NVIDIA Teases Possibility Of Their Future GPUs Being Made By Intel

NVIDIA has mentioned that it is "open" to the possibility of using Intel's fabs to make their future GPUs.

This was made clear by company CEO Jensen Huang, who commented with regards to discussions of Team Blue's foundries, PCGamer reports. Huang was quoted as saying that the company is "open-minded" to considering having Intel build their chips. He also mentioned that he is "delighted" with what Team Blue has been making in recent times.

It is also worth noting that these statements were made right ahead of Intel's imminent entry into the discrete GPU market: a market that NVIDIA co-dominates with AMD. What could be even more telling is that Intel CEO Pat Gelsinger reciprocated, saying that the company is "thrilled" for Team Green's interest in their manufacturing capabilities (via Reuters).

But while this might sound a bit off-kilter to some, Intel's foundries have been increasingly becoming more advanced. Not to mention, they're also still looking to build more fabs.

Recently, Team Blue is planning to build a new "mega-site" in Germany, while also looking to expand their current fab in Ireland. This is part of their recently announced €33 billion ($36.2 billion USD) investment to improve R&D and manufacturing services in the EU, as mentioned in a news release on the Intel website.

As such, it makes sense that NVIDIA would want to outsource either part or all of their chipmaking requirements to Intel, instead of sticking to TSMC who's also making chips for long-time competitor, AMD. They're also in partnership with Samsung.

Despite these comments, though, don't expect future NVIDIA graphics cards to come straight out of Team Blue's fabs. According to Huang, discussions for a potential deal will "take an extended period of time," as reported by Tom's Hardware.

What's Next For Upcoming NVIDIA GPUs?

For now, there are a total of three upcoming NVIDIA graphics cards in the pipeline: Ada Lovelace, Hopper, and Blackwell in ascending order of release. And at least two of them will NOT be made using Intel's fabs.

Lovelace has been confirmed as based on TSMC's 5nm node. Late last year, WCCFTech reported that the company paid "big money" to get the Taiwanese chipmaker's 5nm wafers in preparation for the Lovelace GPUs (which by now should be the RTX 4000 series). There was no exact figure given for the money that NVIDIA spent, though.

Hopper is the other one which will be made by TSMC, but using their much more advanced 4nm node, writes PCWorld. Hopper has already made its debut in the H100 GPU, a data center chip which uses third-generation High-Bandwidth Memory (HBM3) and features an insane 80 million transistors. It also features PCIe Gen 5.0 speeds among other specs, prompting NVIDIA to say that it is their "largest generational leap ever."

Lastly, Blackwell remains in the dark as to what it will bring to the table (or which fab will make it). The only thing confirmed about it is that it will first introduce the GB100 and GB102 GPUs, writes WCCFTech.

This article is owned by Tech Times

Written by RJ Pierce