Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/nvidia-will-invest-5-billion-in-intel%E2%80%99s-common-stock.23647/page-4
        )

    [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] => 2021770
            [XFI] => 1050270
        )

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

NVIDIA will invest $5 billion in Intel’s common stock!

Daniel Nenni said:
Jensen said exactly that. Remember, TSMC will not package other foundry die, right now it is TSMC only due to the tight packaging constraints, which I understand completely. Intel packaging has always been open which is an advantage for Intel Foundry, for now.
Click to expand...
Do you know if Samsung offers this as well? Anything interesting you've heard from Samsung regarding advanced packaging? It seems like they are not even a contender, despite touting some interesting technologies on their website.
 
eding42 said:
Not sure if location makes a big difference since TSMC probably chose Phoenix for the same reasons Intel chose it in the 90s - cheap and consistent power, and cheap land. Intel packages its dies in Vietnam anyways, the cost of shipping is very lower relative to other factors.

Plus, Intel's headquarters in Santa Clara is actually just across Highway 101 from AMD's headquarters! They can actually wave to each other from across the freeway. Nvidia's is just a few minutes down San Tomas Expressway. They call it Silicon Valley for a reason.
Click to expand...
it's not close to Phoenix... (or maybe it's far away?)but intel is also planning to build a Advanced Packaging factory in the United States in New Mexico.
 
hist78 said:
But it doesn't mean the jointly developed products will completely do away with TSMC, especially if they are based on TSMC's technologies and ecosystem. Looking ahead, in some cases Intel products may prefer to use TSMC in order to remain competitive when Intel Foundry lacks the capacity, capabilities, or capital expenditure to handle them. Glass substrate based advanced packaging may be one such case. As Lip-Bu Tan said, the best product wins.
Click to expand...
There is still much we don't know about glass substrates.
Recent news has said that Intel has not discontinued glass substrates at all.
In fact, regarding the previous discussion about outsourcing glass substrates, I don't think a glass substrate supplier was specified at that time.
Of course, I don't think it was TSMC.
 
I wouldn't want to be working in Xe/GPU product department at intel today. Must be a lot of people working on updating their resumes or learning NVlink and CUDA. Any bets on how long before they announce Falcon Crest is cancelled? Although Lip-Bu closed the deal, maybe something else Pat deserves some credit for, since work started nearly a year ago.

I suspect this will be positive for intel client sales but negative for margins due to higher cost of nvidia iGPU chiplet. Huge opportunity for intel to grow their CPU/IPU product footprint in AI datacenter and potentially grow the relationship to position foundry to win some of nvidia's business in the future, if they can execute 18A/14A as planned.
 
To counter my own last post, MLID, finally, after years of it actually not happening, claims today is the day Arc actually dies

There’s good points in there, but I’m unconvinced. It’s too big and too lucrative and too existentially important.
 
Brady said:
To counter my own last post, MLID, finally, after years of it actually not happening, claims today is the day Arc actually dies
Click to expand...
> MLID
I will not take anything that person says like there is a public list of his failed prediction in some thread.
Brady said:

There’s good points in there, but I’m unconvinced. It’s too big and too lucrative and too existentially important.
Click to expand...
Intel ARC dGPU may die but not the IP.
 
Maximus said:
www.tomshardware.com

Former Intel CEO lays out radical plan to rescue the company — Suggests ousting Lip-Bu Tan and asks for $40 billion investment from Nvidia, Apple, and others

Craig Barrett was Intel's CEO for seven years.
www.tomshardware.com www.tomshardware.com

"He proposed that Intel’s eight largest customers, including Apple, Google, and Nvidia, should each contribute $5 billion in return for guaranteed domestic supply and pricing leverage against Asian competitors."
----

Jensen listen to Craig Barrett to invest 5 billions on Intel. Next will be other 7 largest customers?
Click to expand...
siliconbruh999 said:
The monopolistic part is if Intel gives up GPU IP in exchange for Money
Click to expand...
TSM's US fabs will be able to provide domestic supply with a proven ecosystem. Intel still has to catch up, if this is even possible for even TSM has to import employees for lack of a domestic supply in many areas.
 
Arthur Hanson said:
TSM's US fabs will be able to provide domestic supply with a proven ecosystem. Intel still has to catch up, if this is even possible for even TSM has to import employees for lack of a domestic supply in many areas.
Click to expand...
Intel will never choose TSMC for their own products unless absolutely necessary you will see the outsourcing to TSMC % dropping by a decent margin and TSMC US fabs are behind Intel US Fabs. There is not a single advantage for Intel to use TSMC fabs in US for their products.
 
Last edited:
Arthur Hanson said:
TSM's US fabs will be able to provide domestic supply with a proven ecosystem. Intel still has to catch up, if this is even possible for even TSM has to import employees for lack of a domestic supply in many areas.
Click to expand...
TSMC's US fabs have too little capacity to make a significant impact. Most logic chips designed by US companies will continue to be manufactured in Taiwan. TSMC AZ will have an annual production of something like 250,000 wafers per year, while TSMC worldwide wafer production was 17 million in 2024.
 
blueone said:
TSMC's US fabs have too little capacity to make a significant impact. Most logic chips designed by US companies will continue to be manufactured in Taiwan. TSMC AZ will have an annual production of something like 250,000 wafers per year, while TSMC worldwide wafer production was 17 million in 2024.
Click to expand...
17 Million Includes lot's of mature nodes that Skews the number too much
 
siliconbruh999 said:
This is very positive for IFS as well all the packing will be IFS
Click to expand...
I think we need to see details. From what I hear

Intel will do customized x86 datacenter CPUs for Nvidia to allow rackscale (This adds to Intel volume).
Intel will integrate Nvidia graphics into Client products. (This gives Intel better products)

The initial believe is that TSMC is the best choice for now for any co-designed chip. they will keep looking at 18A/14A to see if it works.

Every product will have a decision point on what process to use. There definitely is no commitment currently to IFS on any of this.

TSMCs volume and agility means that you can double the demand six months out and get commitments from them. thats the magic of TSMC.
 
MKWVentures said:
The initial believe is that TSMC is the best choice for now for any co-designed chip. they will keep looking at 18A/14A to see if it works.
Click to expand...
Intel Xeons are 100% IFS including DMR and Coral Rapids so for DC it's a given
MKWVentures said:
Every product will have a decision point on what process to use. There definitely is no commitment currently to IFS on any of this.
Click to expand...
While there is the roadmap is set for 2-3 years and outside of client 2-3 CPU Tile everything is IFS
MKWVentures said:
TSMCs volume and agility means that you can double the demand six months out and get commitments from them. thats the magic of TSMC.
Click to expand...
With TSMC running at capacity you need to pay more or prebook
 
Last edited:
siliconbruh999 said:
Intel Xeons are 100% IFS including DMR and Coral Rapids so for DC it's a given

While there is the roadmap is set for 2-3 years and outside of client 2-3 CPU Tile everything is IFS

With TSMC running at capacity you need to pay more or prebook
Click to expand...
I think its pretty clear from both CEOs what foundry is best at this time

On the thought that everything is IFS in the future, Things might change significantly in the next 1-2 years. We shall see.
 
You must log in or register to reply here.
Back
Top