You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
TSM has been dealing with inference issues at 2nm production. The question is will this experience help TSM in building advanced AI chips in the future? Any clarification or views on this would be greatly appreciated. Thanks
TSMC deals with manufacturing issues, and they don't care if the chip they are fabricating is for AI training or AI inferencing, it doesn't matter for them. TSMC takes orders to manufacture chips, but the chip function doesn't come into play. Yes, TSMC has a library of IP to offer chip design customers, but how the customer assembles those IP blocks is up to each customer. TSMC doesn't tell its customers what to design.
TSM has been dealing with inference issues at 2nm production. The question is will this experience help TSM in building advanced AI chips in the future? Any clarification or views on this would be greatly appreciated. Thanks
Dan, I looked for my original reference and could not find it, but I did run across this and TSM working with inference. If you have any questions I will look further. I read about TSM running into inference at 2nm originally but could not find the same reference. Thanks for the heads up, it made me do some research of value to me.
