The Six Five team discusses Intel’s DCAI Day.
If you are interested in watching the full episode you can check it out here.
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.
Transcript:
Daniel Newman: While MJ was talking to us at HP Amplify, SVP Sandra Rivera that leads the DCAI business, GM of that business, was talking to investors, and I actually was part of that day. I did a generative AI webinar that went out on the same day. I know it was where I talked to Hugging Face and some of Intel’s leadership on the future silicon that’s going to drive this generative AI movement.
But I think first of all, like we say, Pat, this is not investment advice, but this was an investor event. It really focused on a couple of things, in my opinion. One is reassuring and reiterating the opportunity for silicon in the data center and AI market in the larger TAM opportunity. And then of course, Intel really trying to reiterate how well it’s doing with its IDM 2.0 and its overall strategy and hitting its timelines.
Pat, I’ve been steadfast in saying that Intel’s saying a lot of the right things, however its ability to deliver has been suspect at times and that has caused a lot of derision and frustration among this investor community. And I think the company had a really marquis day. First of all, it focused in on what’s driving the TAM and the growth and the heterogeneous computing growth and of course overall, the sort of growth in AI network security, all these different workloads that are going to require more compute. We’ve talked a ton about this with generative AI, Pat. So we’re seeing a 40 billion or so TAM for all the chip sets in the data center that’s going to go something like two and a half times, almost 110 billion by 2027. So just that alone, huge TAM expansion. Even if Intel merely can protect its market share, that’s a ton of growth. And of course plenty of companies gunning after them, both accelerator companies entering the market with specialty chips and of course AMD and NVIDIA have been fierce competitors in both of those particular spaces.
The other big trend they really focused on is just the overall growth in cores. That’s expanding for Intel. It’s expanding not as fast as it’s going to for AMD, but that is a trend line that is going to expand the overall utilization. The company leaned into its diverse portfolio that they have in the business and wanted to let the market remember that hey, yes, X86 CPUs, but they’re more than that. Discreet GPUs, accelerators and connectivity technologies are both currently here and on the road map, and that of course is going to drive growth overall, margin expansion.
And I think an interesting opportunity for Intel to compete with a company like an NVIDIA that’s got 90% almost uncontested market share in the GPU space, there’s going to be questions coming up in the near future about pricing power. Could somebody come in and really disrupt with aggressive pricing? We’ve seen it a little bit with AMD on the gaming side, how much that kind of gaming leapfrog could go between the two companies. We haven’t really seen anyone put a lot of pressure on NVIDIA yet on the GPUs for the data center with the huge cost, huge volume and the sustainability issues that are going to be related to using the – they’re kind of hogs. At scale, there’s going to be an opportunity for some cost efficiency there.
And then of course looking at the road map, Pat, and I’m going to just start here a little bit with the road map and maybe let you talk about it too, we saw Sapphire Rapids get to scale. This was a big thing. It’s not done, there’s more designs coming around Sapphire Rapids. But where the company I think created some excitement with some of its announcements about being able to pull forward – so 5th gen, which is Emerald Rapids, I believe it’s the followed by Granite Rapids and then even Sierra Forest. These names, they just make my head spin. But the bottom line is the company is quickly moving through its paces, what was it, five and four or four and three? Which one was it, Pat?
Patrick Moorhead: Five and four. I get it wrong. I got it wrong about 27 times and I finally hit it.
Daniel Newman: I feel bad about it sometimes but then I remember that there’s absolutely no way I can remember all this stuff. But you’re basically seeing the company pushing through its paces, it’s being able to pull timeline forward, showing its evolution and putting it on a timeline. And right now it’s looking like between now and 2025, it’s going to hit multiple generations. It’s going to hit an E-core. So its efficiency cores are going to come into market. It’s got discreet GPUs, dedicated accelerators and advancements in its FPGAs. I mean if you look at that alone, and then of course you look at the accelerated compute opportunity, both its acceleration on software and then acceleration on silicon, the company showed a really promising day.
Now Pat, Intel has every wall that can be put up being put up in front of it by the investor community. This has not been an easy run for Pat, but I felt like this was a really interesting sort of watershed moment for the company where it finally got to come out and not just say, “Hey, we’ve got a plan.” But it was a little bit more like, “Hey, we’ve got a plan and we are hitting our goals. Actually, we’re not only hitting them, we’re beating them.” If nothing else, Pat, if they can push forward the goal to only land where they originally said, that would be a remarkable improvement from Intel.
Having said that, if they can actually hit this accelerated timeline and some of these augmented technologies that are, like I said, FPGAs, accelerators, discreet graphics that they’re doing, you can’t not be excited when you see a $70 billion TAM growth over five years and say this is very promising for the company. So I thought it was a great day for Sandra and her team. Execution, that will always be in my mind, at least until the company’s done it a few times with no misses, the big question mark. But there was a lot of promise in what she said. And this is one of those where you have to say congrats because you’ve had a lot of headwinds, but this sounds like a good moment for the company.
Patrick Moorhead: Yeah, there’s a couple things going on here. So the company showed up and said, “Hey, we want to talk about future generation of Intel Xeon.” I think everybody expected for there to be a slip, but surprisingly, and I think the stock had a bump on this, Sierra Forest was sooner than I think people had expected. And I think it was not just about Sierra Forest, but that it was the fact that Sierra Forest is the first CPU based on Intel 3. So it basically said the five nodes in four years is on track. So I think it was less about specifically Sierra four, it’s probably more about it being Intel 3.
And also very clear, I mean in writing, this is on schedule, first samples are already out the door. It was a big deal. I think if nothing else, when it comes to CPU core to CPU core, this is going to help Intel a lot. Should it help specifically in the arm space? Because Sierra Forest is targeted as E-cores versus P-core.
One thing that I wish the company would invest a little bit more into is really talking about when they’re talking about, let’s say these data center SOCs like a Sapphire Rapids, that there is actually discreet functionality on that SOC that is 100% dedicated to AI, and in this case, it’s what’s called AMX. But I still don’t think it’s getting through because I think people think that this is just a CPU, not a SOC with a bunch of accelerators on that. And that makes a big deal because when it comes to very heavily latency-related type of inference, you want these discreet blocks. And by the way, very similar to the discreet blocks that are on an H100 or an A100; they’re just smaller. So I wish the company would get into that a little bit more.
I did like the disclosure on software because at the end of the day, I mean, I 100% believe that both AMD and Intel can create a piece of hardware that is competitive with NVIDIA for training. But it’s not just about that, it’s about the lower level software and the programming model. NVIDIA had such a first mover advantage with CUDA, and OpenCL didn’t move quickly enough to fill the gap that everybody could have used so now everybody’s bifurcated doing their own thing. Intel is doing OpenAI, AMD is doing their own thing that leverages their FPGAs, their accelerators and their CPUs.
But I did like that Greg Lavender came out and talked about what is open in democratizing AI, and I do believe that people need to consider this. Do we want one vendor to own everything? What are the byproducts of that? So open programmability, choice and compatibility, and trust and inference at the edge. And Intel is really driving this notion of SYCL, which is open programmability. And I think the claim was that it could address, 90% of CUDA code is migrated automatically to SYCL, and I think that that is a very positive step. I wish AMD and Intel would work together on training to come up with their own programming model, maybe get…even companies like a Groq or a Tenstorrent or somebody looking at-
Daniel Newman: What about the Groq compiler, Pat? The compiler that could then take these large models and move them quickly, efficiently?
Patrick Moorhead: Totally. If what NVIDIA brought out at GTC wasn’t a call for some sort of cooperation collaboration I don’t know what is. So net-net with DCAI from Intel is just by virtue of the momentum and size and scale, I think they’re going to do pretty well. I mean they didn’t slip on anything. Zippo.
Now, granted Sapphire Rapids is arguably two years late, but there’s nothing Intel can do about that. The only thing that Pat and the team can do is to look to the future. There were even some decisions that were made even before Pat came on core architecture that there could be some challenges with down the line, but I’ve heard full steam ahead for the group. I didn’t hear of any slips. I didn’t hear anything that definitively told me that Intel is going to be challenging Intel in training. I saw a strategy that I understand, but I’m not going to say again that it is finished until we see some more evidence of what can happen in the future.
Author Information
Daniel is the CEO of The Futurum Group. Living his life at the intersection of people and technology, Daniel works with the world’s largest technology brands exploring Digital Transformation and how it is influencing the enterprise.
From the leading edge of AI to global technology policy, Daniel makes the connections between business, people and tech that are required for companies to benefit most from their technology investments. Daniel is a top 5 globally ranked industry analyst and his ideas are regularly cited or shared in television appearances by CNBC, Bloomberg, Wall Street Journal and hundreds of other sites around the world.
A 7x Best-Selling Author including his most recent book “Human/Machine.” Daniel is also a Forbes and MarketWatch (Dow Jones) contributor.
An MBA and Former Graduate Adjunct Faculty, Daniel is an Austin Texas transplant after 40 years in Chicago. His speaking takes him around the world each year as he shares his vision of the role technology will play in our future.
