Deep Dive
The Silicon Valley-Taiwan Partnership That Powers AI
Jon Stewart opens with a pointed question: Silicon Valley designs the world's most advanced chips but has outsourced all manufacturing to Taiwan's TSMC. Why didn't anyone plan for the rare earth materials needed but not mined domestically? Chris Miller, historian and author of Chip War, explains the architecture: Nvidia, AMD, and other designers email their chip designs to Taiwan, where TSMC manufactures them with near-atomic precision. The Japanese supply ultra-pure silicon wafers, and the Dutch provide the only advanced photolithography tools that can etch transistors at nanometer scale. This global supply chain, fragmented across three continents and three countries, works brilliantly for innovation but creates singular points of failure. The entire AI industry depends on TSMC's continued operation in Taiwan, yet China has military designs on the island and has grown capable enough in recent years to potentially succeed in an invasion. Stewart and Miller establish the core paradox: the US has the most valuable chip design firms in the world but owns almost none of the critical manufacturing and material infrastructure.
How Taiwan Became the Chip Manufacturing Powerhouse
Miller recounts the 1987 origin story: Morris Chang, passed over for CEO at Texas Instruments, pitched the Taiwanese government on a radical idea — a company that only manufactures chips, never designs them. He modeled it on Gutenberg's printing press, which reproduced books but didn't write them. Texas Instruments and Intel rejected this model, believing their in-house design-and-manufacturing approach made sense. They were wrong. Chang's TSMC bet paid off because it unlocked economies of scale unavailable to single-company operations. Once Nvidia and AMD emerged as pure-design firms, they could all manufacture at TSMC, and TSMC could spread fixed costs across an entire industry. Stewart pushes back on why US companies missed this: it's obvious in hindsight, but at the time there were no pure-design customers because the model didn't exist yet. Chang had to say if I build it, they will come — a bet that could have failed. The financial incentive also cut against the model; designers capture more margin than foundries do. Nvidia's now a $5 trillion company while TSMC is $2 trillion, so financially Silicon Valley made the right choice even if it created strategic vulnerability.
China's Rare Earth Stranglehold and US Blindness
Miller pivots to materials: China produces virtually all refined rare earth elements — yttrium, dysprosium, and others with names that sound made up but are critical to electronics. The US mines almost none and refines even less. When Stewart asks if these materials are actually rare, Miller clarifies they're not rare geologically but rare in economically viable concentrations — found everywhere but expensive to extract. China doesn't make money on rare earths; it treats them as a loss leader, producing at razor-thin margins to lock in global dependence. In 2010 and 2011, China cut off rare earth exports to Japan over a territorial dispute, disrupting the entire tech supply chain. Japan responded immediately by investing in mines and processing facilities in Australia and Malaysia. The US response was nothing. Sixteen years later, with a trade war now active and China retaliating with rare earth restrictions, the US still has no domestic production. Miller notes the US literally invented many rare earth processing techniques in military labs but abandoned the industry. Stewart, incredulous, points out that if you designed cars without securing rubber, you'd be called insane — yet the US bet its entire semiconductor and AI future on materials it doesn't control and didn't bother to secure.
The Trade War's Backfire and Allied Fracture
Stewart raises the geopolitical mess: the president just returned from China empty-handed and triggered a trade war. Miller explains China understood immediately what tool to use — rare earths. The administration apparently didn't expect retaliation along the supply chain's weakest link. What makes this worse is the trade war's second-order effects: Chinese exports didn't collapse to zero, they just diversified to Europe, India, and other markets. BYD, China's EV giant, is now entering Canada. Companies that were hedging toward US partnerships are now hedging away. Miller is direct: the US needed to strengthen alliances with Japan, Europe, Australia, and India — countries equally vulnerable to Chinese supply leverage. Instead, the trade war alienated them. Japan and the US together could source rare earths from multiple friendly countries and lock China out of a massive combined market. But tariffs on Canadian steel and other blunt tools pushed allies toward accommodation with China instead of coordination against it. Stewart doesn't pull punches: if the stakes are so high that we can't regulate AI or we'll lose to China, how is the government's failure to plan this supply chain anything but unforgivable? Miller agrees the situation is dire but argues the problem is solvable because US strength lies in R&D and innovation, not in mining. China's advantage is in basic material processing, which improves slowly. The US advantage — chip design and advanced tooling — moves at light speed.
Why the US Problem Is Easier to Fix Than China's
Miller lays out the asymmetry: the US needs to rebuild rare earth mining and processing domestically or with allies. This takes 20 years and money but the raw materials exist in the US, Australia, and Africa. Processing chemistries were invented in American labs, so we can relearn them. The moving target works in America's favor — Taiwan and the US improve chip manufacturing every year. If China catches up to 2026-level technology but it's now 2030, they're light years behind again. Competition in chips moves at nanometer scales and happens in nanoseconds; competition in materials happens in chemistry labs at human timescales. Miller would rather have America's problem. Stewart pushes further: how would he design a strategy? Miller's answer is deceptively simple: leverage allied market size. The combined US, European, Japanese, and Indian economies dwarf China's. If these countries coordinated to source rare earths only from friendly nations and locked China out, Beijing would have to negotiate access to a market larger than its own. The Chips Act was a step toward Taiwan independence but didn't address material resilience. A real strategy would coordinate with allies on mining, processing, and manufacturing — not tariffs on friends but collective defense against dependence.
The Broader Geopolitical Reckoning
Stewart steers toward the strategic picture: If we're fracturing with allies while a trade war backfires, is the dollar — our last superpower — also weakening? Miller says not yet, but watches it carefully. The dollar's value hasn't collapsed and alternatives are limited. China can't easily internationalize the yuan because doing so would expose its currency manipulation and wreck its competitiveness. But the real risk isn't the dollar collapsing; it's allies deciding that working with China is safer than working with an unpredictable US. Miller then broadens to China's own fragility: Xi Jinping faces slowing growth, rising youth unemployment, energy shocks from the Middle East, and paranoia about his military — he just purged top generals. China looks strong on rare earths but weak on many other metrics. Stewart presses on whether we can split Russia and China apart, as some in the administration hope. Miller: the logic is sound in theory but implausible in practice. Russia has few reliable partners besides China; moving closer to Russia doesn't break the Russia-China alliance. What Stewart really wants to know is whether we're abandoning the post-WWII order the US built — an order where the US Navy patrolled seas, dollars funded trade, and rules constrained great powers. Are we becoming like Russia and China, seizing what we can because we're powerful? Miller notes the irony: the US built the most valuable ecosystem in the world precisely because it outsourced manufacturing and relied on Taiwan and allies. Silicon Valley's $5 trillion value exists because of that system. Dismantling it won't make us richer, it'll fragment us further.