How big are your concerns that these barriers will encourage China to spin out competing AI chips?

There are things in China that are competitive.

Okay fine. It’s not data center scale, but the Huawei Mate 60 smartphone that came out last year got some attention for its native 7-nanometer chip.

Really, really good company. They are limited by whatever semiconductor processing technology they have, but they will still be able to put many of these chips together to build very large systems.

How concerned are you in general, though, that China will be able to match the US in generative AI?

This regulation will limit China’s ability to access the latest technology, which means that the Western world, countries that are not restricted by export controls, will have access to much better technology, which advances much faster. is growing So I think this limit puts a huge cost burden on China. You can always, technically, assemble more and more chipmaking systems to do the job. But this only increases their cost per unit. That’s probably the easiest way to think about it.

Does the fact that you are making compatible chips to continue selling in China affect your relationship? TSMCTaiwan’s semiconductor pride and joy?

The number one rule is specific. This is no different than the speed limit.

You’ve said many times that eight of the 35,000 components in your supercomputer are from TSMC. When I hear that, I think it must be a small part. Are you reducing your reliance on TSMC?

no not at all. Absolutely not.

So what point are you trying to make with this?

I’m just emphasizing that to build an AI supercomputer, there are many other components involved. In fact, almost the entire semiconductor industry partners with us in our AI supercomputers. We are already very close partners with Samsung, SK Hynix, Intel, AMD, Broadcom, Marvell, and so on. In our AI supercomputers, when we succeed, companies succeed with us, and that makes us happy.

How often do you talk to Morris Chang or Mark Liu at TSMC?

all the time. Constantly. Yes. Constantly.

How is your conversation?

These days we are talking about advanced packaging, capacity planning for years to come, for advanced computing capacity. CoWoS [TSMC’s proprietary method for cramming chip dies and memory modules into a single package] New factories, new manufacturing lines, new equipment are needed. So their support is really, really important.

I recently had a conversation with a generative-AI-focused CEO. I asked who Nvidia’s competitors might be down the road, and the person suggested Google’s TPU. Others mention AMD. I imagine it’s not such a binary for you, but who do you see as your biggest competitor? Who wakes you up at night?

Lauren, they all do. The TPU team is phenomenal. Most importantly, the TPU team is really great, the AWS Trainium team and the AWS Inferentia team are really exceptional, really great. Microsoft has an internal ASIC development that continues, called Maia. Every cloud service provider in China is making chips in-house, and then there are lots of startups making great chips, as well as existing semiconductor companies. Everyone is making chips.