NVIDIA CFO Hints at Intel Foundry Services Partnership

[btarunr]

NVIDIA CFO Colette Kress, responding to a question in the Q&A session of the recent UBS Global Technology Conference, hinted at the possibility of NVIDIA onboarding a third semiconductor foundry partner besides its current TSMC and Samsung, with the implication being Intel Foundry Services (IFS). "We would love a third one. And that takes a work of what are they interested in terms of the services. Keep in mind, there is other ones that may come to the U.S. TSMC in the U.S. may be an option for us as well. Not necessarily different, but again in terms of the different region. Nothing that stops us from potentially adding another foundry."

NVIDIA currently sources its chips from TSMC and Samsung. It uses the premier Taiwanese fab for its latest "Ada" GPUs and "Hopper" AI processors, while using Samsung for its older generation "Ampere" GPUs. The addition of IFS as a third foundry partner could improve the company's supply-chain resilience in an uncertain geopolitical environment; given that IFS fabs are predominantly based in the US and the EU.



View at TechPowerUp Main Site | Source

 

[Crackong]

So to circumvent the density related restrictions of the ban ?

 

[wolf]

Interesting if true, I wonder if we'll see split lineups, like the top end chips on TSMC and lower stuff spread across IFS and Samsung.

 

[Assimilator]

So to circumvent the density related restrictions of the ban ?

You post this stupid buillshit in literally every thread about NVIDIA. Go away with your FUD.

 

[Daven]

So Nvidia would love (not actually using) a third supplier which they didn’t name. That’s some great work IFS. Now if you could actually win some fab work that would be something.

BTW isn’t there only three fabs in the world that can offer 10 nm and below. Third place is last place after a company in a country always on the brink of foreign invasion and a dishwasher company.

 

[Eternit]

Intel can't even produce its own GPU.

 

[stimpy88]

nVidia has very few positive relationships with the companies it works with.

 

[Steevo]

Nvidia on 14nm+++++++++++ doublegood.

 

[Squared]

Nvidia's CEO expressed an interest in using Intel's foundries a long time ago, I think in the 22nm days when Intel was dominant. At the time the two companies were almost more like partners than competitors because they had very little overlap in their product offerings and a lot of products sold had an Intel CPU and Nvidia GPU.

Now that Intel's foundries aren't the best, it'd look very good for Intel if Nvidia became an IFS customer.

 

[Daven]

Nvidia's CEO expressed an interest in using Intel's foundries a long time ago, I think in the 22nm days when Intel was dominant. At the time the two companies were almost more like partners than competitors because they had very little overlap in their product offerings and a lot of products sold had an Intel CPU and Nvidia GPU.

Now that Intel's foundries aren't the best, it'd look very good for Intel if Nvidia became an IFS customer.

I was thinking the opposite. Intel could bow out of the GPU business and fab/license GPU tech for/from AMD.

Btw, Intel, AMD and Nvidia 100% overlap as of right now. There is no product line that isn’t made or about to be made by one of those three.

 

[Crackong]

You post this stupid buillshit in literally every thread about NVIDIA. Go away with your FUD.

'Performance density' is one of the US export ban criteria.
No matter what you think how Stupid bulls***** it is, it is fact and should be considered.
As Nvidia themselves are already tried the traditional 'cut down' method (RTX 5880 / 4090D) and received harsh feedback from the US department.

Finding a way to produce equal performance but in lower density is one legit way to circumvent the export ban.

If it is your fanboyism somehow swaying you away from looking at facts and details, maybe it is you who should go away.

 

[unwind-protect]

nVidia has very few positive relationships with the companies it works with.

Like eVGA?

 

[Dr. Dro]

'Performance density' is one of the US export ban criteria.
No matter what you think how Stupid bulls***** it is, it is fact and should be considered.
As Nvidia themselves are already tried the traditional 'cut down' method (RTX 5880 / 4090D) and received hash feedback from the US department.

Finding a way to produce equal performance but in lower density is one legit way to circumvent the export ban.

If it is your fanboyism somehow swaying you away from looking at facts and details, maybe it is you who should go away.

What harsh feedback? Releasing neutered products at just the edge of the legislation, that just about meet the demands of the government is not circumvention, it's compliance. Besides, this does effectively nothing but stifle innovation anyway. If Xi wants to use GPUs for mischief, he will.

Nvidia on 14nm+++++++++++ doublegood.

The hilarious part about it is that no, Intel actually has their nodes in working order, they just don't have a working CPU design for them. It's very unlikely that they will downgrade the node, given Ada is on TSMC N4, and even Samsung 8N used on Ampere was an improvement on Turing's TSMC 12FFN node. Nvidia, like Apple, never goes backwards.

Gelsinger's literally betting the company on IFS at this point, as Intel's got multiple stalled projects and isn't being able to deliver consistently.

 

[Crackong]

What harsh feedback?

This feedback from the US commerce department just a week ago.

US Commerce Secretary had pointed words for Nvidia’s strategy of designing new variations of chips to slide under the TOPS power limit. “If you redesign a chip around a particular cut line that enables them to do AI, I’m going to control it the very next day.”

 

[Dr. Dro]

This feedback from the US commerce department just a week ago.

They'll do it, Nvidia will release something slightly weaker to comply, and it'll repeat until we've got nothing but calculators LMAO

 

[Crackong]

They'll do it, Nvidia will release something slightly weaker to comply, and it'll repeat until we've got nothing but calculators LMAO

It will get to a point it is not financially feasible to Nvidia and their partners to keep that many SKUs co-existed and supported for the life cycle.

But I think they won't come to that extreme.
Nvidia may come up with some kind of low density/high energy consumption solution for China export.
So Xi could do its AI, but in a slow and inefficient way.
The commerce department might look over China energy crisis in the winter/industrial segment and accept this as an intermediate solution on the AI ban / trade / sweet money from China problem.

After all their intention is to slow down China's AI development.
And a stop some/allow some measure is always better than a total ban.
A total ban just isn't viable and induce smuggling which hurts everyone but the smugglers.

 

[Why_Me]

Like eVGA?

Is EVGA still in business? It's like they just disappeared off the face of the planet.

 

[unwind-protect]

Is EVGA still in business? It's like they just disappeared off the face of the planet.

They continued all their products except videocards. But that seems to be a smaller overall business now. In any case they seemed to be deeply dissatisfied with NVidia.