When we see the opportunity of — we are competing with others for the same opportunity, the opportunity is qualified. It’s got to be qualified for NVIDIA to give the supply or to suggest the supply will be available. So I think there’s a lot of control points in place that suggests there’s great fidelity in the demand signal that we’re seeing. But I can’t sit here and tell you with that 100% certainty that there isn’t 1 or 2, but I think it’s a really good demand signal. And given where our team is and what we see across the globe in enterprise with CSPs, we know where the opportunities are. I hope that helps, Sidney.
Rob Williams: Thanks, Sidney. Next question.
Operator: We’ll take our next question from Krish Sankar with TD Cowen.
Krish Sankar : Just had 1 for you. I’m kind of curious, you spoke about AI PC. What is your definition of an AI PC? And also, how much do you expect the ASP uplift to be for an AI PC? And how much incremental growth could it drive from the current $250 million unit run that you’re going through?
Jeff Clarke: Our definition of an AI PC is going to be a PC that has the performance capability to run the workload on the PC. Not a cloud service, the capability of running an AI workload in an app that has been AI-optimized and to be able to run it locally. The architectures that are coming all meet that criteria. There’s variations in performance, but they’re all meeting the threshold that we believe is required to call them PCs that they’ll be AI-enabled and AI PCs, if that’s a name. And I think we’re quite comfortable with that definition. Whether it drives tremendous growth to the TAM is to be seen. What I believe at is — you don’t want to be a PC user that doesn’t have an AI-enabled chip in it. You’re going to be next to people that have 1, and yours doesn’t, and it will perform differently.
It won’t be able to take advantage of some of the new exciting workloads, whether that’s new forms of search, new forms of security, new forms of interacting with your PC itself. The ability to put some form of assistant around you. Those are going to be huge productivity uplift, and I think it’s going to drive a refresh cycle. And again, the scale of this business is so large with 1.5 billion-plus units in the installed base, 950 million units sold over the last 3 years. This is a catalyst to increase the — or reduce the refresh cycle or increase the refresh rate, whatever you prefer. And I think that’s an exciting time, but I can’t call what that number is. Our estimate for the marketplace is low single-digit growth. I think I mentioned 3% to 4%.
That would take the market from roughly [250] to [260] next year. Let’s get there. Let’s get it growing again after 8 quarters of decline.
Rob Williams: Well, said. Next question.
Operator: We’ll take our next question from Aaron Rakers with Wells Fargo.
Aaron Rakers : There’s been a lot of questions already answered. But Jeff, I just want to ask you simplistically on the AI narrative. You mentioned 39 weeks of delivery time or lead times. Are you at all surprised that that’s not starting to change? Have you seen any indications that lead times have pulled back with some of the China restrictions put in place? And I guess that with that context also, would you expect diversity, i.e., another large GPU supplier to be a factor in unlocking and converting some of that pipeline as we move into next year?
Jeff Clarke: A couple of things. The pipeline is NVIDIA pipeline today. It’s not an alternative pipeline, it’s NVIDIA pipeline. And I wish I could tell you Aaron, that the backlog was less than — or excuse me, the lead time was less than 39 weeks. I can’t today. We are on the phone, working every available channel opportunity, as you might imagine, with our supply chain capabilities, to improve supply, to improve supply availability. We’ve offered our services. We’ll help where we can. I’m hopeful for the day to tell you that supply has improved greatly. Lead times have reduced, and we can work the backlog down faster. That’s our job. I don’t have those answers today. It’s 39 weeks. We’re trying to continue to get more supply.
That’s where we are. As we look forward into calendar year ’24, there’s clearly alternatives coming. There’s work to be done in those alternatives, software stacks have to be taken care of, resolve the opportunities around them. But there’s more options coming. Their adoption rate, we’ll see. But right now, that multibillion-dollar pipeline that I referenced, the backlog that we’ve talked about is NVIDIA-based 39-week lead time, we’re working our behinds off every day to get more supply.
Rob Williams: Next question?
Operator: We’ll take our next question from Samik Chatterjee with JPMorgan.
Samik Chatterjee : I guess sticking on the AI subject, you shipped over $0.5 billion of AI-optimized servers in the quarter. Just curious, with lead times holding where they are, how should we think about the trajectory of sort of what we expect for shipments in the coming quarters? Is it fairly stable in that sort of $0.5 billion range? Or should we expect some level of sort of ramp up if it’s supply easing? And then just a quick clarification, Jeff. You’ve mentioned a few times now, the improvement you’re seeing in traditional servers, even as you sort of called out that customers are pulling back spending in most recent months, you’ve seen some signs of that. Just curious, what’s the driver there? Like why what’s your insight in terms of gas customers are pulling back? Why is that sort of this green shoot and traditional servers that runs a bit counter to that overall theme?
Jeff Clarke: Well, we don’t forecast shipments in terms of specific dollars by product line for the next quarter. Our shipments are — for AI are implied in the guidance that Yvonne gave. I really don’t want to break it down into a specific of how many that will ship of AI service. But 39-week lead time, 9 months. Can kind of look at the math and go, maybe that’s probably the right sort of ZIP code to be in, but — we’re hand-in-mouth for parts. Traditional servers and why we think there’s a green shoot. I think it’s — the comments around — it’s been an 8-quarter digestion period, the longest that I can recollect in the server business. Data centers have aged product in. They worked through what they have bought in the early — the first half of the COVID era.
They now need to add capacity, and we’re seeing that in different areas. And to the point now, again, to be slightly repetitive, 2 quarters in a row of sequential growth, a building pipeline, which hasn’t happened this year. Improved — increased activity — improved conversion makes us still comfortable that something’s changed. Again, I’m not using that word recovery, but something’s changed. It’s an inflection point that Yvonne mentioned earlier, and we’re going to continue to obviously feed that inflection and continue to look at the opportunity.
