The Six Five team discusses the Intel AI Event.
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:
Patrick Moorhead: Dan, you attended onsite in New York City.
Daniel Newman: Got a great selfie, which that’s all I care about, right?
Patrick Moorhead: No, I know. I mean I’ve heard some people say that the best research is selfie research. So we did it.
Daniel Newman: Self-analysis.
Patrick Moorhead: Yeah, you can check out my runup. I had a conversation with Pat Gelsinger on the runup. You can read that article on Forbes and then I probably tweeted too much, but it was a very provocative event. And I haven’t done my summary, but let me kind of net out what they did. So they announced their first AIPC, which is the core ultra. They introduced fifth generation of Xeon, same power envelope, a lot more AI performance. And by the way, I feel like this fifth gen was the fourth gen that everybody wanted in the first place, but they had to get that fourth gen out. It was two years late, but very quickly following up with fifth gen.
So I think the AIPC, Dan is going to have three chapters in this year. First phase is going to be limited apps and spreading all the AI goodness across CPU/GPU in a smaller NPU. And then maybe toward the middle of the year we’re going to see these giant NPUs, more apps. The OS will likely take better advantage of it, easier for ISVs to target and characterize because it’s going to be a fixed amount of NPU tops. And that is dramatically easier to do than trying to size, literally in the Windows world, it could be 150 different configurations. Maybe there’s 1,500 different types of configurations that you need to fit in there. And then the third phase, which might be at the very end, might be into 2025 is where it’s all going to come together. So people really need to look at this as a continuum. Intel showed up with some pretty good tops across CPU, GPU and NPU. I think the NPU is like 11 or something like that. They had a couple videos of customers that were on stage.
And anyways, I think this is kind of the, I’ll call the beginning of shipping PCs for AI, and let’s move to fifth Gen Zon that San Rivera brought out. Big performance, big TCO claims, Signal 65 didn’t validate them yet. But those were some super big numbers and I liked the fact that Google Cloud, IBM and Zoom testimonials were super supportive. Final thing, Pat G. came out, snuck out a card of Goudy three and this essentially is going to do industrial grade data center AI training and inference.
It’s an ASIC, which means it’s not a GPU, which means it’s going to be higher efficient and higher performance for a set workload, but harder to program. And the program is built into a combination of the compiler and the API. But listen, it’s good to see Intel doing this. I’m going to hold my breath until 2025 until we see an AI-optimized GPU that people still think just by what they’re buying with Nvidia and now with AMD for the data center. So I don’t know. Intel is delivering on their promises. They’re not getting any credit for AI even though their stock has gone up 63% this year alone. But it’s good to see that the execution continues.
Daniel Newman: Yeah, I had a good chat with Pat Gelsinger yesterday, a small group of analysts, and one of the things I really like about Pat is that he’s confident in what he’s doing, but he’s also okay to admit where the company has gotten things wrong. And he talked about with the Alterra acquisition for a while because he’s obviously they’re spinning off the FPGA business and he basically said, “We got it all wrong.” And now we’re going to fix it and get it right. And the thing is is there’s an opportunity there. And while this has nothing to do with the event yesterday, but I just like a CEO that can say, “Here’s what we did wrong, and this is how we’re going to fix it.” Pat, I want to spend most of the time talking about the AIPC because we only have so much time, and I want to dive in a little bit and even have a bit more of a discussion on this with you.
But so the AIPC is probably one of the most hotly contested, exciting supercycle creators in the next 12 to 24 months. And yet it still seems that when you ask questions of executives about the AIPC, no one can really tell you what you’re going to be able to do with it yet. And I’m not saying running an LLM local, I get that, okay, everybody gets that, but how many times do you need to do that? And how many LLMs are you going to need to run at the same time? And how many of the people that are going to run four LLMs at one time to do some sort of crazy graphic rendering with content and context and versus the average knowledge worker that’s just inferencing on Salesforce or copilot on Microsoft. And so I know there’s this debate and of course in semis, Pat, by the way, this is why we built Signal 65 is because claims will need to be validated, but this first generation, Meteor Lake, AIPC, Pat Gelsinger. So he’s like, “Hey, you can go to B and H down the street today and buy this.”
Is that an AIPC or is it 10 top MPU with an AIPC? Does it need to be over 25? Is it 50? When does it become an AIPC? And what does this market outlook really look like and what is the compelling killer use case that is going to drive enterprises and then consumers in droves to the store to buy an AIPC? I get that AI is cool, I get that inferencing more locally is important. But I mean, by the way, this was a question that was being asked in the audience. Gelsinger gave a good example when he talked about a Zoom meeting where you could be literally real time transcribing and getting action items versus right now when you use a Zoom tool and it gives you the summary after, but there aren’t that many use cases yet that are really well understood.
