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!
It was previously reported (incorrectly) that TSMC would only raise AI wafers to monetize the AI surge. Personally, I agree with Morgan Stanley here, it will be an across the board price increase with 5nm, 4nm, and 3nm. Remember, this will take time as there are wafer agreements in place with the top customers but new agreements or customer modified agreements (for more wafers) are fair game. You should also expect a rise in packaging pricing (CoWoS) since demand exceeds supply.
The bottom line: Intel Foundry has a real opportunity to get some business from the big TSMC customers who want a trusted alternative.
When Intel entered the foundry business with IDM 2.0 I was impressed. Yes, Intel had tried the foundry business before but this time they changed the face of the company with IDM 2.0 and went “all-in” so to speak. The progress has been impressive and today I think Intel is well positioned to...
It was previously reported (incorrectly) that TSMC would only raise AI wafers to monetize the AI surge. Personally, I agree with Morgan Stanley here, it will be an across the board price increase with 5nm, 4nm, and 3nm. Remember, this will take time as there are wafer agreements in place with the top customers but new agreements or customer modified agreements (for more wafers) are fair game. You should also expect a rise in packaging pricing (CoWoS) since demand exceeds supply.
The bottom line: Intel Foundry has a real opportunity to get some business from the big TSMC customers who want a trusted alternative.
Couple questions @Daniel Nenni
How do price increases work. My history on outsourcing is that prices are renegotiated frequently (Quarterly or yearly at least), But I think you have told me at TSMC things are negotiated long term with pricing and volume on leading edge. little flexibility.
I do know that pricing is very different between customers. Apple and AMD do not pay the same price. Low volume customers pay much more. but does it change a lot?
Intel's opportunity is there IF they demonstrate ability to deliver as committed. When does Fab 52 start shipping production? It isnt 2024.
TSMC will more than double packaging capacity this year and will more than double it again next year. Brilliant move by TSMC, packaging is a big differentiator.
Again with these 20k per wafer numbers from financial analyst. No way TSMC charges more than 3-4x the wafer cost on N5 family wafers. If I was a betting man based on the process complexity and TSMC margins, wafer price is closer to the 10-12k range and N7 might be sold for like 5-8k.
Couple questions @Daniel Nenni
How do price increases work. My history on outsourcing is that prices are renegotiated frequently (Quarterly or yearly at least), But I think you have told me at TSMC things are negotiated long term with pricing and volume on leading edge. little flexibility.
I do know that pricing is very different between customers. Apple and AMD do not pay the same price. Low volume customers pay much more. but does it change a lot?
Intel's opportunity is there IF they demonstrate ability to deliver as committed. When does Fab 52 start shipping production? It isnt 2024.
thanks!
Depends on the contract, but yes there is price protection. Apple has most favored nation status so no one gets wafers for cheaper than Apple.
The funny thing is that everytime I read about TSMC wafer pricing in the media it is wrong. Scott Jones can tell you what it costs to manufacture wafers at a process node and if you look at TSMC's margins you can get a general idea of pricing but nothing customer specific.
I don't know how packaging works in the contracts but I do know TSMC packaging pricing is going up.
My understanding is advance package like CoWoS changes the price structure and bundles or integrates the business into next level. For example in AI CoWoS chip, if there is low yield issue, can we do re-work? or need to scrap all chiplets? If all chiplets needs to be scrapped, who will pay the lost in cost and delivery. Can OSAT afford it, pay and request tsmc/Samsung/intel/Micron/hynix to deliver the 3nm/2nm/HBM chips in hot run? It is tricky that the foundries provides the whole solutions stand on choking points. Will intel become trust alternative of advanced package supplier for AI chips? It is not easy and takes long time. I believe it will not happen in the short term.
x.com
semiwiki.com