But Intel was doing all their discrete GPU chiplets at TSMC as well. As @Xebec suggests, they might keep their ARC series for on-chip iGPUs / APUs.
NVIDIA will invest $5 billion in Intel’s common stock!
- Thread starter Daniel Nenni
- Start date
I didn't know that. Do you have a link to the story?
Never mind. I found it. A bunch of companies have licenses. China is on really shaky ground with their monopoly accusations.
NVIDIA Unveils NVLink Fusion for Industry to Build Semi-Custom AI Infrastructure With NVIDIA Partner Ecosystem
NVIDIA today unveiled NVIDIA NVLink Fusion™ — new silicon that lets industries build semi-custom AI infrastructure with the vast ecosystem of partners building with NVIDIA NVLink™, the world’s most advanced and widely adopted computing fabric.
nvidianews.nvidia.com
I was wrong about this one - a friend reminded me that Hart-Rodino has a fairly low monetary trigger and does include a 30 day period.
The monetary trigger for filing under the Hart-Scott-Rodino (HSR) Act in 2025 is a transaction value that exceeds $126.4 million. If the value of the acquisition or investment meets or exceeds this threshold, a pre-merger notification must be filed with the Federal Trade Commission (FTC) and the Department of Justice (DOJ) for antitrust review before the transaction can be completed. The filing must be done before closing the transaction, and regulators have a waiting period (usually up to 30 days) to review.
MKWVentures
Moderator
@Daniel Nenni said Lip Bu is the deal maker. Tan is making changes. Intel using Nvidia graphics is great.
5B is a small investment (they are not a top 3 investor) but it depends on how the partnership goes.
Is it Intel/MS from the 80s?
or is it Itanic.
5B is a small investment (they are not a top 3 investor) but it depends on how the partnership goes.
Is it Intel/MS from the 80s?
or is it Itanic.
IMO, it is not at all like Itanic.
China isn't making a monopolist argument though. They are arguing that NVIDIA is being discriminatory - can't offer to Chinese companies. Non-discrimnatory licensing of Mellanox technology was presumably one of the approval terms for the merger. Of course the discriminatory part is stemming from US government restrictions, not monopolistic behavior, hence Chinese claims are really just part of the overall sprawling and sporadic trade negotiations.Never mind. I found it. A bunch of companies have licenses. China is on really shaky ground with their monopoly accusations.
NVIDIA Unveils NVLink Fusion for Industry to Build Semi-Custom AI Infrastructure With NVIDIA Partner Ecosystem
NVIDIA today unveiled NVIDIA NVLink Fusion™ — new silicon that lets industries build semi-custom AI infrastructure with the vast ecosystem of partners building with NVIDIA NVLink™, the world’s most advanced and widely adopted computing fabric.
Got it, but that makes no sense. Mellanox technology is entirely about Ethernet and InfiniBand, and both classes of products are available literally through retail channels and connect to servers via PCIe. The Mellanox (now Nvidia Networking) business unit doesn't license any of their technology, to my knowledge, outside of the RAND requirements for inclusion of their patented IP in the IB and Ethernet specs. The Chinese claims seem like BS to me.
MKWVentures
Moderator
Lets hope not. I had a close up view of that one!
Partnering the best run tech company of the last 20 years (IMHO) with Intel in any capacity is a win.
hist78
Well-known member
t’s a smart move for both Intel and Nvidia.
For Nvidia, amidst the “Made in America” political atmosphere, this $5 billion investment in Intel is a strategic decision. The most recent data shows Nvidia holds about $57 billion in cash on hand after already spending $24.7 billion on stock buybacks in the first half of 2025. If Intel makes a strong comeback, Nvidia will be rewarded nicely. If Intel struggles, Nvidia can say: “Don’t blame me, I did try to help with $5 billion!”
This approach is much simpler than committing $1–2 billion in purchase agreements with Intel Products or Intel Foundry.
On the other hand, this investment and collaboration increase the likelihood that Nvidia will place manufacturing orders with Intel Foundry. More importantly, it opens the door for jointly developed Intel–Nvidia products to one day be manufactured at TSMC, since both Intel and Nvidia are already major TSMC partners and customers.
