Ask any industrial buyer which parts give them the most trouble and the answer is rarely the exotic ones. It's the boring ones: the 8-bit MCU that's been in the design for a decade, the op-amp, the CAN transceiver, the voltage reference. Parts built on mature process nodes - 40nm, 90nm, 180nm and older - that should be commodity items, and increasingly aren't.

The structural mismatch

The economics are simple and unforgiving. Foundries earn their best returns on advanced nodes, so that's where capital expenditure goes. Mature-node capacity mostly doesn't get expanded - it gets maintained, sweated, and eventually retired. Meanwhile, demand for mature-node silicon keeps growing: cars carry more electronics every year, factories keep automating, the grid is being rebuilt, and none of that runs on bleeding-edge logic. More demand, static-to-shrinking supply. That's not a cycle; it's a trajectory.

Why nobody builds their way out

A new fab takes years and billions of dollars, and a mature-node fab earns commodity margins on parts that sell for cents. No board approves that investment without guaranteed long-term offtake - which customers of commodity parts historically refuse to commit to. Some capacity has been added in recent years where governments subsidized it, but the additions are modest against demand growth, and much of the new capacity is already spoken for by automotive long-term agreements.

What it looks like from the buyer's seat

  • Lead times that never came back down. After each disruption, mature-node parts recover slower than advanced ones - and settle at a longer baseline.
  • Sudden single-sourcing. A second source exits a product line, and a dual-sourced design quietly becomes single-sourced without anyone updating the risk register.
  • EOL by process retirement. When a supplier retires an old fab process, every part on it dies together - which is why discontinuation notices arrive in batches.
  • Price firmness in a soft market. Even when the broader market corrects, mature-node pricing holds, because the capacity discipline is structural.

Designing and buying around it

  • Score BOM risk by node, not just by part. A design with thirty parts on retiring processes carries different risk than one with three.
  • Hold strategic inventory on single-sourced legacy lines - bonded stock with scheduled releases keeps cash impact manageable.
  • Qualify alternates before you need them. The cost of qualification is predictable; the cost of a line-down isn't.
  • Watch supplier roadmaps - process-retirement announcements are the earliest EOL warning you'll get.

The takeaway

The legacy node squeeze rewards buyers who plan in years and punishes those who buy quarter to quarter. We help customers map node exposure across their BOMs, hold buffer stock against single-sourced lines, and move early on last-time-buys - because for these parts, the market doesn't forgive late reactions.