Latest / International Business with Fexingo: Global Trade, Cross-Border Deals, and Multinational Operations / How Taiwan Semiconductor Is Reshaping Global Supply Chains
Transcript
- Lucas: So Taiwan Semiconductor — TSMC — just shipped its first commercial 4-nanometer chips from the new Arizona fab. Apple and NVIDIA are the first customers. It's a genuinely big deal, and not just for the companies involved. Luna: Right, I saw that headline. The fab's been under construction for years, but this is the first time advanced chips are actually being made on US soil at scale. Lucas: Exactly. And 'at scale' is the key phrase. TSMC's Arizona facility — Phase 1 alone cost $40 billion. It's one of the largest foreign direct investments in US history. But the question is whether this marks a genuine rebalancing of the global chip supply chain or just a symbolic hedge. Luna: Well, the numbers suggest it's more than symbolic. The Taiwan Weighted Index is up 14 percent in the last week alone, partly on this news. But if you look at TSMC's overall production, Arizona is still a drop in the bucket compared to their fabs in Taiwan. Lucas: Right. TSMC produces something like 90 percent of the world's most advanced chips — 5 nanometer and below — all from Taiwan. So Arizona adds capacity, but it doesn't eliminate the concentration risk. The real strategic shift is that customers like Apple and NVIDIA now have a backup source outside the Taiwan Strait. Luna: And that's the geopolitical angle everyone talks about. But I'm more interested in the economic ripple effects. What does a $40 billion fab in the Arizona desert do to the local economy? Lucas: It's transformative, but not in a simple way. Phoenix is now competing with Austin and Columbus for semiconductor talent. TSMC is importing hundreds of engineers from Taiwan — which created some local friction. But over time, it's building a skilled workforce. Community colleges in Arizona have tripled their semiconductor technician programs. Luna: That's the kind of detail that gets overlooked. It's not just about the factory itself, it's about the ecosystem. And this doesn't happen in a vacuum — other countries are watching closely. Lucas: Absolutely. South Korea's Samsung is expanding its Taylor, Texas fab — also competing for the same talent pool. And Japan's Rapidus, a government-backed consortium, is trying to build a 2-nanometer fab in Hokkaido. The global race for chip sovereignty is on, and it's incredibly capital-intensive. Luna: Speaking of capital — the US CHIPS Act allocated $52 billion in subsidies, but that's already being stretched thin. TSMC alone got $6.6 billion in grants and $5 billion in loans. At this rate, a single advanced fab costs more than the entire annual budget of some small countries. Lucas: Yeah, and that raises a question: is this the most efficient use of resources? From a pure free-trade perspective, it would be cheaper to keep making chips in Taiwan and ship them. But the security premium is now baked into the calculus. Companies are willing to pay more for supply chain resilience. Luna: And consumers feel that, eventually. If chips cost more to produce in Arizona, the price gets passed down — to phones, cars, data centers. But the trade-off is avoiding a catastrophic disruption if something happens in the Taiwan Strait. Lucas: Look, this is the kind of conversation that makes our show possible — we get to explore these trade-offs without a sponsor telling us to keep it upbeat. And that's actually something we're intentional about. Luna: Yeah. We don't run ads because we think the topic deserves a clear, uninterrupted look. If you agree and want to support that, there's a link — buy me a coffee dot com slash fexingo. Lucas: It's a small way to keep this independent. And if it's moved your thinking forward, that's the goal. Now, back to TSMC — the question about long-term efficiency versus security is one thing, but there's also the technology roadmap. Luna: Right. TSMC's next big leap is 2-nanometer, expected by 2028. Where will that be manufactured? Lucas: Both Taiwan and Arizona. TSMC has already announced a second phase in Arizona for 3-nanometer and eventually 2-nanometer. But the most advanced R&D stays in Taiwan. So we're seeing a split: high-volume manufacturing gets distributed, but the bleeding edge stays home. Luna: That sounds like a smart hedge for TSMC itself. They keep their core IP in Taiwan while serving key customers in the US. But it also means the US will always be a generation or two behind, unless something changes. Lucas: Exactly. And that's the tension. The US government wants leading-edge production onshore, but the economics and the expertise are still deeply concentrated in East Asia. The new export data we saw this morning — US imports of semiconductors and related equipment