Latest / Business Strategy Talks with Fexingo: Competitive Advantage, Market Positioning, and Growth Planning / How Taiwan Semiconductor Built a Moat in Advanced Chip Manufacturing
Transcript
- Lucas: There is a company you have almost certainly never bought a product from, yet it makes the chips inside your phone, your laptop, your car, and probably the data center that powers whatever AI tool you used this morning. Luna: You're talking about TSMC — Taiwan Semiconductor Manufacturing Company. Lucas: Exactly. And the reason I want to talk about TSMC today is that its competitive moat may be the widest and deepest of any company on earth right now. We're talking about a firm that controls roughly 90 percent of the market for the most advanced chips in existence. Luna: That's an astonishing number. And before we go deep, I'll just say — if this kind of analytical deep dive is useful to you, if it's saved you time or helped you think more clearly about business strategy, that's exactly why we keep the show ad-free. The only support we ask is at buy me a coffee dot com slash fexingo. Five bucks, one coffee, whatever feels right. And now back to the moat. Lucas: Appreciate that. So TSMC was founded in 1987 by Morris Chang, who was then in his mid-fifties and had spent years at Texas Instruments. He saw that the semiconductor industry was moving toward a model where you could separate chip design from chip manufacturing. That was a radical idea at the time. Luna: Because back then, every major chip company — Intel, Motorola, TI — they all designed and manufactured their own chips. It was an integrated model. Lucas: Right. Chang's insight was that if you could manufacture chips for hundreds of different designers, you could spread the enormous fixed cost of a fab across many customers. And you could get better at manufacturing faster than any single company could, because you were learning from a much wider set of designs. Luna: So the moat starts with scale. But it's not just scale, right? Because Samsung also has massive scale, yet they're not winning the advanced foundry business. Lucas: That's the key question. Let me give you a sense of the numbers. In 2025, TSMC spent roughly thirty-six billion dollars on capital expenditure. That's more than Intel and Samsung combined in that year. And they've been outspending everyone for at least a decade. Luna: Thirty-six billion in a single year. That's more than the GDP of some small countries. Lucas: It is. And that spending isn't just about building more fabs. It's about process technology. TSMC currently has a two-year lead over Intel and a three-year lead over Samsung in the most advanced node, which they call N2, set to enter volume production in late 2025. By the time competitors get there, TSMC will be shipping N1.5 or even N1. Luna: So it's a moving target. And the moat is that you can't catch up by just spending money, because the learning curve is cumulative. TSMC has been refining these processes for nearly four decades. Lucas: Exactly. There's a concept called 'process knowledge' that's incredibly hard to transfer. You can buy the same lithography machines from ASML that TSMC uses. But the recipes, the tweaks, the yield optimization — that's proprietary. TSMC's defect density per square millimeter is significantly lower than any competitor's. That translates directly into higher yields and lower costs. Luna: And that's why Apple, NVIDIA, AMD, Qualcomm — they all design their most advanced chips with TSMC. They can't afford to risk their flagship products on a node that might have lower yields. Lucas: Which brings us to customer trust as a moat. These companies share their future product roadmaps with TSMC years in advance. They co-optimize designs with TSMC's process engineers. That creates a switching cost that's almost impossible to overcome. If Apple wanted to move its A-series chips to Samsung, they'd have to redesign the entire chip for a different process, and they'd lose two years of performance leadership. Luna: So the moat has multiple layers: capital intensity, process technology lead, customer lock-in, and the sheer accumulated learning. But there's also a geographic concentration risk that people talk about a lot. Lucas: Absolutely. Over 90 percent of the world's most advanced chips are made in Taiwan. That's a massive geopolitical concentration. TSMC has started building fabs in Arizona, Japan, and Germany, but those are for less advanced nodes. The crown jewels remain in Hsinchu, Tainan, and Taichung. If anything disrupted production there, the global economy would seize up. Luna: That's the flip side of the moat. The very thing that makes TSMC so dominant — being the only place where you can reliably make these chips — also makes the supply chain fragile. Lucas: Right. And that's why Intel and Samsung have tried so hard to catch up, and why governments are pouring subsidies into domestic chip manufacturing. The CHIPS Act in the US, similar programs in Europe and Japan — they're all trying to create alternatives. But building a foundry ecosystem takes more than money. It takes talent, culture, and time. Luna: Let's talk about culture for a second. Morris Chang instilled a very specific culture at TSMC — he called it 'manufacturing excellence'. It's almost military in its discipline. Engineers work long hours, there's intense focus on process control, and the company has a policy of not competing with its customers. They don't design chips, so they're never a threat. Lucas: That last point is crucial. Intel tried to enter the foundry business, but many potential customers didn't trust them because Intel also designs chips. They worried Intel would prioritize its own products or steal their designs. TSMC never has that conflict. They are a pure-play foundry, period. Luna: And that trust took years to build. But once it was built, it became another barrier to entry. You can't replicate a forty-year track record of not competing with your clients. Lucas: Let me give you a concrete example of how deep the moat is. In 2020, when the pandemic hit, demand for chips surged. Every foundry was strained. But TSMC managed to keep its yields high and its delivery times relatively stable. Meanwhile, Samsung's foundry had major yield issues on its 5-nanometer node, losing Qualcomm as a customer for some products. That kind of reliability gap is a moat all by itself. Luna: So the moat is self-reinforcing. Better yields attract more customers, which generates more revenue, which funds more R&D, which leads to better process technology, which attracts more customers. It's a virtuous cycle. Lucas: Exactly. And the cycle has been running for decades. In 2025, TSMC's revenue was over eighty-seven billion dollars, with net margins above 35 percent. That kind of profitability funds the capex we talked about, and also allows them to pay employees well, which attracts the best talent. Luna: But what about the future? Is there any scenario where the moat erodes? We've seen other seemingly unassailable moats get disrupted — Kodak, Nokia, even Intel itself. Lucas: There are a few risks. One is geopolitical: a Chinese invasion of Taiwan, or a major blockade, would be catastrophic. But that's an exogenous shock, not a competitive erosion. Another risk is technological discontinuity. If someone invents a completely new way to make chips — say, using carbon nanotubes or quantum effects — TSMC's expertise in silicon lithography could become less relevant. Luna: But that's a long shot, and TSMC is investing heavily in research on those alternatives anyway. Lucas: They are. They have one of the largest central R&D organizations in the world, with thousands of PhDs working on next-generation technologies. So they're not resting on their laurels. The more immediate risk is that the cost of building fabs becomes so high that even TSMC struggles to finance it. A single leading-edge fab now costs over twenty billion dollars. There may come a point where the return on that investment starts to decline. Luna: But for now, the moat is as strong as ever. And it's a fascinating case study because the moat isn't based on a brand or a patent or a network effect — it's based on manufacturing excellence and a virtuous cycle of investment and learning. Lucas: It's a reminder that competitive advantage can come from places that aren't glamorous. TSMC doesn't have a flashy product. But they have something that is arguably more valuable: the ability to make the most complex physical objects on earth at a scale and reliability that no one else can match. Luna: And that ability has made them indispensable to the entire technology industry. I think that's the definition of a moat. Lucas: It is. And the next time you pick up your phone, just remember — that chip inside it was probably made by a company you never see, but without which the modern world would look very different.