This is a pretty dumb take, honestly. Intel for basically forever operated using their own fab exclusively. After failures to maintain good yield rates at their 10nm node, they had the option of continuing to delay new product lines and be eaten by the competition in AMD, or give in to TSMC temporarily while they worked on fixing their fab in parallel. In fact, they were criticized greatly for not switching to TSMC much earlier.
How long do you think fabs take to build and upgrade? Intel was working on fixing 10nm for years, this isn’t a software situation where turnaround times are measured in days or weeks. Going from tapeout to silicon for a single line is a 6 month process after the technology process is solidified, forget if you’re doing it while trying to figure out yield problems.
This is a pretty dumb take, honestly. Intel for basically forever operated using their own fab exclusively. After failures to maintain good yield rates at their 10nm node, they had the option of continuing to delay new product lines and be eaten by the competition in AMD, or give in to TSMC temporarily while they worked on fixing their fab in parallel. In fact, they were criticized greatly for not switching to TSMC much earlier.
The key word is temporarily. How long ago was this?
Calling people dumb then throwing a weak argument doesn’t make it stronger.
They’re on wafer thin margins with vendor lock in. The strategy was not successful.
How long do you think fabs take to build and upgrade? Intel was working on fixing 10nm for years, this isn’t a software situation where turnaround times are measured in days or weeks. Going from tapeout to silicon for a single line is a 6 month process after the technology process is solidified, forget if you’re doing it while trying to figure out yield problems.