Nvidia CEO: Intel's Next-Gen Process Has Delivered Promising Results

Intel famously lost Apple's business for the Mac when it fell behind TSMC while trying to move from 14nm to 10m. In response to this public humiliation, Intel launched an aggressive rebuilding strategy that involved attempting to leapfrog through five nodes in four years, which had never been done before. With that plan in place, Intel is now apparently talking to Nvidia about its need for future chips to lure it away from its biggest partner—TSMC. The notion that Intel could acquire such a scalp seemed ludicrous a year ago, but now it's a possibility based on comments made by Jensen Huang, Nvidia's CEO.

Huang held a Q&A with the press at Computex, and he received some questions about the company's reliance on Taiwan-based TSMC for the lion's share of its current and future GPU manufacturing. Putting all of one's silicon eggs in a Taiwanese basket is starting to look like risky behavior for geopolitical reasons, and Jensen assured the press it was looking at other options. "You know that we also manufacture with Samsung, and we're open to manufacturing with Intel," he told the assembled media. "Pat [Gelsinger] has said in the past that we're evaluating the process, and we recently received the test chip results of their next-generation process, and the results look good," he said, according to Tom's Hardware.

Intel's Foundry Services are prepping a diverse menu of technologies in an attempt to lure companies like Nvidia away from its biggest rivals. Credit: Intel

Jensen did not state what this mysterious next-generation process is, but the company is currently manufacturing Intel 4, formerly known as 7nm, for Meteor Lake. Ironically, Intel is also using TSMC for most of that product's tiles, but that's neither here nor there. Although 7nm is Intel's current bleeding edge process, TSMC was cranking out 7nm products four years ago and 5nm more than two years ago. So, as Jensen said, it's something "next-generation," which could be Intel 20a, 18a, or smaller.

Nvidia has previously stated it would be open to using Intel Foundry Services (IFS) for future products if they were competitive with TSMC's offerings. Post-pandemic, more companies are looking at diversifying their supply chains, and if Intel has the goods, such a partnership would make a lot of sense. However, Nvidia makes a broad range of chips these days for a wide range of applications spanning GeForce to automotive and the data center, so it's possible it could tap Intel for a particular piece of upcoming silicon while still using TSMC for its most ambitious products.

A possible arrangement between the two companies is even more intriguing given Intel's fledgling discrete GPU business, which currently uses silicon fabbed by TSMC. The company has stated it chose TSMC over its own fabs due to favorable pricing and volume, and that likely won't change for its next GPU architecture, code-named Battlemage.

Still, it sets up an interesting turf war for 2024 and beyond when Intel's next-gen nodes come online to face off against TSMC's next-generation processes. Intel has previously stated that its IDM 2.0 strategy would see it achieving market dominance by 2025, so we're getting closer to the endgame.