News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products

[Saylick]

Charlie says JHH says that it is Windows

JHH would blame his own mother if it meant saving Nvidia from a PR mishap. /s

 

[Joe NYC]

The SoC die is 18A the iGPU is 18AP as for cancellation of 4+0 unlikely cause it succeds 4+0+4 PTL SKUs.

I didn't mean cancelled such as nothing would replace it. I meant like what happened with Arrow Lake, where the bottom end die on Intel 20A was cancelled and replaced by TSMC version of the same die.

 

[511]

I didn't mean cancelled such as nothing would replace it. I meant like what happened with Arrow Lake, where the bottom end die on Intel 20A was cancelled and replaced by TSMC version of the same die.

Tbh ARL was always planned to be N3B it wasn't until 2021-22 they decided to have 20A variant than they had a record quarter.

 

[mikegg]

Why would Nvidia do this?

1. Appease Trump

2. Keep Intel alive as a potential fab to negotiate better prices from TSMC

3. It's a way to test Intel 14A without fully committing. If their iGPU port to 14A does well, they might considering moving more of their chips over.

4. They want to put RTX in as many devices as possible to counter Qualcomm, AMD, and Apple. They want AI inferencing to work on an RTX GPU, integrated or not.

 

[poke01]

They want to put RTX in as many devices as possible to counter Qualcomm, AMD, and Apple. They want AI inferencing to work on an RTX GPU, integrated or not.

This. Quickest way would be to partner with Intel. Kill 3 birds with one stone.

 

[511]

I hope Intel is not as dumb cause inference market is pretty much game on right now unlike training

 

[mikegg]

This. Quickest way would be to partner with Intel. Kill 3 birds with one stone.

I thought Nvidia would exclusively partner with Mediatek. But clearly, they see themselves as open to any CPU maker to integrate their RTX GPUs into the SoC.

I'm guessing if AMD, Qualcomm, Apple came calling, Nvidia would be happy to add an RTX GPU to their SoCs.

Anyway, Nvidia clearly knows that SoCs with their unified memory architecture is the way to go in the AI era. They need to put an RTX GPU in as many SoCs as possible to maintain relevance in the changing laptop/desktop world.

This is the result of the failed Arm acquisition. Had that been successful, I'd bet Nvidia would have just designed and sold the entire SoC with RTX + Arm CPU without using 3rd parties like Intel and Mediatek.

 

[Saylick]

This is the result of the failed Arm acquisition. Had that been successful, I'd bet Nvidia would have just designed and sold the entire SoC with RTX + Arm CPU without using 3rd parties like Intel and Mediatek.

Nothing stopping Nvidia from making their own ARM SoC already. Look at their Tegra lineup. Grace already uses off-the-shelf ARM cores. What Nvidia don’t have is an x86 core.

 

[mikegg]

Nothing stopping Nvidia from making their own ARM SoC already. Look at their Tegra lineup. Grace already uses off-the-shelf ARM cores. What Nvidia don’t have is an x86 core.

Nothing is stopping them but they'd have to make a better SoC than Qualcomm and Mediatek. They'd also be competing against AMD and Apple. Intel is a meh. Not that easy without Arm's talent in house.

They've clearly chosen a very interesting strategy which is to integrate with any SoC maker who wants an RTX GPU.

 

[Saylick]

Nothing is stopping them but they'd have to make a better SoC than Qualcomm and Mediatek. They'd also be competing against AMD and Apple. Intel is a meh. Not that easy without Arm's talent in house.

They've clearly chosen a very interesting strategy which is to integrate with any SoC maker who wants an RTX GPU.

Frankly, I’m surprised they haven’t bought out a small design firm with SoC experience, a la what Apple did with PA Semi. Nvidia have a metric crapton of cash now, yet the best they can do with it is issue a 60B stock buyback.

 

[jpiniero]

Nothing is stopping them but they'd have to make a better SoC than Qualcomm and Mediatek. They'd also be competing against AMD and Apple. Intel is a meh. Not that easy without Arm's talent in house.

