The Six Five Pod | EP 290: Celebrating 100,000 Subscribers: AI Milestones, Custom Silicon Debates, and Apple’s AI Gamble

The handpicked topics for this week are:

1. Celebrating 100K Subscribers: Hosts open the special episode, celebrating 100,000 YouTube subscribers, thanking the audience and introducing the YouTube Creator Award. A montage of show highlights, including funny moments, diverse locations, shirtless episodes, and memorable guest appearances.
2. Apple, Google, and the AI Race: Pat and Dan transition into news analysis: Apple’s AI strategy, Gemini integration, CapEx, and the broader implications for device form factors
3. AI Chip Wars: OpenAI, Cerberus, Nvidia & Heterogeneous Computing: Hosts discuss major AI chip deals, the future of custom vs. merchant silicon, and why heterogeneous compute architectures matter.
4. Data Center Boom, Energy Constraints & U.S. vs. China: Exploring the exponential growth in data centers, energy supply/regulatory bottlenecks, and the U.S.-China competition on infrastructure.
5. Meta Layoffs, Wearables, and Future of XR: Meta’s Reality Labs layoffs and what it signals for the Metaverse, AI wearables, and the XR industry shift toward AI-powered augmentation.
6. China/PRC: Nvidia H200 Ban & Tech Sovereignty Rumors: Analysis on China’s restrictions on Nvidia H200 chips, sovereign innovation, and the “cat and mouse” of supply chains and government posturing.
7. The Flip - Live Debate Custom vs. Merchant Silicon, Google, Apple: A special, in-person, rapid-fire debate segment with spicy Texas sausage and coin flips: custom silicon’s rise, Google TPUs, Apple’s semiconductor strategy.
8. TSMC Earnings, AI Ecosystem, & Chip Market Trends: Macro discussion on TSMC’s results, CapEx, implications for Nvidia, AMD, Apple, Intel, and the ongoing AI-led semiconductor boom.
9. Infosys, GSIs, and the AI Implementation Curve: Hosts trade insights on Infosys’ strong quarter, what it means for enterprise digital transformation, and the role of GSIs as AI reshapes services.
10. 2026 Tech Predictions: Dan and Pat share predictions for enterprise AI, ROI, key AI milestones, and potential for AI-driven layoffs.

‍

For a deeper dive into each topic, please click on the links above. Be sure to subscribe to The Six Five Pod so you never miss an episode.

‍

Listen to the audio:

‍Disclaimer: The Six Five Pod 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.

Patrick Moorhead

Welcome to the six five weekly podcast. It is a special celebration. You can see here we are live in studio. I think one of the first times, maybe we've done a few before that, but we are here to celebrate 100,000 YouTube subscribers. We want to thank everybody out there for making this happen and this YouTube Creator award we are very proud of. So thanks for supporting all of us along the way.

‍

Daniel Newman

It's been quite a journey. I think the last time you and I did a regular Friday show in person together, we were crammed into, like, your little workspace office together. Exactly. Sitting shoulder to shoulder. This close? Um, it's been a journey. I mean, this is episode 290, and I think we cracked the 100,000, maybe ten episodes or so ago, but we finally could get off the road for enough time just before we get back on the road.

‍

Exactly. Once again, to be here in person. And it's you. All the all the fans, friends, followers, haters, all of you out there. I'm sure there's some of all of you watching the show right now that have made this happen. So thank you all so much.

‍

Patrick Moorhead

Yeah, what I'm particularly happy with is, is who watches the show. And I think we went out to a couple analytics companies and it's pretty much everybody who we wanted and we expected right from VCs, tech investors, enterprises and executives of of tech companies. And that's it. It's great. When a plan comes together, you put out the right content.

‍

They're not exactly the most interactive. They're kind of this lean back audience, as you totally would expect. How many times have you commented? How have you personally commented on YouTube?

‍

Daniel Newman

Not that often. And what I think gives me the best confidence in this audience, Pat, is when you and I have been on the road and people have randomly seen us at a at a, you know, restaurant in an airport, and they come up and they're like, oh my God, we watch your podcast. I, you know, there's been more than a handful of times where it's like, who are you?

