As a chip supplier, Nvidia will not refuse Intel to bring additional AI chips onto the market in the future.

Today, the majority of Nvidia chips are manufactured by TSMC, the world’s largest chip supplier. Samsung produces the previous generation of 8-nanometer graphics chips (RTX-30 series), but was unable to impress with its 5-nanometer process. The RTX 40 series and the H100 and H200 AI chips, which are highly sought after in data centers, roll off the TSMC production line.

In an interview during the UBS Global Technology Conference, Nvidia CFA Colette Kress said she would consider Intel as a chip supplier for the next generation of chips, which will likely be baked in 3 nanometers.

“I think there are a lot of good chip suppliers. TSMC is a great partner. As you know, we also use Samsung these days. Would we like a third? Sure, we would like a third. And then you have to find out what interest they have in services. Remember that there are others who could come to the U.S.,” Kress said.

“TSMC in the USA could also be an option for us. Not necessarily different, but again with regard to a different region. There’s nothing stopping us from potentially adding another chip maker.”

Nvidia cannot follow suit

Jensen Huang, CEO of Nvidia, announced earlier this year that it was considering Intel as a supplier. It seems clear that the brand is keeping all options open for the next generation of chips and wants to produce maximum quantities. Last quarter, Nvidia sold half a million H100 GPUs, reportedly costing €50,000 each.

Last month also saw the launch of the H200 Hopper chip with faster HBM3e memory. Every manufacturer is begging for chips from Nvidia. Dell previously announced that it would have to wait at least 39 weeks for delivery of a new AI server today because Nvidia cannot deliver chips fast enough.

In the meantime, AMD can benefit from this. As a counterpart to Nvidia’s AI chips, the company launched the MI300X and MI300A last week. The chips are available in mass production and can help manufacturers reduce their backlogs if customers also choose AMD and therefore have to wait less.