Intel and NVIDIA: Important, But Not Fundamentally Reshaping Anything

NVIDIA GPUs were probably used somewhere in the chain for AI to generate this image

What WAS ANNOUNCED

  • Intel will build custom x86 CPUs for use with NVIDA AI infrastructure platforms via NVIDIA NVLink.

  • Intel will build x86 system-on-chips (SOCs) that integrate NVIDIA RTX GPU chiplets.

  • NVIDIA is buying $5 billion of Intel stock.

Press release: Intel and NVIDIA to Jointly Develop AI Infrastructure and Personal Computing Products

Analysis

The Intel NVIDIA combination gives x86 a healthy dose of confidence and should help Intel's foundry ambitions in the abstract, because it depends both on 14a being a success and on Intel's sophisticated packaging capabilities. It also gives Intel a new channel into AI datacenters that it badly needed. Meanwhile, NVIDIA gets more help in CPUs, an area where it has never fully invested, and further cements NVLink as an industry standard. NVIDIA has probably generated some political good will as well. The deal also gave Intel's stock price a healthy 23% boost and, at least for now, has already proven to be an excellent use of $5 billion of NVIDIA's extensive cash flow.

However, despite the hype, this deal doesn't reshape anything at a fundamental level. It isn't a tariff play or a national security investment. NVIDIA's silicon is still being fabbed at TSMC and then packaged by Intel; depending on how tariff rules and exemptions change in the coming years this could actually increase tariffs not reduce them. If it makes Intel more competitive that's certainly good for Intel's need for revenue bridging the gap until the Foundry business is viable, but NVIDIA is not making any promises to use Intel Foundry. Nor can this investment go towards Foundray; NVIDIA is buying Intel's public stock, and while a higher stock price could give Intel room to raise more capital from investors, the $5 billion from NVIDIA does not go to Intel's bottom line.

The joint silicon should result in more potent competition in PCs and laptops against a surging AMD once it ships, but Intel CPUs and NVIDIA GPUs are already a common combination in PCs today. It may allow Intel to pull out of the high end of its Arc GPU investments if it wants to, but Xe will still be needed at the low end; the Intel/NVIDIA chiplet SoC will only make sense at the high end of the price curve.

In the datacenter, Intel could well displace some Arm-based MediaTek processors, but likely not all of them. NVIDIA and MediaTek are also building silicon for AI on the desktop, and are rumored to be working on Windows laptop chips together. I suspect that NVIDIA will continue building DIGITs on Arm, and, if the PC plans are real, they'll continue as planned as well. NVIDIA is not a small company limited to a handful of initiatives, and having multiple CPU architectures to pair with its GPUs makes strategic sense.

For Techsponential clients, a report is a springboard to personalized discussions and strategic advice. To discuss the implications of this report on your business, product, or investment strategies, contact Techsponential at avi@techsponential.com.

AI, SiliconAvi Greengart