You do need very high-performance storage because the GPUs that drive those algorithms need very, very fast parallel access to data. And with our unstructured data scale-out file system, we feel very well positioned for that. And then the third is with regard to inferencing, inferencing is the part of the data life cycle where you’ve taken a model and now you want to put it into production on our data set could be on a factory floor, could be in a distributed office. There it really depends on the data set and the use case what type of storage you need. You may need it for larger environments like shop floors, but you may not need super high-performance storage and compute for a very small office, like a claim’s office, for example, in insurance.
Aaron Rakers: That’s very helpful. And as a quick follow-up, real quickly, who is your predominant competitor that you see on the block optimized C-level all-flash arrays or C-Series arrays?
George Kurian: It’s all — the block market is a crowded market. We feel very good about our offerings in the mid-range, especially and the ability to offer a single solution with common automation, common administration, common life cycle management for both fire and block. And no other vendor in the market can do that. And so we feel very good. I think the large competitors are clearly Dell and HPE are in the mid-range and then you occasionally see some pure.
Aaron Rakers: Yes. Thank you.
Operator: Thank you. And our next question today comes from Samik Chatterjee with JPMorgan. Please go ahead.
Samik Chatterjee: Hi. Thanks for taking my questions. I guess if I start with the Public Cloud strategic review that you discussed, just wanted to clarify, based on the changes you’re making there, are there any cost implications other changes? I’m assuming that allows you to focus your go-to-market a bit more, but is there any sort of cost implication where you’re enabling some cost take out there? And secondly, Mike, you mentioned sort of the 58% to 60% gross margin on the product revenue in the second half. When we think about, sort of, prebuys that you might do going into fiscal ’25, how should we think about the trajectory of the gross margin into fiscal ’25. Any levers that you have to offset sort of the flow-through of that to the operating margin that you’re really reporting at a very strong level right now? Thank you.
George Kurian: On the cost side, listen, I think we will, for the most part, just repurpose those resources to drive growth in our first-party cloud storage, we feel good about the demand environment there, and we want to continue to accelerate that. Whatever cost opportunities they are, it’s been factored into the second half guidance that we gave you.
Mike Berry: Okay. And then thanks for the question, Samik. On the question, especially going into fiscal ’25 from the product margin, I’ll go over a little bit my answer to Meta when he asked the same question. So as we’ve all seen from the analyst reports, we have — we do expect that NAND is largely bottomed. We’re at the loss component pricing we’ve been at in a while. We did a really nice job and kudos to the supply chain team for doing the prebuys and purchase agreements for ’24. We will certainly look at those and continue to look at those in ’25, but it has to make sense for both NetApp and our suppliers. So as we go into ’21, there are certainly some levers. The big question there, Samik that I don’t have an answer to is what does the market do and pricing do over the next six months.
That’s why we’re going to wait, will guide the margin number when we guide for the full year because we really want to get through the next couple of quarters to see how component pricing goes, what we expect the mix to be. And then what happens from a pricing environment perspective.
Samik Chatterjee: Okay. Thank you.
George Kurian: On the cost side, listen, I think we will, for the most part, just repurpose those resources Thank you.
Operator: Thank you. And our next question comes from Ananda Baruah with Loop Capital. Please go ahead.
Ananda Baruah: Yes. Good afternoon. Thanks for taking the questions. I guess, George, an inference related question, maybe two quick parts. Are you hearing any of your customers talk about legislation around inferencing, having an impact on their plans? And I guess, just generally speaking, any opinion you have on sort of what that ramp over what time period kind of looks like even. If just anecdotally would be helpful. I appreciate it. Thanks.
George Kurian: Yes. I think, first of all, we are in the stages of generated AI. Predictive AI is quite mature and has very strong use cases. We have done really well in health care and life sciences in manufacturing, in parts of financial services, lots of use cases, right? And so I think that’s mature. It requires good data sets and good data management to make it have the right outcome. With Generative AI, there is obviously a lot of discussion on both regulation as well as judicious use of the technology, everything from fairness to epic to privacy to all kind cybersecurity, all of those things. And I think it will take time. Where we are with most clients today is proof of concepts, right? They are trying to put their data sets together, they are trying to learn what these models, will help them do.
