I felt this was strongly, and felt incredibly sure of that, until I read Stratechery on it last week:
https://stratechery.com/2024/intel-honesty/It slipped by me how much AMD eating a little more every quarter really added up over the last ~5 years. (~9% to ~50% of datacenter revenue)
Also, the foundry stuff just isn't working as a financial exercise: they can't split it out as a division and massively subsidize it and seem responsible. And there's ~0 light on the horizon. No one wants to use it, even Intel is falling back to using TSMC's leading nodes now, which is letting it tread water against ARM in laptops.
If their confusing plan to do 5 nodes in 2 years or whatever works out, that'd enable them to start reversing the tide: they'd still have to build the same muscles TSMC has from always being a foundry, like having generic designs that are already in the market available, and convince people to switch suppliers, which is always risky, and usually done over years.