> 2) Missed too many ticks and replaced them with tocks. Fell far behind because of ridiculous tilt-at-windmill folly around Itanium and other architectural decisions that absolutely didn't pay off. Ended up taking way too long to get further down on process size.
I applied to Intel's architecture group in 2015, and when I asked one interviewer "does Intel have a strategy if TSMC catches up on process tech?" I was left with a blank stare, as though nobody in the group had contemplated the possibility. I didn't end up working there...
That said, Intel got exceptionally unlucky on which technologies they pursued and which they didn't, so I hope they can turn things around. They missed on through-silicon vias and chiplets, EUV, Cobalt wires, and several low-power transistor technologies.
The fab business is very sensitive to scale, with enough money and discipline you can mostly make your own luck. TSMC does way more volume than Intel and they have a very aggressive customer (Apple) that can help work out early quirks and guarantee orders on new nodes before they're economical for anyone else. I don't know what a turn around might look like but it's a tough problem.
That is revenue based, and not based who makes the chips. In the chart Nvidia, Broadcom and Qualcomm are all fabless (outsourced production) and their chips are made by TSMC, Samsung, or other foundries.
By wafer doesn't sound like the correct comparison as we mostly care about top nodes for this. Revenue actually sounds better to correct for that. Your link also shows that Intel sells their old nodes to others further skewing the per-wafer comparison.
Wafers at 14nm or better would probably be a good metric for who has the volume behind top nodes to then be able to stay in front. If that's TSMC by a large margin losing the mobile and GPU markets has really done a bigger number on Intel than I had assumed before.
I applied to Intel's architecture group in 2015, and when I asked one interviewer "does Intel have a strategy if TSMC catches up on process tech?" I was left with a blank stare, as though nobody in the group had contemplated the possibility. I didn't end up working there...
That said, Intel got exceptionally unlucky on which technologies they pursued and which they didn't, so I hope they can turn things around. They missed on through-silicon vias and chiplets, EUV, Cobalt wires, and several low-power transistor technologies.