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!
Couple Items:
The Fab 34 Apollo deal does not change Intel capacity. Intel was the only customer. Intel is given cash to Apollo to go away
I think Intel made announcements on the plans for Ohio about a month ago. It was a push out or clarification on pushout. 2030 best case. I am not sure if...
I (and many media people) total missed that. thank you.
So terafab is all things, everywhere, all at once..... but it doesnt exist yet.
Lets wait until someone presents a actual agreement and committment.
This is actually good for Intel as they could possible unload all their unused shells...
Terafab is supposed to be 25B..... with memory and Logic and packaging. That is a pretty small factory (Micro Fab? Kilo Fab?) REALITY? Its a made up affirmation. It should say "Elon plans to do something someday and Intel plans to talk with Elon about it"
Something major may be in the works...
One thing I went through at Intel and as a consultant to other compainies.... Once you skip a node (or 3) the cost goes up. Its very hard to get back on. its like starting greenfield fab and greenfield R&D. Maybe buy old Intel facilities or old TSMC/Samsung facilities. Intel took way longer...
I tihnk I will wait to see what this really is before concluding. We have a mythical memory and logic fab with packaging. Then we have a twitter of a partnership between the mythical fab and the manufacturing company that loses 8B per year.
My current model is that this will have no positive...
What exactly will Intel be doing on this? I think we need to know a few more details. Was there a filing or even official announcement? (twitter is not an official announcement) @Daniel Nenni
So the only announcement is on Twitter? Is there a actual announcement? SEC filing?
the word " refactor" is a interesting term.
Announcement Seems too vague at this point
I will be interested in the details. If Intel is licensing/sharing the technology, this is a great win. Got to find a way to pay for the billions in spending.
100% agree with you. I think Intel needs to look at "value options" to end this Intel 10nm and Intel 7 nonsense (I am shocked at how many CPUs Intel ships on 5 year old node). Per CFO, over 10% of Intel revenue comes from EUV in Q4 2025 LOL.
@Fred Chen
the die size is actually well published. techinsights has tons of info (Public). you don't need a detailed teardown to get die size. you can get it from xrays or working units. you can even get it from package layout to show the die dize is not 3x.
The HBM3 die size is is less...
I have met Pat before. He is probably the most optimistic executive I worked with at Intel (Many years ago). I like that trait.... just not in a CEO....
Actually we have a cost model for the final product. We have a model for TSM prices, die size, yield etc.
IFS sells wafers to Intel Products are market price. Once Fab 52 is fully loaded, and assuming they hit the yield roadmap, the IFS gross margin on the CPU will be slightly positve. OM is...
I would take 50% of LBT vs 100% of PG. The stockholders (owners of the company) are very happy with 50% LBT. If we need failed strategies, 30% decrease in revenue and a 60% decrease in stock price, we know who can deliver that.
They NAND companies need DRAM to stack in packages. Nanya and other companies are saying we will only guarantee capacity if you buy into company. Its a fairly standard procedure in these tight times
I have a pretty good model for 18A wafer cost today and what it will be when fully loaded in Fab 52.
I do not have great model for Intel 3 other than the fact that the ramp of Fab 34 does not seem to be helping IFS margins. and Intel says wafers are sold at market price (I used N3 price even...
It was not fully ramped last time I checked (December). According to IFS metrics (sell at market price to business unit), it was not profitable in our model. Intel has been sticking with the commentary that Intel 7 has higher margins than INtel 3, 4, 18A currently.
Let me know if you have...
It was mostly Pat. DZ is a good soldier. It would work if the ramp ever came but it did not.
We need to make sure we look at actuals vs plans and models. the amount of spending is legally documented in 10Q. total was about 10B brookfield, 10B Intel total to date last quarter (correct me if I am...