They've clearly chosen a very interesting strategy which is to integrate with any SoC maker who wants an RTX GPU.

That's not it at all. The threat (to nVidia, and Intel) is XBoxPC gaining momentum. No guarantee that will actually happen but you can see that they would want to be prepared just in case.

 

[mikegg]

That's not it at all. The threat (to nVidia, and Intel) is XBoxPC gaining momentum. No guarantee that will actually happen but you can see that they would want to be prepared just in case.

Eh... That's not the biggest threat. Gamers always think they're more important than they really are. Xbox PC gaining momentum would increase PC discrete GPU demand. Good for Nvidia.

This is AI inference mostly. Period. The biggest threat to Nvidia is Apple (and AMD) winning local AI inference due to superior SoCs.

 

[Joe NYC]

This is AI inference mostly. Period. The biggest threat to Nvidia is Apple (and AMD) winning local AI inference due to superior SoCs.

True, and also, in general, NVidia recognizes that on client, their 2 chip (CPU, GPU) solution is doomed and will inevitably be surpassed by single chip APU.

 

[adroc_thurston]

NVidia recognizes that on client, their 2 chip (CPU, GPU) solution is doomed and will inevitably be surpassed by single chip APU.

Less doomed and more like "sooner or later AMD will prybar the market via die reuse".
Just that it's happening sooner.

 

[Dana599]

I have many thoughts but isn't this anti-trust... but I guess in todays political climate that is useless to consider

You make a good point—on paper it does look like an antitrust issue, but with how regulations are enforced (or ignored) today, it often feels like those concerns don’t get much traction anymore.

 

[mikegg]

True, and also, in general, NVidia recognizes that on client, their 2 chip (CPU, GPU) solution is doomed and will inevitably be surpassed by single chip APU.

It was doomed the moment Apple announced the M1.

It became more doomed when unified memory SoCs are way more economical for local AI inference.

Are discrete GPUs market going away? No. But its importance will continue to get greatly diminished.

 

[511]

This is AI inference mostly. Period. The biggest threat to Nvidia is Apple (and AMD) winning local AI inference due to superior SoCs.

You forget Intel their SoC are fine for this task for small LLM they just need to make a bigger fatter SoC.

 

[DavidC1]

1. Appease Trump

Nvidia said the decision was made more than a year ago, when Gelsinger was still CEO. Nothing to do with Trump.

Teams at Nvidia and Intel have been working in secret on jointly developed processors for a year — 'The Trump administration has no involvement in this partnership at all'

Targeting hundreds of millions of client PCs.

www.tomshardware.com

said Nvidia's Huang said, during the joint press conference with Nvidia on Thursday. "They would have been very supportive, of course. Today I had the opportunity to tell Secretary [of Commerce Howard] Lutnick and he was very excited and very supportive of seeing two American technology companies working together."

The work began around a year ago, and preliminary agreements were reached by Intel's then-CEO Pat Gelsinger and Nvidia's Jensen Huang even before that. (A year ago, Joe Biden was president, though no one suggested his administration was involved, either.) Intel and Nvidia are working on custom data center CPUs that Nvidia will integrate into its AI platforms as well as GPU tiles that Intel will integrate into its upcoming client processors. In both cases CPUs and GPUs will use Nvidia's NVLink technology as an I/O interface. By now, there are three teams working together on the joint projects.

It was doomed the moment Apple announced the M1.

End of desktops and PCs were "foretold" for more almost 30 years now. First, Desktops and DIY are alive as ever, because if it wasn't gaming market wouldn't be thriving and dozens of players making everything including chairs(chairs!!!). If anything that will kill dGPU, it'll kill all the PC market, including desktops AND notebooks, it's the Smartphone market.

Even Gabe Newell said gaming was what drove PC purchases. If that goes away, expect Windows share to drop quick. Because when the passion and enthusiasm dies, the marketshare will follow.

 

[mikegg]

Nvidia said the decision was made more than a year ago, when Gelsinger was still CEO. Nothing to do with Trump.

Teams at Nvidia and Intel have been working in secret on jointly developed processors for a year — 'The Trump administration has no involvement in this partnership at all'

Targeting hundreds of millions of client PCs.

www.tomshardware.com

The $5b investment was likely to appease Trump. Maybe the technical partnership was not to appease Trump.

End of desktops and PCs were "foretold" for more almost 30 years now. First, Desktops and DIY are alive as ever, because if it wasn't gaming market wouldn't be thriving and dozens of players making everything including chairs(chairs!!!). If anything that will kill dGPU, it'll kill all the PC market, including desktops AND notebooks, it's the Smartphone market.

They're not as alive as ever. They're greatly diminished. Desktops and DIY used to be a way bigger market as a percentage of computers.

 

[Doug S]

The $5b investment was likely to appease Trump. Maybe the technical partnership was not to appease Trump.


They're not as alive as ever. They're greatly diminished. Desktops and DIY used to be a way bigger market as a percentage of computers.

Already most desktops ship with iGPU, and that's been true for a while.

This trend has little to do with gaming. It has been obvious ever since Intel started including a GPU in their chipset. As the number of transistors you can include on a chip increases, it was obvious the GPU would eventually move on chip, and would grow more powerful over time. This is why the AMD/ATI deal was good for both companies. First iGPUs killed the entry level dGPU market, then started slowly eating their way up through the midrange. Apple has shown Intel/AMD the path to finishing off the midrange and starting to nibble into the high end dGPU market.

Nvidia (at least those who are truly in "GPU" division and aren't doing Scrooge McDuck dives into piles of cash like the GPGPU/AI guys are) can read the writing on the wall. dGPUs will become ever more expensive the fewer customers there are to amortize the cost of developing the hardware and the drivers. It can't survive in a form where it is affordable for gamers if the only true GPUs Nvidia sells are at the very high end, the volume is too small. They'd end up pricing them at $10K for the CAD market. This deal gives Nvidia a way to preserve that midrange GPU market and maintaining volume by selling them as iGPUs packaged with Intel CPUs.

 

[adroc_thurston]

dGPUs will become ever more expensive the fewer customers there are to amortize the cost of developing the hardware and the drivers

ah no, there are other markets to amortize volumes with.

This deal gives Nvidia a way to preserve that midrange GPU market and maintaining volume by selling them as iGPUs packaged with Intel CPUs.

Wrong because that would be an expensive Halo product.
Anything but mainstream.

 

[Doug S]

Wrong because that would be an expensive Halo product.
Anything but mainstream.

You underestimate the marketing power of Nvidia's name in 2025. They have all the power Intel did in the heyday of "Intel Inside" without having to spend a dime on advertising or "market development funds".

Businesses will go apes--- over that, buying it just because they will figure that will somehow make the PCs they're buying more future proof, and pay for the added cost by figuring on adding another year to the replacement cycle.

You're assuming that because AMD's Halo stuff has been a market failure that this will be too, but that failed because AMD isn't Nvidia and even in Intel's depleted state to the people making these purchasing decisions AMD isn't Intel, either.

 

[adroc_thurston]

You underestimate the marketing power of Nvidia's name in 2025. They have all the power Intel did in the heyday of "Intel Inside" without having to spend a dime on advertising or "market development funds".

It's not about marketing or whatever, fat APUs are just that.

You're assuming that because AMD's Halo stuff has been a market failure that this will be too

a) not a failure
b) big iGFX is inherently niche because it offers a lot less config/pop flexibility than discrete graphics

 

[coercitiv]

You underestimate the marketing power of Nvidia's name in 2025. They have all the power Intel did in the heyday of "Intel Inside" without having to spend a dime on advertising or "market development funds".

Is your stance that Nvidia does not spend money on advertising and market development? Everything we know about them points to massive spending for both advertising and market adoption.

 

[LightningZ71]

While NVidia spends a lot on marketing, as a percentage of revenue, it's not so much.