Almost everyone working in this field operates at a level of abstraction where compute is a line item. You request a GPU, it appears, you are billed. If it is expensive you complain about the price. If it is unavailable you wait.
The physical reality underneath, that these devices are manufactured by a handful of firms using machines built by essentially one company in the Netherlands, on an island whose political status is contested, is a set of facts that occasionally surfaces in the news and rarely gets integrated into how people think. This book is the correction, and it is the best one available. The central achievement is making the concentration comprehensible rather than just asserting it. Anyone can tell you that most advanced chips come from one place.
This explains how that happened over seventy years through a chain of decisions, each locally reasonable, that produced a supply chain with no redundancy at several critical points. It is a story about compounding advantage, capital intensity, and the way expertise concentrates and then cannot be reproduced elsewhere by simply spending money. That last point is where most casual commentary goes wrong, and the book demolishes it thoroughly. The lithography chapters are the ones I would make required reading.
The machines that print circuit patterns at the current node are arguably the most complex objects ever manufactured, they come from one supplier, that supplier depends on its own web of specialist suppliers, and the knowledge involved is substantially tacit, held in the practices of specific teams rather than written down. Once you understand this, the strategy of building a domestic industry by allocating a large budget looks very different, and so does the effectiveness of export controls. The historical research is real. This is a historian's book with archives, interviews and primary sources, not a journalist's assembly of recent coverage.
The Cold War material on how semiconductor policy was an instrument of superpower competition is genuinely illuminating and mostly unfamiliar even to people who follow the industry. The offset strategy, the role of chips in precision weapons, and the way commercial and military development intertwined are told properly. And it will change how you read the news. Export controls on particular chips, a fabrication plant announced in a new location, a dispute over equipment sales, all of these stop being isolated stories and become moves in a structure you now understand.
That is what a good book about a complex system does. Now the caveats, and the first is one of expectation. People arrive at this book from AI and expect it to be about AI. It is not, particularly.
The AI-specific material is a relatively small portion, concentrated towards the end, and if what you want is an analysis of the compute economics of frontier model training you will find this book spends far more time on the 1970s than on that. The history is the point, and the connection to current AI is largely left for the reader to draw. It is worth knowing that going in. The final chapters date fastest, which is unavoidable in a book about a fast-moving policy area.
The export control regime has been revised repeatedly since publication, new restrictions have been added and some relaxed, fabrication plants announced then have progressed or stalled, and the competitive position of several firms has shifted. The structural analysis holds up well and the specific policy details should be read as a snapshot rather than a current account. There is a narrative tendency I found slightly grating. The book is organised heavily around individuals, the founder who saw it first, the executive who made the bet, the engineer who solved the problem.
It is good storytelling and it sometimes credits people for outcomes that structural forces, capital availability and state policy were doing most of the work on. The book is aware of this and does not always resist it. And it is not technical. If you want to understand what a transistor does, how photolithography works physically, or why smaller nodes are harder, this gives you the shape and not the mechanism.
That is a legitimate choice for a general audience book and it does mean an engineer looking for depth will want something else alongside it. My four point three is for a genuinely important book that fills a real gap in how people in this field understand their own supply chain, researched properly, written clearly, and with an explanation of manufacturing concentration that no other popular account matches. Marked down for being less about AI than its current readership assumes, for policy chapters that have aged, and for a narrative style that occasionally overstates individual agency. If you have never thought about where your GPUs come from, this is the book.