Book Review: The Chip War | Chris Miller
We Engineered Efficiency. We Forgot to Build Redundancy.
There’s a machine that costs $350 million, lives in the Netherlands, and is made by exactly one company in the world. Every advanced chip — in your phone, your car, a fighter jet’s targeting system — gets printed using light that only that machine can produce. If something happens to that supply chain, we don’t have a backup plan.
That’s the book. That’s also the world we’re living in right now.
What it’s actually about
Chris Miller jumps between Cold War espionage, the biography of Morris Chang, the physics of lithography, and decades of Pentagon strategy — and somehow it all connects. The argument is simple: chips aren’t a component of modern power. They are modern power.
If you’ve read Yergin’s The Prize on oil geopolitics, or The Code Breaker on the CRISPR race, this sits in the same category. Dense, specific, genuinely illuminating. By page 99 you’re deep into EUV lithography and starting to understand why a $350 million light machine matters more for national security than most weapons programs.
What I kept thinking about after
He opens with a US destroyer. Ninety-six launch cells, looks terrifying. But it’s not the missiles that make it lethal — it’s the Aegis radar system that sees, tracks, and guides them. The whole book is really that one image scaled up. Raw firepower stopped mattering. Computing power is what wins now.
There’s a chapter on the Thanh Hoa Bridge in Vietnam that stuck with me. The US tried to bring it down for years — 800 unguided bombs, couldn’t do it. Then in 1972, 24 smart bombs dropped it in one mission. That’s the moment the US military started betting everything on precision over mass. Everything that followed — the chip race, Taiwan’s importance, the export controls we’re seeing today — traces back to that decision.
The part about why the Soviets failed is worth the price of admission alone. They weren’t short on talent or resources or even stolen designs. What they couldn’t replicate was the accumulated know-how of a factory floor that had been iterating for decades. Chipmaking isn’t a blueprint. It’s institutional knowledge that compounds over time. By the time you copy the current generation, the original is already two generations ahead.
And then TSMC. Morris Chang built what is genuinely the most important factory on earth — roughly 90% of the world’s advanced chips come from one company, on one island, 100 miles from China. Miller doesn’t dramatize this. He just shows you the dependency and lets it sit there.
Where it gets shakier
The China section feels a bit dated. Miller wrote this in 2022, and Chinese progress on lower-end chip manufacturing has moved faster than he anticipated. The core argument still holds, but some of the specific capability gaps he describes have narrowed. Worth keeping in mind as you read.
The policy section at the end also left me wanting more. He covers the CHIPS Act, export controls, allied fab investment — but doesn’t really wrestle with whether spending billions to build fabs in Arizona actually solves the concentration problem, or just relocates it. That question is sitting right there in his thesis. He acknowledges it and moves on.
Should you read it?
If you work anywhere near defense, enterprise AI, supply chain, infrastructure, or policy — yes. It’s one of those books that quietly reframes how you read the news. You finish it and suddenly the chip-related headlines that used to feel technical and distant start making sense.
It’s not a quick read. Miller assumes you’ll stay with him through the technical details. If you’re looking for a high-level take, there are shorter paths. But if you put in the time, the payoff is real.
What made me write this now
I read Chip War a while back and had it sitting on my list to review. Then this week happened.
On February 26, Pakistan declared open war on Taliban-controlled Afghanistan. Airstrikes on Kabul and Kandahar by February 27. Actual state-on-state military conflict, not a proxy situation. And then on February 28, military strikes escalated between Israel and Iran.
What hit me reading both stories through Miller’s lens: neither Pakistan nor Afghanistan manufactures semiconductors. The weapons being used — the guidance systems, the radar, the communications infrastructure — run entirely on chips made somewhere else. That dependency is invisible until a conflict makes it visible.
On the US, Israel, Iran side, there’s a chip plant under construction in Ashkelon, designed for AI and defense applications. Israel’s defense exports hit $14.8 billion in 2024. When conflict reaches regions that are starting to build semiconductor infrastructure, the questions Miller raises aren’t academic anymore.
Underneath all of it, TSMC still makes roughly 90% of the world’s most advanced chips. One disruption to that supply chain — a car misses one $1 chip and a $50,000 vehicle can’t be completed. An AI training run stalls. A weapons system goes offline.
Miller’s point wasn’t that this would definitely happen. It was that we built an incredibly efficient global system with almost no redundancy, and at some point that tradeoff would get tested. This week felt like a test.
Moore’s Law — the idea that chip density doubles roughly every two years — gets talked about like it’s a force of nature.
It isn’t.
It’s the product of continuous investment, compounding expertise, and stable conditions for innovation. Sustained instability doesn’t just affect chip supply today. It erodes the conditions that make future chips possible.
I’m not drawing geopolitical conclusions here.
I don’t have an angle on who’s right or wrong in any of these conflicts. But I do think Miller gave us a useful lens, and right now that lens is clarifying rather than distorting.
The question I keep coming back to
We’ve spent years optimizing for efficiency in our supply chains — chips, components, software dependencies. Miller’s book is essentially a long argument for why that optimization has a hidden cost that we’ve been deferring.
What would it actually look like to build resilience into the systems your organization depends on?
Not as a thought experiment — as a real decision with real tradeoffs?
That’s the question I’m sitting with.
Reviewed for the next reader, not the author.
References: