So I think Pinnacle shows me a lot of what that would look like.

So there's a couple of interesting points, right? One is like existing benchmarks that are out there that measure model capabilities or tokens per second throughput and all that.

They all have a place and it all depends on what your focus is as a practitioner.

Maybe you are measuring only silicon to silicon and you only care about that kind of maximum throughput.

But what we were targeting was more people who have to make financial decisions on what AI to go implement.

And so, you know, it is kind of easy to look at something and say, OK, well, I'm going to take a model with some intelligence grading that I like and then some infrastructure that has a tokens per second throughput that I like and see what we get.

But what we did was the pinnacle benchmark goes a step further and measures correctness of performance.

But but really, how do you answer that? I mean, the word correct is doing a lot of work here.

What was the incremental need to even do anything?

I think because I think the genesis of it was really that everything was evolving and we wanted something that we could have better visibility into the roadmap for and more control over, right? I think all those benchmarks that you mentioned, they all have their value, right? And they all have specific areas and things that they touch on that they do extremely well.

And that could be across, you know, consortiums or speed to data or whatever it happens to be.

But the gap that we were trying to solve was how do we make a benchmark that is applicable to enterprise? Like people who are actually deploying and paying for AI out of the market.

And these conversations came up all the time in our engagements with customers, right? Whether or not you're talking to Dell HP Lenovo or one of the CSPs, at the end of the day, the goal was always, how do we convince people or tell people about our advantages for whatever their workload is? Now it's gonna improve their business or improve their time to market or make their business more efficient and more cost effective.

And there was really nothing that did that effectively and on the trajectory in which we saw was coming, especially around agents.

So the idea behind the Signal 65 Pinnacle Benchmark is really to allow us to test three specific categories of things.

So yeah, you could try to build it, but talk a little bit about the, anyways, react to that, but maybe we close with the roadmap and what's next.

what are we expecting from Dell tonight? I mean, what have they got to say? Are we going to keep seeing AI demand offsetting the memory pressures?

Yeah, it does feel like that's going to be the case.

I think memory pricing and pressure has really been the bear case for a lot of this from the infrastructure names for a long time.

But I think if you really look at the demand curve that we continue to see increasing, the Nvidia earnings gave everybody a very good preview of what we could expect there.

I do think Dell is in a very good position to figure out a way to offset that.

They're the largest of those OEM system integrators.

in this space and they have a lot more pricing power than some of the smaller options.

Is it a bit of a similar story to what we heard with, I guess, Lenovo? Is it the fact that some of these companies, they have the AI server business that's different to an HPQ, which suffered last week on higher memory input costs that's saving the day here? How long do you think that can last?

How much do you anticipate the conversation, or at least the communication from Jensen Huang, to be centered around, [laughs] there's no doubt the analysts are gonna ask him about this, the circularity of financing and depreciation?

I, I, I think we'll hear a lot about it.

I think both of those topics are super interesting, right? The de- the depreciation or shelf life of these GPUs is a really interesting topic.

You know, they've talked a lot about in, in their financial dealings that suddenly, you know, these are gonna be six to nine-year lives for GPUs.

Uh, you know, in the previous way of the world and how this worked in the last couple of decades of us tracking it, GPUs were kind of outdated in two to three years.

And what they're saying, and what we're seeing even in our own discussions with, uh, NVIDIA customers, like the clouds and neo cloud providers, they'll s- they're still renting out these A100 GPUs that are probably five to six years old at this point.

Um, and so I think that will come up and, you know, can they, can they make that case again and make that case very strongly that these are still, uh, gonna be in high demand? And all the technology, Blackwell, Hopper, uh, all, all the GPUs they've released this, to this point, will still have that same shelf life.

Do you view these as potentially being a real threat to NVIDIA's dominance over the near term, the next two to three years? Or is the CUDA ecosystem just simply too difficult to displace?

So which signal then should investors be trusting here?

Look, i- i- it's my view that, um, as we move from training to inference, the demand for storage is, is going to only increase, right? The, the training was the area- era of HBM and kind of this need for super high performance, high bandwidth memory, and inference is the age of how do we increase context sizes? How do we increase KV cache? How do we give the consumers of this AI, you know, uh, uh, systems that have kind of longer term memory? And that's all gonna be based on, on flash and memory.

Uh, so to me, it's, it's a, a huge boom cycle coming up, still to come, for, for NAND and flash and companies like SanDisk.

Does that signal to you that management still believes that the stock is undervalued even after its remarkable run-up?

I mean, that, that would be the indication, right? I don't think as a company that you would, you would propose a buyback of that scale if you didn't believe that there was more up, up, uh, uh, upward momentum to come.

It also tells you a little bit that they're trying to avoid some of the scares of the cyclical nature of the, of the memory cycle by over-investing in, say, new manufacturing or new facilities or new technologies too early, too quickly.

And this is maybe a little bit more safer bet, and it should, I think, calm a lot of the fears of, uh, of the investor community.

A- has this investment thesis fundaly- fundamentally changed with this most recent earnings and guidance from Micron?

I, I think it's fundamentally changed, uh, at least in the next kind of 24-month window.

I, I agree that the forward multiples that we're seeing on there don't take into account this, and they kind of are, are kind of looking back at a previous way that these memory cycles worked.

But I think a lot of what you're seeing in the pullbacks, uh, is probably also attributed to some of the, the unrest and questions around this new Korean investment, uh, where they're gonna invest over $500 billion into new production facilities, and what does that do for this position that Micron, SK, Samsung have had in terms of severe, uh, um, limits of inventory, and will that get addressed with some of these investments?

Is this scarcity becoming a permanent fixture in this market?

I, I think it, HBM uses a lot more wafer capacity than standard DRAM uses.

Uh, I look at it as this is, this is about as permanent as you can look out into the future, '28, '29.

I think those investments we talked about earlier are in that timeframe of even coming online to having, to having any impact.

I mean, just walk us through how much we should be paying attention to that in this earnings.

I feel like they've built that into their projections and earnings over the last quarter or so.

They did pay out their penalty that they had to pay for the transgressions that occurred in the 2019 to 2022 realm.

And I think that's kind of behind them.

the the restrictions and export restrictions into china are a risk factor for them i'd probably put that as the largest risk factor and i think you know there's a possibility that some of this is a topic of conversation for the president while he's over there in china this week yeah

But I'm just wondering what you make of the entire trade right now, given this rally.

I was looking this up today and Applied Materials is still kind of below forward earnings ratios than their competitors like Lamb Research and a couple of others.

They're in line with kind of the ASML, TSMCs of the world, which I think makes sense because they're one of the primary you know, kind of competitors, augmenters, suppliers to a lot of those organizations.

What's going to be a detail that might really make a difference?

would look towards indications of kind of what the sell-through is going to look like for ruben right in the next generation of product i think you're right the narrative about blackwell being sold out is kind of something we've heard multiple times now and i think for the market to really get some kind of reaction we've got to know what is coming after that and what the impact of that product is going to be on the market so indications about sell-through sell-out you know, kind of quantity and demand from some of the cloud providers on Rubin is going to be the key next thing to keep an eye out for.

I mean, what areas may be more lucrative than others, Ryan?

I don't think there's any doubt that data center and AI is gonna be the most lucrative highest margin area for Nvidia to sell into.

And that's what you're seeing play out inside the company itself is the kind of the gaming division gets pushed back into the pipeline a little bit in favor of that.

I think robotics at an industrial level is gonna be the next player in that space.

Automotive still has some pricing pressure in it because it is a consumer sold good at the end of the day, whereas industrial robotics, you have a much, much larger entities to deal with in terms of pricing structure.

So that's how I still think it would play out.

What's giving you the confidence that the progress is real this time and that it's sustainable?

I think there's some truth to the idea that the PC segment is going to be on slow growth, but they still have confidence that as the AI push and for AI PC to be an ASP driver, that's something that Intel will benefit from.

I do think there's significantly more upside in the data center for every memory, set and GPU you sell, you're going to need a CPU of some kind.

Lots of those are x86.

Intel is still the market share leader in that space, so that gives them a lot of upside.

I think their announced product partnership with Nvidia creates a really interesting product roadmap and pipeline for them over the next 24 months or 24 months out that gives them sustained growth opportunities.

And I just think all of that is working in their favor now, while the foundry side of the business keeps them I'll say interesting in the market for their long-term impact and value on the ecosystem.

And Ryan, last question to you, you know, with this confidence and optimistic tone that they took on the call, do you think that with this report they've earned back some market credibility or is Intel still very much in this show me set up for now?