The migration of technological centers.

If you treat civilization as a system that must continually access the laws of the universe — material, mathematical, biological, and now informational — then the centers in this archive are not cities at all. They are interface layers. Mesopotamia is the read-write head for arithmetic and law. Alexandria is the read-write head for geometry and astronomy. Baghdad is the translator. London is the steam-and-finance kernel. Silicon Valley is the device-driver. Each center is the place where civilization briefly aligns its hardware (population, geography, capital, energy) with a previously inaccessible region of universal law.

From Tyre and Carthage through Venice, Antwerp, Amsterdam, London, New York, Hong Kong, Shanghai, Singapore, and Shenzhen, the pattern is consistent: the centers of innovation cluster at harbors. The reason is not romance; it is information. A port is the place where the densest variety of strangers, goods, and ideas converges per unit of land. Variety per unit area is the operational definition of an idea ecosystem.

Almost every major technological discontinuity in this archive runs through a war. The pattern is not romantic. Wars concentrate state capital, suspend peacetime budget discipline, mobilize talent on deadlines, and reward results over publications. Whether one approves of the mechanism or not, the historical record is consistent enough that any complete account of technological centers must reckon with it.

Every center in this archive eventually develops a sophisticated financial layer adjacent to its productive one. The Medici bank finances the Florentine workshop. The VOC finances Dutch shipping. The Bank of England finances the East India Company. Wall Street finances the Erie Railroad and AT&T. Sand Hill Road finances NVIDIA. The pattern is not coincidence. It is an iron law of how civilization-scale innovation gets paid for.

Drop the cost of reproducing a unit of knowledge by 100x, and every layer above the reproduction layer reorganizes. This happened with Babylonian clay tablets (vs. oral tradition), papyrus (vs. clay), parchment (vs. papyrus, in a few respects), paper from China to Baghdad to Europe (vs. parchment), Gutenberg's movable type in 1450 (vs. manuscript copying), and the internet in the 1990s (vs. print). Every one of those step-changes opened a new center somewhere and broke an old one elsewhere.

Three places on earth — Hsinchu (TSMC), Veldhoven (ASML), and Santa Clara (NVIDIA) — control the supply of the silicon that every frontier AI model trains on. No other strategic commodity in modern history has been this concentrated. Oil's choke-points are at least ten cities; rare-earth refining is more concentrated than oil but on commodities with substitutes; compute, currently, has none.

Traditional geopolitical competition has been about territory, energy reserves, populations, and trade routes. Each had a clear unit of account: square kilometers, barrels per day, GDP per capita, shipping tonnage. AI as a strategic resource fits none of those units cleanly. It is closer to a service than a stock, closer to electricity than to oil, closer to a research program than to a finished good. If it becomes the central strategic asset, the form of state competition changes, possibly fundamentally.