‍

And it's like, great. It's so cool. Like people recognize you and then they'll often comment on your health journey, because somehow this has become also the Pat Moorhead Health bonanza.

‍

Patrick Moorhead

But I don't know, we pretty much share that at word. Hey. But let's get let's I want that that cold hard award in my hand. Let's go.

‍

Daniel Newman

I just want to know why you get to get to hold it. Oh, we both get.

‍

Patrick Moorhead

Both get one.

‍

Daniel Newman

I was expecting just you to get one. So we opened them up. Yeah. Together.

‍

Patrick Moorhead

Yeah. Let's do this time. Um, so I also want to thank the production and the editorial staff as well, because quite frankly, none of this could happen without them. And I feel like we really upped our game. Um, gosh, I don't know, a year and a half ago on production and got all serious about this. Check this out. Whoa.

‍

Daniel Newman

Thank you, YouTube, and thank you all.

‍

Patrick Moorhead

Appreciate.

‍

Patrick Moorhead

Look at this. Beautiful I love it.

Daniel Newman

We tap them high tap.

‍

Patrick Moorhead

Yeah.

‍

Daniel Newman

We did it.

‍

Patrick Moorhead

Great job bestie.

‍

Daniel Newman

Ready to go. And here's to hundreds of thousands more. Hey, we should ask the producers while we're here. When do we get the next ones? In a million. We gotta get to a million. Something like nothing between here and a million. Yeah. All right.

‍

Patrick Moorhead

All right. Well, hey, uh, our crack production team, the editors put together a montage historical view of the pod. Let's check it out.

‍

Patrick Moorhead

Well so funny. Like I remember you saying like you used to be so fat and I was like, you were never fat. And then I saw a picture of you where your face was like, round.

‍

[Montage]

‍

Voiceover:

Technology podcast. Patrick Moynihan How are you doing today, buddy? Daniel? How are you, my friend? Live for the first time ever for our regular show. Live in beautiful Maui. And here in your living room. Um, I'm in my backseat in my car. When you open, you look at the really show.

‍

That's the shirtless six five. You call the show this time? Thanks. Gosh, the screen wasn't big enough to see that pathetic effort after you mentioned my work at work. I don't know, I heard I heard the shirt, Rick, over your high intensity microphone. I walk around with you and it's like, oh, you got it. Let's go off the record.

‍

Let's bring on. Mr. Arms himself. Alex Katayama

‍

President and Chief Product Officer, Cisco, G2, Battelle Dan Ives how the heck are you? Remember that to be here. to go. Let's go to our next segment. Live

‍

With that, Daniel. You can try. I'm just protecting my insides. You want to hear my opinion?

‍

Hold up. Don't you love when we ramble on?

‍

Are you not entertained? This is a wrap. Thanks, everybody, for tuning in. We do appreciate it. Thank you for keeping up with the awesome growing community of us sponsors and listeners. Take care. Hello, guys.

[End Montage]

‍

Patrick Moorhead

So, Daniel, you know, a little bit of laughing. A little bit of crying. A little bit is sad. I mean, I was really fat and ugly for a long time, and I still am, and I just don't know it.

‍

Daniel Newman

And you're just not fat.

‍

Patrick Moorhead

know you're pretty sexy, and, I mean, you were always in shape. Uh, but no, seriously, those those were those.

‍

Daniel Newman

Well, we brought that element, like the first 4 or 5 years. We just. It wasn't like, this is something that sort of naturally came out of our show was, you know, especially because it became an interest of yours. You're it actually became almost an obsession of yours, which was something that was kind of just part of my life. And then you actually brought it back in me, and now it's become an obsession with me. It was like the other night I sent my wife an

‍

Daniel Newman

Instagram reel, and it was something along the lines of, you know, I may have protected my piece just a little too much because now all I do is basically work out, work, and send you memes. But like, it's become, you know, the obsession or the peptide show or the gun show. But I still think we do a pretty good, you know, pretty damn good job talking about what's going on in tech, too.

‍

Patrick Moorhead

Yeah. I mean, in the end, it's it is about the content is about the, uh, somewhat no absolutely informed opinions.

‍

Daniel Newman

But at least one of us.

