In the autumn of 2025, a shipment of a few thousand graphics processing units became a matter of state. It was tracked the way an earlier era might have tracked a tanker of crude oil — flagged by intelligence agencies, debated in congressional hearings, and eventually caught up in a diplomatic tug-of-war between Washington and Beijing. The chips in question were not weapons in any conventional sense.

They could not be fired, dropped, or detonated. Yet in the calculus of twenty-first-century power, a rack of advanced processors is increasingly treated as strategic cargo, no less consequential than a shipment of enriched uranium. Welcome to the age of silicon sovereignty, where the ability to compute has quietly become the ability to command.

For most of the twentieth century, geopolitics was organised around the control of physical resources — oil fields, shipping lanes, rare metals buried under contested soil. Nations rose and fell on their ability to secure and refine these inputs. Today, a parallel resource has emerged, one that is manufactured rather than extracted, but which behaves in remarkably similar ways: computing power, or what industry insiders simply call "compute." The training of advanced artificial intelligence models requires staggering quantities of specialised chips, arranged in vast data centres that consume as much electricity as small cities. Whoever controls the supply of these chips, and the factories that make them, effectively controls the pace at which the rest of the world can build the technologies of the future.

This is not a metaphor stretched for effect. Since 2022, the United States has imposed a widening set of export controls on advanced AI chips and the equipment used to manufacture them, aimed overwhelmingly at slowing China's access to frontier computing capacity. The restrictions have evolved in fits and starts —........

© The Nation