As we know, Intel’s product division lacks highly demanded and competitive products. With this Intel–Nvidia collaboration, Intel’s product division could strengthen its ability to grow and compete independently, whether or not Intel Foundry succeeds, and without requiring massive Capex. A growing and self sustaining Intel product division is good for both Intel Foundry and TSMC. Without a strong Intel product division to fund Intel Foundry, no matter who manufactures Intel’s products, Intel Foundry will be in serious trouble.
Here are the current NVLink Fusion Program partners announced by Nvidia:
- Intel – Just announced.
- Qualcomm – Integrating NVLink into Arm-based server CPUs for AI data centers.
- Fujitsu – AI and high-performance computing (HPC) systems.
- MediaTek – AI and edge computing applications.
- Marvell – Developing custom silicon for data center connectivity.
- Alchip – NVLink integration for ASIC clients.
- Cadence – Providing design tools and IP for NVLink-enabled chips.
- Synopsys – Providing design tools and IP for NVLink-enabled chips.
Only one. Integrating NVLink will probably make any Intel datacenter GPU programs unviable. I can't believe Nvidia will license any other GPU maker, say AMD, to use NVLink.
Seems like this is Intel's path to stepping away from their loss-making discrete GPU architectures (Xe/ARC discrete) that had to be fabbed at TSMC and required huge software/driver efforts to remain in the game with NVIDIA and AMD. On-CPU ARC will probably stay.
Tom's Hardware sounds like they had a pre-brief - more details.
Nvidia and Intel announce jointly developed 'Intel x86 RTX SOCs' for PCs with Nvidia graphics, also custom Nvidia data center x86 processors — Nvidia buys $5 billion in Intel stock in seismic deal
Cats and Dogs, living together!
Q&A was the best part:
Jim Cramer from CNBC asked the first Q in the Q&A. Asking about the landscape of the partnership. He asked about Nvidia using Intel Foundry. Jensen deflected. Nvidia will be a large Intel customer and resell x86 chiplets.
The question about Nvidia using Intel Foundry was asked several times!
Jensen:
Trump had no involvement in this deal.
Sang the praises of TSMC when asked about Nvidia using Intel Foundry again.
$25-50B in annual partnership opportunity.
Still committed to Arm.
The investment in Intel is an incredible opportunity
The investment is linked to the product announcement with a big ROI expected
Intel has Foveros packaging that will connect Nvidia (TSMC) and Intel made chiplets (TSMC does not do that)
Jim Cramer from CNBC asked the first Q in the Q&A. Asking about the landscape of the partnership. He asked about Nvidia using Intel Foundry. Jensen deflected. Nvidia will be a large Intel customer and resell x86 chiplets.
The question about Nvidia using Intel Foundry was asked several times!
Jensen:
Trump had no involvement in this deal.
Sang the praises of TSMC when asked about Nvidia using Intel Foundry again.
$25-50B in annual partnership opportunity.
Still committed to Arm.
The investment in Intel is an incredible opportunity
The investment is linked to the product announcement with a big ROI expected
Intel has Foveros packaging that will connect Nvidia (TSMC) and Intel made chiplets (TSMC does not do that)
Hot mike at the start caught Jensen and Lip Bu talking about the amazing cognac they had last night !
It's a positive development--Craig Barret's plan someone mentioned above--but vague. I don't see this as an Intel Foundry endorsement. It may limit Intel graphics products, from integrated to discrete. It most likely will divide Intel Products development efforts between the JV and non-JV, which could mean less competitive Intel Products.
It could limit Nvidia too, since they want the best server processor to go with their graphics chips, and it is doubtful that is Intel anymore. At least it is possible AMD has pulled ahead decisively. This JV may increase Nvidia costs compared to go-it-alone.
Pretty explicit in press conference that this partnership is about end-product, not process or foundry relationship. Mostly about NVLink, and NVIDIA products that will be sold with X86 and Intel products that will be sold with NVIDA chiplet onboard.