‍

Patrick Moorhead

Yeah, exactly.

‍

Daniel Newman

I won't say which one.

‍

Patrick Moorhead

No, but seriously, we we this is a tech podcast show, and let's jump into the real content. So let's kick off with the decode. We're essentially, uh, separating the signal from the noise on the top news that was out there for the week. Let's dive in.

‍

Daniel Newman

Let's do it.

‍

Patrick Moorhead

All right. All right. Daniel probably the top news story here. You and I have been hemming and hawing about, you know, Apple's inability to do much with AI. And they cut a deal. Is it capitulation or just pragmatic or both?

‍

Daniel Newman

Oh, you're talking about Apple and Google.

‍

Patrick Moorhead

Exactly.

‍

Daniel Newman

Well I thought Apple I thought.

‍

Patrick Moorhead

Did I mess it up? No, no you did.

‍

Daniel Newman

I'm just. No. Oh, okay. Do it again if you want or I'll just come out. Just let me start. You don't have to set it up. I can start over. Yeah. I mean, you're talking Apple, Google, Gemini, Apple. Once again, tapping Google to solve a problem that it can't seem to solve itself. Although this one, I'm sure that I'm not sure the terms are quite as good, because Apple has been paid a fortune by Google for prioritizing its search, but from what I understand, Apple's going to have to pay something to Gemini, although the terms are still somewhat murky for the ability to basically put Gemini inside of the platform.

‍

Now, a little bit confusing and maybe a bit of a knock for OpenAI because, you know, OpenAI was deeply integrated a year ago. That was going to be the focal point. I think this has a lot to do with one. Google's played a big game of catch up over the last year, I think. A year ago, it was clearly OpenAI had a better model.

‍

It was more popular, it was more interesting. Google. Everything has been doing full stack, it's infrastructure, all the things that can offer. Clearly playing a role in Apple making this call. But Pat, I don't know if regime change is what's necessary. You and I have talked about this endlessly, but what the heck is Apple doing?

‍

I mean, they're spending nothing on CapEx. I guess that's great if you're an investor and you want them to just buy stock back for the next hundred years, but isn't this risky? I mean, are we only an inflection on device type away from Apple relevance, falling off sharply without having anything to speak of in terms of its own AI capabilities?

‍

Patrick Moorhead

Yeah, I mean, potentially I think we all need though, a device that has a display. And then the question is, is it going to be a device like a smartphone or something? We we wear on our head and look at. So and I think we've seen the industry and what, you know, how far we have to go with actually wearing something on our face.

‍

But I think for, I mean, at least a decade, I can't imagine not having a smartphone type of, of display in front of us. I think that gives Apple an advantage. I think this is a pragmatic capitulation. They realized that their own Ajax models, we're not going to cut it, and they didn't have a future and they couldn't attract the right talent.

‍

I am very surprised that this they couldn't pull it off for on device models. I totally get they couldn't do cloud models or whatever. Apple private cloud model. I'm not surprised about that, but I am really surprised because those don't have to be the same models. Um, but in the end, I think the, you know, the market liked it and, uh, on both sides also shows just how dialed in Google is on, on on what they're doing.

‍

They seem to be firing on all cylinders.

‍

Daniel Newman

Do you think I mean, I mean, my read on it has been that Apple can play and come in late and do. Okay. Is this a signal that Apple's just given up? I mean they've had the revolving door of AI leaders. Like I said, regime change looks like Tim Cook's time at the helm. He's he's absolutely executed flawlessly in many ways against Steve Jobs vision.

‍

Right now we're two decades into it. At some point when did it become like, what's his vision? I mean, a failed car, um, a failed wearable, uh, headset. Right. I mean, maybe AirPods. I mean, these AirPods, kind of the lasting legacy. It's time for a visionary, probably to come in and figure out what's next, because, I mean, I don't know if it's going to be the Johnny Five and Sam Altman show, because Sam is going to be wrapped up in a lawsuit with Elon Musk for the next 200 years.

‍

Yeah, trying to take that company, you know, into to a public, uh, not only public but a for profit entity. But like, we know there's a lot of other form factors being worked on. But Apple's mode is so deep. People love the product. They're still very committed to it. I don't know if they care if Apple makes the model or not.

