There is such a thing as "good enough".
For a smart phone? Yes.
We are paying a high price to keep them from getting the last iota of performance.
The difference between "7 nm" and later process nodes is a lot more than an "iota". When you account for not only the single-thread performance, but also the aggregate performance you get when scaling up multi-core CPUs, GPUs, and AI processors on newer nodes, it's actually quite a big difference!
For instance, just look at the difference between Nvidia's A100, which used TSMC N7 (unlike the rest of the Ampere line), compared with the H100 (which used "4N" - a custom variant of TSMC N5).
Or, look at the difference between AMD's Genoa, which scales up to 96 cores at much higher performance levels, compared with Milan. That's essentially the difference between TSMC N5 vs. N7 - just
one node.
Now, you might claim that there other differences between them than just the node, but those differences were largely
enabled by the power and density improvements delivered by the new node. When you try to backport a newer design to an older node, it runs way hotter and uses a lot more area, if it's even viable at all. Just look at Rocket Lake.
It has cost US companies 25 or 30 percent of their customer base, it will not be made up by setting up new expensive fabs in the US, they still need customers.
That's not the point. The stated purpose of the sanctions is aimed at the
military applications of advanced technology.