How Semiconductors Became Strategic Infrastructure

01From commodity to chokepoint

For decades, chips were bought like any industrial input. Then the industry consolidated to a handful of firms at the leading edge, concentrated in a few square miles of the planet — and every government noticed at once.

• The foundry model and why leading-edge manufacturing concentrated
• One node, one region: the geography of advanced logic
• The pandemic shortage as the moment chips became policy

02What a fab actually demands

A leading-edge fab is one of the most extreme pieces of infrastructure ever built — by capital, precision, and utility demand. Calling it a factory understates it.

• Capex measured in tens of billions per site
• Ultra-pure water, ultra-clean power, and vibration tolerances measured in nanometers
• Campus power demand approaching gigawatt scale — a utility planning event in itself

03The industrial policy era

CHIPS-style programs in the U.S., Europe, and Asia have made fab siting a matter of statecraft. Subsidies move tens of billions; the constraints that remain are physical.

• What subsidies can buy — and what they cannot: talent, water, power, time
• The Arizona, Ohio, and Texas buildouts as live experiments
• The critics case: cost per job, and whether the leading edge can be transplanted

04The feedback loop

Chips need power. The modern grid increasingly needs chips — in inverters, sensors, protection, and controls. And AI needs both. The three systems are converging into one planning problem.

• Semiconductors inside the grid: power electronics as grid infrastructure
• AI demand pulling on both fabs and generation simultaneously
• Why chip, compute, and power planning can no longer be separated

05What we are watching

• Advanced packaging: the quiet bottleneck behind the headline fabs
• Fab–utility co-planning agreements
• Whether announced capacity actually reaches high-volume production