Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/intel-might-get-nvidia-as-a-foundry-customer-soon-likely-to-produce-5000-h100-wafers-per-month.20727/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Intel Might Get NVIDIA As A Foundry Customer Soon, Likely To Produce 5000 H100 Wafers Per Month?

While Jensen has stated that Intel-built chips look good & that he is open to using Intel Foundry

I do look at Nvidia using Intel Foundry with heavy skepticism especially giving that Pat/Intel hasn't exactly been friendly towards Nvidia

Common sense would say you'd treat any whale (big potential customer) with kid gloves but that's apparently not Pat's approach.
 
Guess packaging only, TSMC CoWoS 20,000 wafers per month by end of 2024, Intel Foveros or EMIB 5,000 wafers per month, a similar news article from Taiwan in 2024-Jan
 
Guess packaging only, TSMC CoWoS 20,000 wafers per month by end of 2024, Intel Foveros or EMIB 5,000 wafers per month, a similar news article from Taiwan in 2024-Jan
It says 5000 H100 Wafers only not total capacity btw the source is the worst so i will take it with a grain of salt
 
I think this is the same rumor from 3 months ago. I believe the investigation stated that Intel and Nvidia and TSMC had reached preliminary agreement for intel to do "CoWoS" packaging.... but I understood it to be a backup plan for TSMC overflow. I am not sure how much Advanced Packaging compatability matters.

Considering the issues with current advanced nodes, I think Customers are on a wait and see on using Intel to foundry wafers
 
I had hoped some of the big companies would use Intel 18A for chiplets as a foundry test run. Still possible even though I have not heard of any in the hallways. I had hoped Pat would have better foundry news on the investor call. A packaging win does not move the foundry needle, my opinion.
 
Is Intel sourcing the silicon slivers for TSMC's CoWoS-L by any chance? Would that be the 5000 wafers material spoken of here?

Btw looks like CoWoS-L is having some birthing pains if the Nvidia delay stories are true.
 
I had hoped some of the big companies would use Intel 18A for chiplets as a foundry test run. Still possible even though I have not heard of any in the hallways. I had hoped Pat would have better foundry news on the investor call. A packaging win does not move the foundry needle, my opinion.
How much did the multiple delays announced this year (Ohio, Germany, Israel) impacted the decision to use Intel or not
vs the degrading financial situation
vs just the fact that they need chips sooner than what Intel will be able to mass produce?
 
Last edited:
I had hoped some of the big companies would use Intel 18A for chiplets as a foundry test run. Still possible even though I have not heard of any in the hallways. I had hoped Pat would have better foundry news on the investor call. A packaging win does not move the foundry needle, my opinion.
I have a question: What is the cost to initiate an 18A test run, including design and masks? I'm concerned that the cost of a test run for nodes beyond 2nm might already be prohibitively high, making it impractical even as a backup option.
 
How much did the multiple delays announced this year (Ohio, Germany, Israel) impacted the decision to use Intel or not
vs the degrading financial situation
vs just the fact that they need chips sooner than what Intel will be able to mass produce?
I guess the Israel delay is permanent they would rather leave the deal than to take it in such a situation in which Israel is also they delay must be due to lack of customer Arizona was accelerated due to pre payment if they get the contract they will accelerate it
 
I have a question: What is the cost to initiate an 18A test run, including design and masks? I'm concerned that the cost of a test run for nodes beyond 2nm might already be prohibitively high, making it impractical even as a backup option.
I would behoove Intel to offer test runs for free. Or even subsidize the costs for companies who want to test. As it is not free for companies to come up with test designs to send to Intel in the first place.
 
Back
Top