‍

I think that's why their response was positive because like in the end, you and I do it. I go to the cloud, I use a Gemini app, I use an open AI app. It's like as long as the connection is good, I don't care if it's on device and I don't think most people do. I think that's why the whole edge and on device thing is kind of failed is the network is good, latency is low.

‍

It gives me a pretty quick response. I'm able to put the bicep pictures in. It tells me it's 18in. It looks great. The cuts are the cuts are solid. I mean, I don't care, I don't know and I don't think other people do either.

‍

Patrick Moorhead

Yeah. I mean, what I know for certain is that Apple has a moat and its end users and the experience or the I call it the advantage of not wanting to make a change. And these people don't want to make a change. Not because they don't think better devices or other devices are better. They don't want to go through the hassle.

‍

Daniel Newman

I think the SaaS industry is hoping the same thing.

‍

Patrick Moorhead

Exactly. Hey, we've we've milked this one for all it's worth. Let's let's dive into. It's a new chip announcement with a, um, AI cloud provider. Its open AI partner partnering with Cerberus to do a $10 billion

‍

Patrick Moorhead

deal. And, you know, it's interesting. Everybody was pooping everything that wasn't a GPU, right? And then everybody was saying, the only thing that will survive, uh, will be, uh, TPU, uh, maybe titanium. And then if Microsoft decides it wants to come out with a killer version of Maya, that would be it, but everything else would be dead.

‍

And then Nvidia buys Grok $20 billion. Um, a lot more than I think anybody thought it would be. Um, congrats to the grog, folks. And I guess we were able to take advantage of that deal as well, so. But we're still here, which means we didn't, uh.

‍

Daniel Newman

We didn't have enough.

‍

Patrick Moorhead

Stock. It wasn't. It wasn't for money, that's for sure. But here we have Cerebus, which again, there are another I. I look at them architecturally as a cluster on a wafer. So it is an entire wafer. They tell they they sell an entire system here. But I think you really probably I think you nailed it with your with your tweet talking about, hey, if nothing else is an example, uh, that the, you know, we don't have nearly as much as compute as, as we need.

‍

Daniel Newman

I look at my tweets.

‍

Patrick Moorhead

I do it's the, the at least the fifth place I go. Uh, in the morning.

‍

Daniel Newman

Uh, five babies.

‍

Patrick Moorhead

Exactly. Um, but it does clearly show that a winner take all architecture like a GPU is is is not going to happen, right? Heterogeneous computing is the way to go. And you know, I will take a victory lap. I mean, probably for five years. I'm not saying you didn't, but I was very clear in saying heterogeneous is the way.

‍

Here's when it gets more interesting and it gets more interesting when efficiency, performance, when the gaps get narrowed. Because then it makes sense to take the risk to do the extra work you have to do in software and even in operations to to make this work.

‍

Daniel Newman

And the inference inflection is a real thing. Jensen acknowledged different architectures with the Grok deal, whether he wants to say that or not. His is the histrionics was, he would say they were mostly going to fail and they were at least a year or two behind. So he kind of did the Grok deal. So he got to have his foot in that door without necessarily saying what's going to be in that door.

‍

Um, but the other thing is, Pat, is I think for this next couple of years, and this dovetails nicely into the next topic, is so long as this infrastructure build out is exponential, it's all hands on deck. Everything that can be built will be sold. So whether it's Qualcomm's inference chips in the Middle East, whether it's ARM building a competitive product for Stargate, whether it's Microsoft building a product for itself or or, I think the zero sum stuff makes good headlines.

‍

It's clickbait. It's just not reality. And I still think in the end, though, when Nvidia gets most of the volume and it's going to be that way for a while. Yeah. And meanwhile, though, there's still like 100 or 200 billion worth of revenue to go get for everyone else. So,

‍

Daniel Newman

you know, it's it's fun and, you know, but Cerebus was kind of a blast from the past. Like I was like, okay, you know, but it just goes to show that Sam Altman and by the way, I don't know if we're covering this on the on the docket. We're not. So I'll just mention this, but like Meta mentioned, their whole like, Meta compute effort or Meta sort of is deferring away from LMS, acknowledging to some extent that they are becoming increasingly commodity, but they're investing massively to try to understand their compute requirements over the next decade, to be able to maintain market leadership, which basically means buying all the freaking compute they can get their hands on and and just figuring out how much they need, when they need it, where they need it, and open AI.

