expand globally while keeping the core R&D and most advanced manufacturing in Taiwan.' So they're building these fabs as a kind of geopolitical insurance. If things get worse in the Taiwan Strait, they have some chip capacity outside. But the cutting edge — the 2-nanometer and below — stays in Hsinchu. Luna: That makes sense. But the US CHIPS Act — the $52 billion in subsidies — was supposed to cover some of this cost. How much has TSMC actually received? Lucas: As of this month, TSMC has been awarded about $6.6 billion in direct grants and up to $5 billion in loans from the CHIPS Act. But that's far below the initial $12 billion they reportedly hoped for. And the grants are disbursed based on milestones, so they get the money as they hit construction targets. The first milestone payment came through when they completed the roof structure on Fab 1. Luna: So the subsidies help, but they don't cover the full cost premium. And TSMC shareholders are essentially absorbing the difference. But the stock has been doing well — up over 20 percent this year. Lucas: Because the revenue story is even bigger. TSMC's revenue grew about 30 percent year-over-year in the first quarter of 2026, driven by AI chip demand. Their advanced packaging — CoWoS, which is the technology that stacks chips together for AI accelerators — is completely sold out. They're building a dedicated packaging facility in Taiwan just for that. And they're also expanding CoWoS capacity in Japan. Luna: So the AI boom is pulling TSMC's core business forward faster than the geopolitical costs can weigh it down. That's a pretty strong position. Lucas: It is. But there's a longer-term risk. If export controls on advanced chip equipment to China get tighter — and they've been tightening every year — TSMC could lose access to some Chinese customers. Currently about 12 percent of TSMC's revenue comes from China-based chip designers. That's not huge, but it's a growth market. Luna: And NVIDIA said this week that their forecast for a $200 billion CPU market includes China. So the AI chip demand in China is real. But if TSMC can't serve those customers due to US rules, that business goes to Chinese foundries like SMIC, even if their technology is a generation behind. Lucas: Right. SMIC can make 7-nanometer chips now, using older equipment. They're not competitive on the cutting edge, but for many AI inference chips, 7-nanometer is good enough. So there's a bifurcation happening: TSMC gets the premium, leading-edge business; Chinese foundries get the mainstream stuff. And the US export controls are basically accelerating that split. Luna: It's fascinating. TSMC is simultaneously the most important company in global tech and the most geopolitically entangled. And their response is to build factories everywhere while keeping the crown jewels in Taiwan. Lucas: And if today's conversation gave you a clearer picture of how that works, that's exactly what we're here for. These episodes stay ad-free because of listener support — if you found it useful, you can help keep it going at buy me a coffee dot com slash fexingo. No pressure, just a way to say this kind of deep dive matters to you. Luna: Yeah, and we really appreciate it. It lets us spend the time on stories like this without chasing sponsors. Lucas: So back to the bigger picture — when you look at TSMC's global expansion, it's a case study in how a company can turn geopolitical risk into a moat. Competitors can't replicate the scale, the technology, or the customer relationships. And by building fabs in key markets, TSMC makes itself more valuable to those governments, which gives it political leverage too. Luna: It's almost like they're using the factories as diplomatic assets. If you're the US government, you're less likely to impose crippling sanctions on a company that's building a $40 billion complex in Arizona and employing thousands of Americans. Lucas: Exactly. And that's the sophisticated play. TSMC isn't just a chipmaker — they're a geopolitical actor. And so far, they're playing the game very well.