Latest / Business Strategy Talks with Fexingo: Competitive Advantage, Market Positioning, and Growth Planning / How TSMC Built a Moat on Manufacturing Precision
Transcript
- Lucas: When you look at the world's most valuable companies, most of them are asset-light — platforms, software, subscriptions. And then there's TSMC. Luna: The Taiwanese chipmaker that makes the brains inside everything from iPhones to fighter jets. Lucas: Exactly. TSMC is the opposite of asset-light. It spends roughly thirty billion dollars a year on capital expenditures. That's more than most countries' GDP. And that spending is the foundation of what might be the most durable moat in modern manufacturing. Luna: I've heard people call TSMC a monopoly. But it's not like they have a patent on chipmaking. Intel and Samsung both make chips. So what's the actual moat? Lucas: It's a combination of three things. First, sheer scale of investment in leading-edge nodes — the smallest transistor sizes. TSMC's N3 node, the three-nanometer process, took years and billions in R&D. Second, manufacturing discipline — they've perfected yield rates that competitors can't match. And third, trust. Luna: Trust meaning customers know TSMC won't compete with them in design. Lucas: Precisely. TSMC is a pure-play foundry. They don't design chips. Intel and Samsung both design and manufacture — that creates a conflict of interest for companies like Apple, AMD, or NVIDIA. If you're Apple, you don't want your chip supplier also designing competing products. TSMC removes that fear entirely. Luna: That's a real structural advantage. But let's talk about the manufacturing edge. What makes TSMC's factories so hard to replicate? Lucas: It's the yield. Building chips at three nanometers involves hundreds of steps, each with near-zero defect rates. TSMC's yield on a new node is often ninety percent or higher within a couple of years. Competitors — Samsung in particular — have struggled to get above sixty percent on equivalent nodes. If your yield is low, your cost per good chip skyrockets. And customers like Apple need billions of defect-free chips. Luna: So the learning curve is huge. And TSMC has been at this since 1987. Lucas: Right. Founder Morris Chang created the pure-play foundry model. Before TSMC, most chip companies designed and manufactured in-house. Chang saw that design and manufacturing would split, and he bet the company on being the manufacturer for everyone. That bet took decades to pay off, but by the 2010s, it became clear that no one could match TSMC's combination of scale, process technology, and neutrality. Luna: And now they're building factories in Arizona and Japan and Germany. Doesn't that dilute the moat? Spreading the secret sauce around? Lucas: Interesting question. But TSMC isn't transferring its crown jewels. The Arizona fab is behind the leading edge — it's making five-nanometer chips, while the latest N2 node remains in Taiwan. The moat is still anchored in Hsinchu and Tainan, where decades of accumulated process knowledge reside. The overseas fabs are more about geopolitical risk mitigation and customer demand for supply chain resilience. Luna: Let's talk about that risk. TSMC is headquartered in Taiwan, which China claims. Doesn't that create an existential threat to the moat? Lucas: It's the biggest risk. If Taiwan were invaded, global semiconductor supply would collapse. But the moat itself — the institutional knowledge, the culture of precision — can't be seized. You could take the equipment, but you can't replicate thirty years of learning overnight. That said, TSMC has been working to diversify its geographic footprint, exactly because customers like Apple and NVIDIA are nervous. They're building redundancies. Luna: So the moat is partly geographic concentration, which is also the vulnerability. Lucas: Exactly. It's a double-edged sword. But the concentration is also what gives them the edge. Having all your advanced R&D and manufacturing in one place allows for tighter coordination and faster iteration. If you spread too thin, you lose that. Luna: I want to talk about capital intensity. TSMC spends more on capex than the entire GDP of many countries. How does that create a moat beyond just deep pockets? Lucas: It's about the barrier to entry. To compete at the leading edge, you need to invest tens of billions and accept that you might not see a return for years. Intel has tried, but they've stumbled. Samsung has tried, but their yields lag. The sheer sunk cost discourages new entrants. And because TSMC has so many customers, they can spread that cost across a huge volume, making their cost per chip lower than anyone else's. Luna: That's a classic scale moat. Lucas: Yes, but combined with a technology moat and a trust moat. The three together are incredibly rare. Most companies have one, maybe two. TSMC has all three, and each reinforces the others. Luna: Before we move on — a quick honest thing. These deep dives take real work to research and produce, and we keep the show ad-free by design. A small group of listeners already chips in monthly over at buy me a coffee dot com slash fexingo, and that support is what makes this possible. Lucas: Yeah, it's listener-funded, no ads, no sponsors. If these conversations have added something to how you think about business strategy, that's the whole point. No pressure — just putting it out there. Luna: Exactly. Okay, back to TSMC. Lucas, you mentioned the trust moat. Can you give a concrete example of how that plays out? Lucas: Sure. Consider AMD. In the mid-2000s, AMD was struggling with its own fabs. They decided to go fabless — spin off their manufacturing into GlobalFoundries and eventually rely on TSMC. That decision was a bet that TSMC wouldn't prioritize Intel or NVIDIA over them. Today, AMD's most competitive chips, like the EPYC server processors, are made by TSMC. If TSMC played favorites, AMD's entire business could be at risk. But they don't. They treat every customer with the same discipline. Luna: And that neutrality is baked into their corporate structure. Lucas: Right. They don't have a design division. They don't compete with their customers. That's a strategic choice that has become a core part of their moat. It's hard for Intel or Samsung to copy that because they have decades of culture built around designing their own chips. Luna: Let's look ahead. TSMC is already investing in N2, which is expected to start production in 2025 or 2026. Is there any competitor that could catch up in the next five years? Lucas: The consensus among analysts is no. Intel's foundry ambitions are real, but they're years behind. Samsung has the ambition but not the yields. And no new entrant can afford the capex. TSMC's lead is estimated at two to three process nodes — that's a four-to-six-year gap. And that gap is widening, not narrowing, because TSMC's R&D spend grows every year. Luna: So the moat is actually getting deeper. Lucas: It is. But there are two wild cards. One is geopolitics — if conflict in Taiwan disrupts production, even temporarily, customers will be forced to dual-source, which could erode TSMC's volume advantage. The other is technology — if a new paradigm like quantum computing or optical chips replaces silicon, TSMC's accumulated knowledge could become less relevant. Luna: But those shifts are likely a decade or more away. Lucas: Probably. And TSMC is investing in advanced packaging and other technologies to stay relevant even beyond Moore's Law. They're not resting. Luna: So the big takeaway: TSMC's moat is a rare triple threat — scale from capex, technology from decades of process refinement, and trust from being a pure-play foundry. Hard to replicate, and getting harder every year. Lucas: Exactly. And that's why it's one of the most interesting business strategy stories of the last forty years. A company that makes things — atoms, not bits — has built a moat as strong as any software platform. That's not supposed to happen in the age of asset-light business models. Luna: It's a reminder that manufacturing, done right, can be just as defensible as code. Lucas: Absolutely. And with that, thanks for listening. We'll be back next time with another deep dive into how companies build lasting competitive advantage.