Intel is going to have to earn it back with CPUs. They have shredded their credibility, and AMD is eating their lunch. Gonna be a long uphill battle for them.
I share your point of view, RTN. Intel lost a lot of market share this gen but they lost way more mind share. People who didn't upgrade in the last year plan to buy AMD, when they do. Intel can come back, and I hope they do, but its a steep uphill climb. I believe it is very likely, Intel will end up in chapter 9 or 11 receivership. The government seems very likely to bail them out for national security reasons but I assume the fab will be split from the design side. I'm trying to figure out what would be made in Intel fabs. Not sure. Intel doesn't have a 3nm node. They struggled with their 7nm node and I don't think it is yielding well, even now. I'm not even sure they scaled it, much. If Intel get 18A online in time for Zen 6 in 2026, and they very well might, AMD could produce their cores on 18A and have TSMC do the IOD, NPU, GPU. At some point, TSMC will probably be forced to bring their 2nm node into the line cost wise, so they can sell it to companies other than Apple. That won't leave a lot for the rest of the industry.
Its going to come down to Samsung. If Samsung can scale their 2nm node, Intel is toast. If Samsung can't, and their track record isn't good lately, Intel could benefit from the insane price model of TMSC's N2P node.
Rumors are that Intel will bring Nova Lake to market in 2026. That's a long, long time to not have a competitive product in the market. What's more, the top spec Nova Lake part will have 16 performance cores, 32 efficiency cores, and 4 low power cores. It will be a new socket. That's a lot of cores but we've seen that people ignore non-performance core count in this generation. Intel has a lot of education and software to take care of before the release.
I just read an article that was incredulous that Intel is basing their entire recovery on the 18A process. We've arrived at the point where it's not about outlier articles which are bad. It's about finding the odd, rare, article which is reasonably close to reality. Before someone throws down their issue with Intel's need for a 2nm process node, they should be required to explain how a CPU vendor can remain relevant selling CPUs that are more than one process generation old. It can't be done. The only two viable possibilities are that Intel drops their fabs or they keep their fabs close to the leading edge. I'm not even sure the best possible outcome is for 18A to succeed. If 18A is a viable process, that will trigger jubilation, parades, and Altoid blowjobs but that is the first step of a long journey. From there, they will have to keep up with TSMC and, to a far lesser extent, Samsung in the process race. If they drop back again, a successful 18A won't make all that much difference because it's not like Intel is going to flood the world with a significant number of 18A chips. If Intel fabs were sold to AMD, they could make AMD's IOD and possibly infinity cache on their 4nm class process. It might be good enough for this purpose (keep in mind their 4nm class process is a renamed 7nm class process). This would allow AMD to move IOD production in house and let them use contracted TSMC capacity for CPU/GPU production. AMD can't succeed with old process nodes, either. If they purchase Intel Foundry, it will just shift the funding burden to the more wealthy company. AMD might do better by shifting some TSMC production to Samsung. No capital cost and far, far less risk.
