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!
Panther Lake and whatever Microsoft is building there will be the proof points. If MSFT (the largest buyer of Nvidia chips) builds an AI chip on 1.8nm while everyone else is on TSMC 4nm that will be a big deal.
Recent commentary on HNA at industry conferences has been good from both Intel and ASML, although both could be talking their books. Guess we'll see.
It's notable that the Intel executives (who are steeped in the technology much more than the board) seem to be against a deal.
Panther Lake and whatever Microsoft is building there will be the proof points. If MSFT (the largest buyer of Nvidia chips) builds an AI chip on 1.8nm while everyone else is on TSMC 4nm that will be a big deal.
Recent commentary on HNA at industry conferences has been good from both Intel and ASML, although both could be talking their books. Guess we'll see.
It's notable that the Intel executives (who are steeped in the technology much more than the board) seem to be against a deal.
The Intel 18A vs TSMC N2 battle has already been won by TSMC in regards to design starts. The Intel 18A PDK is still behind TSMC N2 as far as maturity which is very important. Again, it all goes back to trust and the top chip makers are risk adverse for obvious reasons.
The NOT TSMC market segment is where Intel 18A is in play now. Since the Samsung 2nm PDK is in last place this business is low hanging fruit for Intel Foundry. I serious doubt TSMC's foundry market share will change as a result of Intel 18A. Intel will take some away from Samsung though.
Everyone I have talked to (fellow insiders) are against this deal. It makes no sense whatsoever. If 18A does well Intel Foundry is off to the races. If Intel 18A fails then manufacturing will be put out to pasture GlobalFoundries style. I see no deal here until we have an answer on 18A.
The Intel 18A vs TSMC N2 battle has already been won by TSMC in regards to design starts. The Intel 18A PDK is still behind TSMC N2 as far as maturity which is very important. Again, it all goes back to trust and the top chip makers are risk adverse for obvious reasons.
The NOT TSMC market segment is where Intel 18A is in play now. Since the Samsung 2nm PDK is in last place this business is low hanging fruit for Intel Foundry. I serious doubt TSMC's foundry market share will change as a result of Intel 18A. Intel will take some away from Samsung though.
It doesn't matter if there are more customers currently designing on TSMC 2nm. IF Microsoft is able to utilize 18A to make a superior AI chip, everyone else will be behind. Which means they will have to come over on 18AP or 14A for competitive reasons. AMZN I think saw this which is why it's dipping its toe in with the networking chip on 18A it announced last September. But obviously, customers coming over to Intel has been slower than Pat expected. Lagging PDK maturity seems like a fixable problem.
It doesn't matter if there are more customers currently designing on TSMC 2nm. IF Microsoft is able to utilize 18A to make a superior AI chip, everyone else will be behind. Which means they will have to come over on 18AP or 14A for competitive reasons. AMZN I think saw this which is why it's dipping its toe in with the networking chip on 18A it announced last September. But obviously, customers coming over to Intel has been slower than Pat expected. Lagging PDK maturity seems like a fixable problem.
I agree except I seriously doubt Microsoft is using Intel 18A for a mission critical chip. That would be high risk and that is not the Microsoft I know and do not love. Same with Amazon. We are talking about billions of dollars in cloud business at risk, right? Customers are dipping toes into 18A, they will not go all-in (my opinion).
PDKs are certainly fixable but time-to-market isn't. Either you make it through the market window or you don't. AI is an arms race so mission critical is an applicable term.
“All of us at Microsoft are committed to supporting Intel’s efforts to build a strong supply chain right here in the United States. That’s why we are so excited to work with Intel Foundry Services, and why we have chosen a chip design that we plan to produce on Intel’s 18A process,” Microsoft CEO Satya Nadella said during remarks at the conference. https://www.manufacturingdive.com/n...,behind Taiwan Semiconductor Manufacturing Co.
Microsoft is currently behind in custom AI chips and at the back of the line at TSMC. They are buying massive amounts of NVDA GPUs. Their chip design team is all ex-Intel. They have not specified where Maia 2 will be made. Intel has also announced a 3P customer tapeout on 18A in 1H 2025. Could be a different chip from an AI processor, but it also could be.
Microsoft is currently behind in custom AI chips and at the back of the line at TSMC. They are buying massive amounts of NVDA GPUs. Their chip design team is all ex-Intel. They have not specified where Maia 2 will be made. Intel has also announced a 3P customer tapeout on 18A in 1H 2025. Could be a different chip from an AI processor, but it also could be.
Microsoft does a lot of chips at TSMC. Maia 100 used TSMC 5nm. If I remember correctly it taped out in 2022. I heard last year (2024) that Maia 200 was taping out on TSMC N3 but nothing has been confirmed publicly. Maybe Maia 300 is in process on Intel 18A? That would be really cool.
Given the close relationship between Intel and Microsoft I always thought they would be one of the first on 18A. Microsoft however is not an aggressive chip designer like Nvidia or AMD so I doubt Microsoft will be first on 18A with a complex AI chip.
Microsoft does a lot of chips at TSMC. Maia 100 used TSMC 5nm. If I remember correctly it taped out in 2022. I heard last year (2024) that Maia 200 was taping out on TSMC N3 but nothing has been confirmed publicly. Maybe Maia 300 is in process on Intel 18A? That would be really cool.
Given the close relationship between Intel and Microsoft I always thought they would be one of the first on 18A. Microsoft however is not an aggressive chip designer like Nvidia or AMD so I doubt Microsoft will be first on 18A with a complex AI chip.
Yes, that single unconfirmed "report" about Maia2 "expected to be" made on TSMC 3 came from a Taiwanese news publication. It may be true, but that's guesswork at this point. Microsoft has been highly secretive about its recent chipmaking developments.
Microsoft does a lot of chips at TSMC. Maia 100 used TSMC 5nm. If I remember correctly it taped out in 2022. I heard last year (2024) that Maia 200 was taping out on TSMC N3 but nothing has been confirmed publicly. Maybe Maia 300 is in process on Intel 18A? That would be really cool.
Given the close relationship between Intel and Microsoft I always thought they would be one of the first on 18A. Microsoft however is not an aggressive chip designer like Nvidia or AMD so I doubt Microsoft will be first on 18A with a complex AI chip.
I would ask "which Microsoft part is taping out on Intel 18A?" Also Is Microsoft doing the tapeout or is it Broadcom doing a customer part for Microsoft?
