TMTPOST -- Nvidia Corporation CEO Jensen Huang on Thursday denied the Trump administration’s involvement in a “historic” partnership with its rival Intel Corporation announced that day.
Credit：Intel
At a press conference along with Intel Chief Executive Li-Bu Tan, Huang told reporters the Trump administration has no involvement with the latest Nvidia-Intel partnership. Calling Tan a “longtime friend”, Huang said Nvidia and Intel held discussions for nearly a year before they reached the investment and technology collaboration. Thursday’s announcement is purely about custom chips the two companies will create, he said.
Nvidia and Intel have discussed potential central processing unit (CPU)partnership since at least the second half of last year, well before Tan was appointed as CEO in March, the Wall Street Journal reported. That discussion was also initiated long before Intel on August 22 confirmed it agreed to give a nearly 10% stake to the Trump administration in exchange for a total of $11.1 billion funding under the CHIPS and Science Act of 2022 (CHIPS Act).
Wolfe Research analyst Chris Caso in a note what’s unclear is if the new partnership represents “token cooperation intended for political purposes, or its if it’s the start of a wider collaboration that would more significantly benefit” Intel.
Nvidia on Thursday said it will invest $5 billion in Intel’s common stock at price of $23.28 apiece, representing a 6.5% discount to Intel shares close on Wednesday. The investment, subject to customary closing conditions including required regulatory approvals, will enable Nvidia to own a 4% stake in Intel.
Nvidia and Intel will also jointly develop multiple generations of custom data center and personal computer (PC) products, focusing on connecting both companies’ architectures using Nvidia NVLink, the high-speed interconnect for graphics processing unit (GPU) and CPU processors in accelerated systems.
For data centers, Intel will build Nvidia-custom x86 CPUs that Nvidia will integrate into its AI infrastructure platforms and offer to the market. For personal computing, Intel will build and deliver x86 system-on-chips (SOCs) that integrate Nvidia RTX GPU chiplets. These new x86 RTX SOCs will power a wide range of PCs that demand integration of world-class CPUs and GPUs.
Huang said both companies believe “it was going to be such an incredible investment.” Tan was excited about building a new era for both companies, claiming the collaboration will be good for the whole tech industry, and customers will also benefit from it.
Asked how this partnership will benefit Intel and Nvidia, Huang said it gives Nvidia an access to x86 architecture, thus offering the flexibility to scale its supercomputers further than it could on its own.
As to how the collaboration will work, Huang said Nvidia will soon support Intel’s CPUs in its NVLink racks for AI. CPUs are sold as separate servers that are then fused with Nvidia GPUs in its data centers. The company will buy the CPUs from Intel and connect them into super chips that then becomes its compute node, that then gets integrated into a rack scale AI supercomputer.
With the new collaboration, Nvidia will sell its GPU chiplets in a passthrough way with Intel or to Intel. Huang said this will expand the market for both companies.
The biggest question is if Nvidia will manufacture at Intel’s fabrications, Wolfe Research’s Chris Caso commented following the announcement of collaboration. Huang at the press conference decline to answer if Nvidia will use Intel’s foundry, stating the announcement is about CPUs, not exactly how these chips will be manufactures.