‍

That's the one thing Sam hasn't wavered from at all. I need more compute. And so he's got to deal with AMD. He's got to deal with with the survivors. He's got to deal with Nvidia. He's got his own chip. He's got to deal with Broadcom like everybody that can sell a chip to him. God just hopes that he figures out how to stay in business.

‍

And Musk doesn't sue him into the abyss because yeah I guess the good news is the demand will go somewhere else.

‍

Patrick Moorhead

Yeah. One of the big, uh, a lot of the headlines out there, in addition to, um, AI provider doing a chip deal is our data centers, and typically it can be Stargate data centers, Microsoft data centers, Meta data centers. There seem to be a ton of data center news. Last week, a $6 billion data center announced in Arkansas.

‍

And it's not just announcing a data center. You've got to come in and talk about how you're going to get the energy, how you're going to get the water, how you're going to deal with the the community. Yeah, right. Um, is this just the next wave of support for AI? Is it as simple as that? Is there something unique here that we might be missing?

‍

Daniel Newman

No, I think the story, you know, I think when we have Martin Casado come on the show, he really did give me at least some new fodder. You know, I've been sort of looking at constraints at every layer the compute constraint, the memory constraint, the concrete constraint, the talent constraint. He basically just said we have a regulatory constraint.

‍

So when you hear about data centers coming out of the ground, I think what's really important is that for us to basically make sure we do maintain our leadership. And I will say very clearly. US is ahead of China in innovation, especially at the silicon level. But US is far behind. China on energy and energy is if we can't energize the data centers, we can ship GPUs all over the place, and we can add capacity and we can cut deals with Taiwan and they can – Micron can build its new fabs. We have more HBM and all these things. But like we have to energize this stuff. So our Manhattan Project is basically building more energy capacity. And a real Manhattan Project is getting our government red tape and bullshit out of the way, where we can actually start a project quickly, because at least when China sort of centralizes an idea on something and decides it wants to lead there, there's no red tape. They just build what? I think you've said it a coal fired plant a week. I mean, they are literally putting ten.

‍

Patrick Moorhead

China did ten nukes in the same time period we did one. Right? In the South.

‍

Daniel Newman

Anything they put their mind to. I mean, you know, I had a great I was on Fox business and one of the contributors on the way when we were talking about this, and he said to me, he said, you know how long the Empire State Building took to build?

‍

Patrick Moorhead

Less than a year.

‍

Daniel Newman

It was 13 months. Yeah. 13 months. Do you imagine what that would take today? What it would take in the middle of New York City to build a building like that? What? Five years? Eight years?

‍

Patrick Moorhead

Yeah. What I need to do more research on, Daniel is I was in the event where the president signed an executive order that cut through the red tape, uh, to build, to build these data centers. And potentially when we're in Davos, we can maybe figure out what happened or what's what's holding them.

‍

Daniel Newman

But one thing I'll just add here at Pat and, great, great point because they were supposed to be going faster. But what I'm hearing is it's just not going faster. And you got like the, um, you know, the different senators like Bernie, you know, wanting to slow it down. But Microsoft did come out with an effort that I do think is notable.

‍

And that was their whole effort to try to make sure costs aren't passed on to the citizens of various communities where data centers are being built. They're basically saying, we want to make sure we're paying for what we're creating the excess of normal capacity, load and requirement. I think all the hyperscalers need to come out and do that.

‍

I mean, affordability is going to be on the docket in 2026. And I think, you know, if Trump wants to see the Republican stay in office, you certainly have to talk about affordability because data centers and costs in those communities, it's meaning it's materially. Energy is the one thing that's been high.

‍

And why not? We're building gigawatts. Yeah.

‍

Patrick Moorhead

Yeah. What I don't understand in that whole thesis, though, is, is how you don't pass it on to to the consumer. I mean, maybe it's augmented, maybe they, they buy credits. It's a mystery.