are still rising, not falling. Luna: Right, the import numbers for goods and services hit $4.4 trillion annualized in the first quarter, up from $4.1 trillion. A big chunk of that is electronics. So the reshoring narrative hasn't yet shown up in the trade balance. Lucas: It takes years. The Arizona fab broke ground in 2021 and is only now shipping in volume. The second phase won't be online until 2028. So the trade data will lag by a decade. But the direction is clear: more fabs in more places. Luna: And that's not just a US story. Look at what's happening in Europe. Intel is building a massive fab complex in Magdeburg, Germany. TSMC is building a specialized fab in Dresden with Bosch and Infineon. The world is fragmenting into regional chip hubs. Lucas: Right. And each hub comes with its own subsidies, its own talent challenges, its own regulatory environment. For a multinational company like Apple, that means managing multiple supply chains instead of one. It's a lot more complex, but also more resilient. Luna: I wonder how smaller companies are coping. If you're a startup designing a chip for a medical device, you used to be able to work with TSMC in Taiwan and get prototypes in weeks. Now, you might need to qualify multiple fabs, which is expensive and slow. Lucas: That's a real pain point. The semiconductor industry is consolidating at the top, but the barriers for new entrants are rising. The cost of a mask set for 4-nanometer is something like $5 million. So the reshoring trend actually favors the largest players. Luna: And that's a concern — because innovation often comes from startups. If the cost of prototyping goes up, we might see fewer new chip companies. That could stifle the next wave of AI hardware or specialized sensors. Lucas: It's a legitimate worry. But on the flip side, government programs like the CHIPS Act include funding for prototyping and R&D. The National Semiconductor Technology Center is supposed to help. But it's early days. Luna: Let's talk about the AI angle specifically. NVIDIA is one of TSMC's biggest customers for AI chips. With the Arizona fab now operational, does that change anything for AI supply? Lucas: It does in terms of geographic risk. If a crisis hits Taiwan, NVIDIA's AI chips can still be made in Arizona — at least partially. But the highest-performance chips, like the ones for training large models, still come from Taiwan. So the AI boom is still heavily dependent on a single island. Luna: And that's why the market reaction was so positive. The iShares MSCI Taiwan ETF — ticker EWT — was up 5.6 percent in the last five days. Investors are pricing in that TSMC's global expansion reduces systemic risk, even if it costs more. Lucas: Yeah. EWT is at $106.47 now, which is near its 52-week high. The market is essentially saying: we trust TSMC to execute this global strategy. And so far, they have. Arizona is on schedule and on budget, which in today's construction environment is almost miraculous. Luna: It really is. Look at the delays in other large infrastructure projects — high-speed rail, offshore wind. TSMC managed to build a state of the art fab in the desert, with a new workforce, in four years. That's impressive execution. Lucas: Part of that is the company's culture. TSMC is famously disciplined. They've been doing this for decades. But they also had strong government support — streamlined permits, infrastructure investment. It's a model of public-private partnership. Luna: So looking ahead, what's the biggest risk to this reshoring story? Lucas: I'd say talent. There simply aren't enough semiconductor engineers in the US. TSMC had to bring in workers from Taiwan, which created some political backlash. Over time, the talent pipeline will grow, but it's a bottleneck. The second risk is cost escalation. If building fabs keeps getting more expensive, the economics might not work without permanent subsidies. Luna: And that becomes a political question. Will the next administration continue to fund the CHIPS Act? The subsidies are not guaranteed forever. If they lapse, some of these projects could stall. Lucas: Exactly. So the reshoring of semiconductor manufacturing is not a done deal. It's a long-term bet that requires sustained commitment. But right now, with the Arizona fab shipping, it's the most tangible sign yet that the bet is paying off. Luna: It's a fascinating moment. For decades, the semiconductor supply chain was optimized for cost. Now it's being reoptimized for resilience. And TSMC is at the center of that transformation. Lucas: And we'll be watching closely how it evolves — especially as the 2-nanometer race heats up and other players like Samsung and Intel try to catch up. For now, though, the Arizona fab is a landmark. It changes the conversation from 'should we reshore' to 'how do we reshore better.'