Intel doesn't have any fabs in China.
What does Intel-China currently create, and what do they have access to from the US and other Intel sites?
You forgot intel 3 product 24
If memory servers these dates above were dates for process completion rather than product launch dates.
Yes and no. Intel 4 will be obsolete when intel 3 comes out and intel 20A will be obsolete with 18A. Not much different than how intel 7 made 10nm SF obsolete or how N5P made N5 obsolete. Intel said that these nodes were forward compatible though so this isn’t an issue. Intel 3 will still serve a purpose after 18A is out because it is their best finFET and it will be farther along it’s yield/depreciation curve(s). To say intel 3 would be obsolete would be like saying N7 was obsolete after Apple made their first N5 chips.
For one AMD still sells zen 2 parts as Ryzen 7000 series and they still fly off the shelf. Two keep in mind the products and dates.
By this logic could they not recoup the rapid pace of investment between 14, 14+, 14++, 14+++, and 14++++ one after the other? Intel 3 is an enhanced version of intel 4 and Intel 18A is an enhanced version of intel 20A, so there is probably very little extra expense to recoup. Intel has also said this was part of the beauty of foundry. That way intel can actually make money after the fabs depreciate rather than hoping tools could be jury rigged to work for new nodes when fabXX converted from say 90nm to 32nm. Heck they even used to be on a 4yr tool deprecation curve because of how rapidly fabs were being converted to their newest nodes.
By this logic we should have 100% confidence intel 4 is launching this year. Tigerlake in September 2020 (4 years later than 2016) was the first top to bottom mobile lineup on the 10nm family and finally had compelling PPW over 14nm revisions. In my opinion intel 4 is a far less ambitious node than 10nm, and is coming 5 years after that initial statement on when 7nm products would ship.
Perhaps (If they continue the 4-5 year delays as you highlight). So when would a 18A product ship in your opinion?
The big difference here is that QCOM is driving 18A so do not lump it in with the others. QCOM is a force of nature for the foundries.
He says as he types his message into a CPU with at least 3 billion transistors...
The clear implication of what you just wrote here is that schedule/delivery failure is tolerated for Intel's internal projects, but not for Qualcomm ... . While I can imagine that might be the case, it certainly shouldn't be.
Don't think I'm disagreeing about the benefit and discipline of working with Qualcomm. I'm merely observing that an external customer appears to be necessary - in my experience it often is. But perhaps Qualcomm has a more immediate market need for 18A than the internal designs.
