I still think the non-TSMC fab play is the best direction to go. Only two or three real players compete there, including IFS. All other areas are super competitive; thanks to ARM, chip design does not have such a high entry barrier anymore.
I don't think it can be worse than Intel 7 in any way
Their Laptop CPUs are fine but not that good outside of Lunar Lake and in DC GNR is very competitive
Good Input. I talked to some more people at IEDM. I think I could make an argument that Intel foundries biggest failure is not getting the mature nodes running. I am not sure what happened to Intel 16 (22nm), and the UMC, Tower deal (not the purchase) but I get the feeling from my discussions they might be gone for a variety of problematic reasons for Intel
Yes, Intel has a decade plus of foundry "experience", but prior to Gelsinger's arrival it wasn't experience that adds value. Intel insisted on using their design workflows using outdated and proprietary systems. No one was seriously interested in that and given what TSMC was offering, no one should blame them. I would argue that Intel's prior "experience" was detrimental and taught lessons that had to be unlearned before they could move forward. So from my perspective, Intel has been at the foundry game for < 4 years and has made great strides. And while they still have a very long way to go, they are moving in the right direction.
The problem is Intel 14 and Intel 7 have to be designed on Intel's proprietary systems. They would require customers to invest in those design systems, knowing that nothing else would ever be build using them. Unless Intel were to invest in completely replacing their design systems for those legacy nodes I don't see how they attract customers. Right now the only viable legacy node I see is Intel 3 given the design concerns.
Intel 3 is not a legacy node it is right there in between N5/N3 and it is thier first EUV the best value will be Intel 16 and than Intel 12 using UMC if it happens also there is not much Intel 3 capacity
Apparently, Intel is still having issues in customers relations by not being flexible and it is costing them revenue.
I'm not a design guy by any means, what I have heard is that their EDA tools on the processes prior to Intel 3 are proprietary, and therefore, not industry standard. It is also my understanding that the process design kits are also not industry standard. Both of these issues were significant impediments to success in Intel's previous foundry efforts, though these weren't the only issues by a long shot.
That is a very fair criticism. I'm viewing any process that isn't cutting edge as legacy. I was looking at recent comments by Naga that 18A development was essentially done and now they were focusing you yield improvements. But 18A isn't in production yet and Intel 3 is Intel's leading edge, so you are right. Even by my definition Intel 3 isn't a a legacy node until 18A is in HVM.
I haven't seen any specific reports that Intel's inflexibility has cost them a specific engagement, but I wouldn't rule it out. Discounting Intel 3 that isn't being pushed very hard for foundry, Intel has one process to offer with no variants. I expect it will take a while to have multiple offerings and to learn how to be flexible enough to satisfy more customers. This is a whole new skill set for Intel and not an easy one to master.
Intel (specifically Gelsinger who was way too optimistic IMHO) expected far more foundry volume on Intel 3 that was reasonable. All their financial expectations were set way too high considering the herculean task ahead of them. Consequently, I'm not even to the point where I'm looking at the financials for signs of progress. Once they rolled out their foundry numbers they set an expectation of break even in 2027. With the first Intel 18A product not launching until the end of 2025 Intel will be taking a financial hit to produce 18A wafers that will not count towards revenue until the product launches. That is a typical ramp cost and should weigh on their financials as every ramp does. I don't expect to see significant improvement in foundry financials until 2026 when the begin selling more 18A volume.
I hope Intel will manage this on its own account, but if not, the Government will be needed.