‍

Daniel Newman

They put a plan and they outlined what it is, but it's not totally clear how they make sure that the excess doesn't end up landing inside the bills of the community. But yeah, I think someone's gonna have to answer to that because the costs are materially higher.

‍

Patrick Moorhead

Yeah. Final comment. Uh, the report came out that says there's just not enough jobs to actually build these data centers and there's not.

‍

Daniel Newman

There’s not enough people?

‍

Patrick Moorhead

Sorry, not enough people to fill the jobs to build all of these data.

‍

Daniel Newman

Temporarily unemployment should drop.

‍

Patrick Moorhead

Rates can measure that. Yeah. If we're actually measuring it, I don't trust any of the numbers and I haven't for a long time. But hey, let's move on here Daniel. Good. Good conversation here. Uh, Meta confirmed more layoffs from Reality Labs. And of course, the focus shifts to AI and also wearables. And you might be like, well, wait a second.

‍

What? Reality labs, what's the difference between that and these wearables? Essentially, Reality Labs is the content and the software to make that happen. I think the people who took the hit were on the business side. Um, but I think the press made a huge deal about it. It was 10% cut, not like a 50% cut, a ten, a 10% cut.

‍

And I'm pretty sure that they over hired, uh, there. They hired way, way too many people. And I don't view this as stepping back, maybe stepping a little bit back from the content, but, um, um, not stepping away from the overall objective because you can't separate AI from the wearable, right? The wearable AI is what makes the wearable relevant, right?

‍

When you're looking at something, it automatically tells you what you're looking at. You ask it a question, you ask it to do something. Um, so I, I see it as a momentary speed bump as opposed to Meta, you know, basically jettisoning the effort.

‍

Daniel Newman

I think it's also doing so many things at one time, doing them well, and then really looking at what where to prioritize. It's a little bit like micron prioritizing data center versus prioritizing consumer right now. I mean right now Meta’s prioritizing its AI and superintelligence lab efforts and compute efforts over the Metaverse, which the Metaverse is like a sub product.

‍

It's like a spoke of the AI future part. I mean, when do you last where your Vision Pro put it on? Lately?

‍

Patrick Moorhead

Gosh, I lived in another condo um, the last time I put that thing on.

‍

Daniel Newman

I'm just saying, like.I just don't, you know, the Metaverse. I still see some practical uses, but I just think that was the wrong pivot. And I think there's still a value in maybe that eventually AI plus is going to create the desire for us to live in another universe. But I think it's going to be more about XR and kind of this augmentation of our physical reality and us being out there and that sort of those bigger, bulkier headsets. I just don't I just don't know. I just, I don't know, I never I never bought any of them. I've tried them. I think they're only good for one thing.

‍

Patrick Moorhead

Yeah. Gaming.

‍

Daniel Newman

Exactly, exactly.

‍

Patrick Moorhead

All right. Hey, let's, uh, I'm going to skip a couple, uh, skip a couple here. I'm going to throw you over, uh, this H200 rumor game. I mean, how many times is this thing gone back and forth like a ping pong match? Uh, supposedly China's customs agents told Nvidia's H200. Chips are not permitted. It's just just another in a long line of negotiation. In a weird way?

‍

Daniel Newman

I don't know. How much. Do you trust the PRC?

‍

Patrick Moorhead

I mean, I don't trust any government.

‍

Daniel Newman

Yeah. But I certainly don't trust that one. So, you know, here here's what I think happens. I mean, there's a couple of different bifurcations here. First of all, I still think there's a large demand within Chinese research labs, enterprises and companies that prefer and would rather use Nvidia. So I think the demand is real.

‍

Right. I think China definitely wants a domestic strategy and has every interest in trying to build its own, its Manhattan project of building its own. It's project is lithography. Yeah. You know, they've cobbled together with old ASML parts and some some former engineers. And you know, they've cobbled together something that they say might be competitive with our current advanced lithography by the end of the decade.

‍

But of course, we're going to be multiple nodes into the future by that point. So I think those restrictions are really meaningful to them. I still think that Nvidia gets used in the cloud, Singapore, wherever it is. I don't think China is not using Nvidia in the cloud. Um, but I think Jinping is very prideful.

