5 ms·
Guess what by putting a factory doing the newest node you educate a whole new generation of engineers that can build and run those next fabs.
by christkv 5y ago
Guess what by putting a factory doing the newest node you educate a whole new generation of engineers that can build and run those next fabs.
- pas 5y agoTSMC buys their next gen stuff from ASML (Netherlands), so the problem is not knowledge transfer, it's good old manufacturing and scaling up. (And that means labor, and traditionally labor laws in the Members States meant that offshoring large scale production was the smart choice ... strategic importance or not, if it doesn't make sense economically and business-wise, then that means it'll be instantly worthless strategically too.)
- lttlrck 5y agoThe buy a tool from ASML. TSMC built up the knowledge to use that tool, to achieve these processes, themselves. Knowledge transfer is key.
- pas 5y agoI'd say the working relationship is the important thing. (yaay, synergy!) Naturally TSMC knows their tools and processes, works with product companies (nvidia, AMD, Qualcomm) to actually implement their design on the target process node. And at the same time works with the various tool vendors to refine the process, the tools themselves. But as lately we saw it's increasingly impossible to do it "alone", both Intel and Samsung works with many-many vendors. (Intel - among others - funded the push for EUV by buying ASML equity in 2012.)
- imtringued 5y agoThat explains why global foundries stopped building leading edge fabs. It only sounds good in the news, in practice it is a money pit, not a competitive advantage. You only do it because of national security.