And then so this point is this battle for tops. What are we winning as consumers? And how much more are we willing to pay to get from 10 to 20 and 20 to 40? I’m just asking the question right now because I also do believe that this is cool. So you have Apple and MD what, around 17 tops in what they’ve recently announced?
Patrick Moorhead: Yeah, I don’t have it on the tip of my tongue.
Daniel Newman: But I think that’s about what Apple announced in their M3 announcement, 30 minutes super announcement, which by the way totally fit my ADD, I think Lisa mentioned 17 on the new Ryzen. And obviously Intel’s first iteration’s at 10, right?
Patrick Moorhead: 11 NPU tops in total. Yeah.
Daniel Newman: Okay, 11. Why do I have 10 in my mind? Anyway, point is-
Patrick Moorhead: I’ve been saying 10 forever. I thought it was 10 and then I got the official word that it’s 11.
Daniel Newman: That’s a trillion extra operations per second. That does matter. You’ll have to talk to Shrout. By the way, we got to bring Shrout in on these conversations sometime. Ryan Shrout’s our president of Signal 65, which is our testing, validation, benchmarking services in case anybody needs that stuff because you do. Okay, so I’m going to pause here and kick this back to you. What I think companies need to do is actually tell us at 10 and at 20 and at 40, what does the experience look like? What changes? Because otherwise Intel’s in it, has a huge advantage because they have a product, it’s shipping in the market right now and they have the best distribution in the world for their silicon.
Patrick Moorhead: So the reason that everybody’s not talking about everything openly, quite frankly, is because they want to keep it secret about those future experiences. And that’s why ISVs and OSVs aren’t talking about it. And interestingly enough, they don’t want to strand and destroy the first half of sales. I mean it’s about as economic as that. And there’s still a lot of work to be done. The way that I like to explain this, Dan, is imagine doing everything that you can do today with the cloud inside a certain parameter model, but make it more performance. It’ll be faster because you’re not going up and down to the cloud. It’s going to be more secure. You’re not passing data up and down. It’s going to be a lot more private. When I think about, there’s an interesting company called Replay that operates on the Mac now it, operates in the public cloud, their cloud or whatever IS provider they use.
But essentially it’s taking snapshots of every single thing on the screen. And think about everything on your screen, every video, every phone call, everything that you really do on your PC and uploading that to the cloud, that’s going to be a hell of a lot of data. It’s going to be slow and there are some things we just don’t want to share. And if you’re an enterprise sharing all that data, I mean, I’d love to say that Microsoft 365, more companies are using that than standard, I’ll call it on-device, Microsoft 365 and Office 365. But it’s not. Big companies are still very reticent to share all that corporate data and send that to the cloud. And if I can do what I need to do with Microsoft 365 and not have to transmit that to Microsoft, I think customers are going to like that. And I think we already see elements of hybrid architecture with Office 365 where it actually determines what’s the best way to do this grammar check, what’s the best way to do this spell check based on the quality levers of that.
The industry’s calling that hybrid AI. It’s going to take a while to get there, but that’s where the ultimate happens, where you can have your cake, you eat it too, depending on your privacy settings, your security settings, and also connectivity. And quite frankly, compute. So long-winded answer to what you asked Dan, but I think that’s my answer right now.
Daniel Newman: Just because I seeded a little bit of my time. I’m going to make one comment and then we’ll get on. We’ve got a lot of ground left to cover, but I still think there’s Intel gains a marketable advantage by the industry holding back this as a secret because they’re getting distribution, they’re getting design, they’re getting into market, they’re basically getting to set the tone because the users will drive. And by the way, I still think there’s a big market here. Someone has to tell me why the number matters in terms of everyday use cases. Like I said, we’ve always understood workstations, I’ve always understood gaming, those kinds of things. People have always understood about the power and the performance requirements. But if I’m just a knowledge worker in Salesforce and Zoom and email and telling me why a doubling tops matters is going to be valuable to make sure that you can get people. Otherwise, it’s just iPhones, iPhone 15, iPhone 16, iPhone 17, and who cares? So anyways, all right, I seed my comments.
Patrick Moorhead: Yeah, not disagreeing. I completely agree. I just think people are keeping it secret right now and what we see with Meteor Lake today, I think it’s going to be important for people to look at the applications and how the experience is better. The reality is most of, I’ll call it if we’re using copilot, most of that is right now and even on Meteor Lake is going to be done through the cloud anyways.
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.
