Latest / The Productivity Podcast with Fexingo: Output, Efficiency, and Long-Term Economic Growth / How Japan Resurrected Its Chip Industry Productivity
Transcript
- Lucas: In 1988, Japan produced more than half of the world's semiconductors. TSMC didn't exist. Samsung was still a decade away from becoming a serious player. Luna: And today Japan's share is below 10 percent. That's a staggering fall. Lucas: It is. And it's the backdrop for what might be the most ambitious productivity revival attempt in modern industrial history. The Japanese government and a consortium of its biggest companies—Toyota, Sony, SoftBank, NTT, and others—have formed a new entity called Rapidus. Luna: Rapidus. I've seen the name. The goal is to produce two-nanometer chips by 2027, right? Lucas: Exactly. Two nanometers. That's the frontier that currently only TSMC and Samsung have reached, and they're both struggling with yields. Rapidus is essentially saying: we're going to leapfrog a generation of process technology and build a megafab in Hokkaido—of all places—to prove Japan can compete again. Luna: Why Hokkaido? That's not exactly a semiconductor hub. Lucas: The logic is actually pretty interesting. Hokkaido has abundant renewable energy—wind and hydro—which is increasingly critical for chip fabs that consume enormous amounts of electricity. Plus, the local government offered significant tax incentives and land. And it's far from potential geopolitical flashpoints like Taiwan. Luna: So the location itself is part of the productivity strategy. But how does a consortium like this avoid the classic pitfalls of government-led industrial policy? I'm thinking of things like Japan's own experience in the 1970s and 80s where they had successful state-backed projects, but also some spectacular failures. Lucas: That's the key question. The Rapidus model is different from the old miti style vertical integration. It's structured as a private company with massive public funding—about $5.4 billion committed so far, with more expected. But the leadership includes veterans from TSMC and IBM, and they're licensing technology from IBM's two-nanometer research. Luna: Licensing from IBM. That's a shortcut, but is it enough? IBM hasn't actually mass-produced chips in years. Lucas: Exactly right. IBM's process design is theoretical in the sense that it's been demonstrated in a lab, not in a high-volume manufacturing environment. The real productivity challenge is transferring that lab recipe to a commercial fab and achieving acceptable yields—north of 80 percent. That's where TSMC has a decade of iterative learning. Luna: So you're saying Japan is trying to compress ten years of learning into about three or four years. That sounds incredibly risky. Lucas: It is. And it's not just technical. There's a cultural dimension. Japanese corporate culture is famously risk-averse. Semiconductor manufacturing requires rapid experimentation and acceptance of failure—you run lots of test wafers, you tweak parameters, you fail fast and learn. That's not the Japanese industrial tradition, which prizes precision and doing it right the first time. Luna: But Japan did dominate chip manufacturing once. How did they lose that edge? Lucas: There are several reasons, but the big one is the 1986 us japan Semiconductor Agreement. The US accused Japan of dumping chips and forced them to open their market and limit exports. That crippled Japan's DRAM industry, which was the volume engine that drove learning and cost reductions. Korean companies like Samsung stepped into the gap. Luna: So essentially the US government inadvertently created Samsung's opportunity. Lucas: That's a fair reading. And then Japan's keiretsu structure made it hard for them to pivot to the foundry model that TSMC pioneered. Japanese chipmakers were mostly integrated device manufacturers—they designed and made chips for their own products. They never built the culture of serving external customers with demanding specs. Luna: Rapidus is explicitly a foundry, though. So they're trying to adopt the TSMC model. But TSMC's productivity comes not just from technology but from a relentless culture of operational efficiency. Can Japan replicate that? Lucas: That's the bet. And there are reasons for optimism. Japan still has an incredibly skilled workforce in materials science and precision engineering. Companies like Tokyo Electron and Shin-Etsu are world leaders in semiconductor equipment and materials. So the ecosystem is there. What's missing is the integration and the manufacturing discipline specific to advanced logic chips. Luna: Speaking of materials, I read that Japan actually supplies about half the global market for the high-purity chemicals needed for chip manufacturing. So the raw inputs are there. Lucas: Absolutely. And that's a competitive advantage that China, for example, lacks. Japan has the supply chain for the building blocks. The question is whether they can assemble them into a world-class fab operation. And there's another dimension: geopolitical urgency. The US and Europe are desperate to reduce dependence on Taiwan for advanced chips. Japan is seen as a reliable ally. Luna: So there's customer demand waiting. TSMC is already building a fab in Japan, actually—in Kumamoto. That's a lower-tech facility, though. Lucas: Right. TSMC's Kumamoto plant is for 12 to 28 nanometer chips, not leading edge. But the fact that TSMC chose Japan says something about the talent and infrastructure. Rapidus is going for the frontier, and they've secured a commitment from Toyota and others to be anchor customers. If they can deliver, they'll have guaranteed demand. Luna: Let me push back a bit. The global chip market is incredibly cyclical. We're coming off a severe downturn in 2023 and 2024. Building a new fab for cutting-edge chips is a multi-billion dollar bet that requires sustained investment for years before seeing returns. If the market turns again, or if AI chip demand shifts, Rapidus could be left with a very expensive white elephant. Lucas: That's