Ben Gilbert: preference

Transcript context

…Exactly. So kiki. The broader point where Moore’s law is going now, it is not just about the transistor. It’s really about the package. You’re seeing a lot more innovation, not just more transistors onto one gigantic chip to drive more performance. You can actually split up chips into multiple chips called chiplets which AMD is doing, and this is a big part of Intel’s future strategy, actually. Also, instead of having one gigantic GPU, you apply for Nvidia, you can have four smaller ships, and you can kind of stitch them together to drive more performance. That’s another way that we’re going to get a lot of benefit from Moore’s law. Actually, one thing I skipped over on the transistor side as we are moving to a new transistor architecture—either a two nanometer or three nanometer, depending on which company you’re talking about—to a gate all-around architecture. Previously, we are on FinFET, which has been around since I guess for the last seven or eight years, I don’t know the exact numbers, [01:18:41] 10. There’s line of sight into from here, more gate architectures, different materials, and then we get to the mid-2030s we’ll kind of see how it goes. It is amazing. I was actually at a virtual chip design conference earlier in the week, and there’s so much focus not just on this packaging idea but even if you have to extract one more layer from that, it’s about system-level performance. If you look at what Nvidia or AMD, and the leading digital companies are talking about, it’s can you make more processors and actually stitch them together into a cluster with some sort of proprietary interconnects. They’re really thinking more like computing companies than just chip companies now, and that’s a big change. But all of these transistor-level innovation, package-level innovation, and then system-level innovation, I think we’ve got a pretty good line of sight into the spirit of Moore’s law continuing at least through the mid-2030s. Let’s just talk about the iPhone because it’s the one I understand the best. Apple became the aggregator rather than the old days of I’m going to go build my computer, I’m going to go buy a motherboard, I’m going to buy a GPU and I’m going to slot it in the PCI slot, and blah-blah-blah. Apple basically says, well we’ve designed this logic board and we’ve designed most of the chips—the important chips—and we’ve situated them together. TSMC manufactures it. They do all the packaging, so you have the bare metal to bare metal packaging of these things so we don’t need to run it through these buses that have low bandwidth to get information from one piece to another. Let’s keep playing that out a little bit based on everything you know. Where in the value chain do you think the point of aggregation shifts to over time? Who gets to own where do we put all this stuff together? And I get to capture a lot of extra margin because I’m the one putting it all together. I think the interesting thing about the semiconductor ecosystem is actually there’s a lot of people capturing margin, and they’re captioning really high margin. This is the sign of a healthy ecosystem. There’s not one company that’s making all the money throughout the whole chain. We see margins come up. Who has higher operating margins—Texas Instruments or Microsoft? Because I asked it, you know the answer. It’s TI. It’s not really appreciated how good these businesses are.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.