When a foundry says a chip is built on a leading node, the number no longer describes a physical measurement. That gap matters for how you read every chip announcement.
A quick read — the essentials, fast.
Read any chip launch and you will see a node name attached to it: a number followed by a letter or two. People treat that number as a size, as if a smaller one means smaller transistors and therefore a better chip. That stopped being true years ago, and the gap between the label and the silicon is now wide enough that you have to ignore the number to understand what was actually announced.
The number is a marketing label, not a ruler#
A node name used to roughly track a physical feature you could point to on the wafer. Once manufacturers moved past the simpler planar transistor designs, the naming detached from any single measurement. Today the figure is closer to a generational badge than a dimension. One company's leading node and another's can carry similar names while differing meaningfully in transistor density, power behavior, and cost.
The practical consequence: comparing two chips by node name alone tells you almost nothing. A chip on a numerically larger node from one foundry can outperform a numerically smaller one from another, depending on the design, the libraries used, and how aggressively the part is clocked.
What to read instead#
When you strip away the label, three things actually move generation to generation:
- Density. How many transistors fit in a given area. This is what lets a design add cores, cache, or accelerators without growing the die.
- How much energy the same work costs. Efficiency gains, not raw speed, are where most of the recent progress lives.
Discussion