the risk. But the Japanese government is essentially treating this as a national security imperative. They've allocated nearly $6 billion so far, and they're prepared to put in more. The calculus is that the cost of not having domestic advanced chip production is higher than the cost of a failed project. Luna: It's a classic productivity paradox: you invest heavily today for uncertain future gains. But if you don't invest, you're locked out forever. Lucas: Exactly. And there's a precedent that gives some hope. In the 1970s, Japan's VLSI consortium—a government-backed research project—helped them leapfrog the US in memory chips. That consortium brought together five competing companies to share research costs. Rapidus is a similar model, but this time the goal is a commercial foundry, not just R&D. Luna: The VLSI consortium worked because it focused on pre-competitive research. Rapidus is asking competitors to collaborate on production, which is much harder. Toyota and Sony are customers, but they'll also be competing for allocation of capacity. Lucas: That tension is real. But the consortium structure includes a commitment from all partners to use Rapidus as their primary foundry for advanced chips. That alignment of incentives is unusual. And the Japanese government has made it clear that Rapidus will be treated as a national champion—defense contracts, infrastructure support, and immigration policies to attract foreign engineers. Luna: You mentioned immigration. Japan has historically been very insular. Are they really willing to open the doors to foreign chip talent? Lucas: They've already started. In 2024, Japan created a special visa for semiconductor engineers and offered fast-track residency. The Rapidus fab in Hokkaido will reportedly have a large contingent of foreign engineers—some from IBM, some from TSMC, some from European equipment makers. That's a significant cultural shift for Japan. Luna: It really is. So the productivity challenge here isn't just about machines and processes. It's about organizational culture, talent flows, and the willingness to learn from outsiders. That's a much harder problem. Lucas: And that's what makes this episode so interesting. The standard productivity narrative is about technology adoption—better tools, better processes. But Japan's chip revival is fundamentally about institutional change. Can a society that values stability and consensus embrace the fast-failure, iterative culture required for leading-edge semiconductor manufacturing? Luna: If they succeed, it would be a huge signal that old industrial powers can reinvent themselves. If they fail, it's a cautionary tale about the limits of top-down productivity pushes. Lucas: Honestly, if today's episode helped you think about productivity in a new way, that's the kind of value we try to deliver every day. And if it was worth a coffee to you, you can support the show at buy me a coffee dot com slash fexingo. We keep the podcast ad-free, and listener support is what makes that possible. Luna: Yeah, it's a small gesture that goes a long way. We really appreciate it. Lucas: So back to Rapidus. One factor that often gets overlooked is the role of extreme ultraviolet lithography, or EUV. TSMC and Samsung use EUV machines from the Dutch company ASML to print the tiny features on chips. Each machine costs about $400 million and requires a cleanroom the size of a football field. Luna: And Japan's Nikon and Canon used to make lithography equipment, but they fell behind ASML on EUV. So is Rapidus dependent on ASML? Lucas: Completely. They've already ordered EUV machines from ASML, and those are scheduled for delivery starting in 2026. But here's the catch: ASML has limited production capacity, and it allocates machines years in advance. TSMC and Samsung have priority. Rapidus may not get enough machines to achieve the volume they need to be cost-competitive. Luna: So even if they solve the technical and cultural challenges, there's a supply constraint beyond their control. That's a real vulnerability. Lucas: It is. And it's why some analysts think Rapidus's real contribution may be in niche applications—custom chips for automotive, IoT, and AI inference—rather than competing head-on with TSMC in high-volume logic. Toyota, for example, needs advanced chips for autonomous driving, but they don't need TSMC's scale. Luna: So maybe the success metric isn't market share. Maybe it's building a self-sufficient capability for Japan's key industries. That's a lower bar, but still significant. Lucas: Exactly. And if Rapidus can achieve that, it would be a meaningful productivity boost for the Japanese economy. The auto industry alone accounts for about 20 percent of Japan's exports, and they're increasingly dependent on chips. Having a domestic source of advanced chips reduces supply chain risk. Luna: It's a hedge, not a moonshot. But the rhetoric from the Japanese government has been very ambitious. Prime Minister Kishida called it a 'national project' to revive the semiconductor industry. That sets expectations high. Lucas: And that's the final piece of the puzzle: political pressure. Public projects often face pressure to show results quickly, which can distort decision-making. If yields are low in 2027, will the government double down or pull the plug? Luna: What's your bet? Lucas: I think they'll double down. The geopolitical stakes are too high. The US and Europe are both investing heavily in domestic chip capacity—the CHIPS Act in the US, the European Chips Act. Japan can't afford to be left out. So I expect Rapidus to survive even if it's not profitable for a decade. The question is whether it can ever become truly productive. Luna: And that's the ultimate productivity paradox: a project that is economically inefficient but necessary for national resilience. Lucas: Right. And that's where we'll leave it. Japan's chip revival is a fascinating case study in how productivity isn't just a technical metric—it's shaped by history, culture, and geopolitics. Thanks for listening.