March 16, 2023 update: Nvidia CEO Jensen Huang's interpretation of comments he made to CNBC reporter Katie Tarasov (which appeared to suggest that he was open to using Intel's Arizona plant for new GPU production) is incorrect. Intel informed us that.

When Juan was quoted as saying, "Oh, absolutely yes: "We will definitely use Arizona," he was referring to the advanced node facility TSMC is building in Arizona, not Intel's own fab in Arizona.

Original text Have you ever seen a man eat his own head?" probably not, but it is very likely that you will see something just as strange.

Nvidia CEO and leather jacket enthusiast Jensen Huang seems to confirm that the company plans to use Intel's new state-of-the-art chip production facility in Arizona (opens in new tab).

At this time, Nvidia has not elaborated on what specific chips it will use. Therefore, it could be something other than a GPU. However, since Nvidia is primarily a GPU company, an Intel-made Nvidia GPU became a very real possibility.

Of course, such oddities have happened before. Nemesises come together to share a practical cause. It wasn't that long ago that Intel teamed up with AMD to include Radeon Vega M graphics in Intel's mobile CPU packages (open in new tab).

This really didn't seem to catch on, and Intel has since embarked on its own graphics chips with the Arc series GPUs (opens in new tab). However, this is an example of how a competitive relationship can become a symbiotic one.

Manufacturing chips for customers is central to Intel's plan to reinvent itself not only as a fully integrated chip design and manufacturing powerhouse, but also as a foundry to rival the mighty TSMC.

As we recently reported, Intel has professed aggressive plans to roll out three or more new chip production nodes within a year (open in new tab).

Intel 20A will follow in early 2024 and Intel 18A in late 2024, at which point Intel claims to have the most advanced chip manufacturing technology.

For Nvidia, a cutting-edge technology to replace TSMC would be really convenient. In part, it helps keep TSMC honest in terms of price. Strategically, however, being a fabless company that relies on one manufacturing option must be a bit frightening. Especially if that option is on an island in the South China Sea, which is currently the subject of very tense geopolitics.

Indeed, Nvidia has hired Samsung for its Ampere generation of GPUs, including the RTX 3090 (opens in new tab). In other words, TSMC is not literally the only option. However, while Ampere was quite effective in terms of performance, it was a fairly inefficient architecture. Thanks in no small part to the characteristics of Samsung's 8nm node, most industry observers consider it far inferior to TSCM's 7nm technology at the time.

In any case, it is not certain that Nvidia GPUs will be made by Intel. Nor is there any guarantee that Intel will improve its manufacturing technology sufficiently that Nvidia will want to use it. But Intel-made Nvidia graphics seems to be exactly what they plan to do. And it will take some getting used to.