‍

And I think what it says about the fact that they can't get parity in their products and that they can do brute force. And you and I, I've talked about the The matrix networking and the fact that they can get somewhat close in some cases with brute forcing their their hardware. I think there's a reason to maybe push the envelope, but I think it sets them back.

‍

I genuinely do think that they probably would prefer to quietly acquire some H200’s to be able to continue to advance some of their interests, but they don't want it to look, This is a posturing thing. They don't want it to look like they're not capable of doing it on their own. And so there's this bit of a cat and mouse game being played. Pat, I don't know what what do you what's your read on that?

‍

Patrick Moorhead

So, um, this is the PRC, the government trying to motivate more innovation. And I feel like it's kind of a how you deal with your kids sometimes, which is you'll, um, if you give them everything that they want, they're not going to try to find unique ways around it.

‍

Daniel Newman

And you did that. And he turned out great.

‍

Patrick Moorhead

He did. Yeah. That's a that's a miracle. That's I'm very proud of my son. Uh, but I just think it's a, that it's the Chinese government saving face and motivating people to do something different. I mean, heck, two years ago the meme was “nobody could do AI in anything but a GPU,” right? And then the industry, the US industry got together and found very novel ways to do things with hard coded, hard coded silicon. So I, uh, I will leave that there.

‍

Daniel Newman

So are you ready to battle?

‍

Patrick Moorhead

I am, I am ready to battle. And we have a unique play here, the in Live version of The Flip, where we're actually going to, uh, look at The Flip in three different ways and we're going to react, but we're going to do a live Flip of the coin here and just want to remind the audience, uh, this is a I don't want to call it a fake debate simulated a simulated debate where we take very, uh, polar binary opposite points of view on a very nuanced topic. All right. Let's get into the flip.

‍

Daniel Newman

All right. But because we're in person, we're going to do this flip a little bit differently. All right. First of all normally we do one topic, but since we're going to be in person, we're going to do a little Texas theme here. We're going to add we're going to add a little spice. We're going to add a little sausage to the theme everybody likes.

‍

When we talk about the sausage, we're going to go from mild to wild. We're going to have three topics. I'm gonna have you flip a coin in a minute. When you flip a coin, it's going to be heads and you're going to be for each of these topics. If it's tails you're going to be against. And I have to be for all these topics and differently than our normal show that everybody out there enjoys so much.

‍

This isn't going to be like a six minute debate from each side. We're going to do like a minute each on each topic. I'm going to win each of them. Pat's going to go crawling into the darkness, and then we will get to our Bulls and Bears. But this is going to be a lot of fun. We're going to do a little different because why not? Because when in Rome or Austin or wherever the heck we are. Right, exactly. All right. Let's fucking go.

‍

Patrick Moorhead

I think this is chocolate. I might eat it afterwards.

‍

Daniel Newman

We'll see. You look like eating. I'm joking. Yeah. I'm sorry.

‍

Patrick Moorhead

I ate a few too many of these. Oh, what is it?

‍

Daniel Newman

It's heads.

‍

Patrick Moorhead

It's heads baby.

‍

Daniel Newman

Each thing. And so the Super Bowl here. That doesn't believe there's a bubble. It doesn't even believe that there's a possibility that one company is overvalued. I do think there could be one. Um, has to argue against this is gonna be hard for me. It'll be hard for me. All right, let's bring out the argue.

‍

Patrick Moorhead

Anything.

‍

Daniel Newman

Let's bring out the sausage first.

‍

Patrick Moorhead

Yeah. So the topic is essentially custom silicon will overtake merchant silicon AI within three years. And let's say it's units versus revenue.

‍

Daniel Newman

I'm going to bring out the mild.

‍

Patrick Moorhead

Let's go with the mild baby.

‍

Patrick Moorhead

So Daniel, this just pretty much nails it, right. We've got all of these cloud companies developing all of this silicon. I mean we've got OpenAI. We've got Google, we've got Microsoft, we've got AWS and even Apple here. And there's really nobody nobody left. Is anybody else relevant after that?

‍

Daniel Newman

Look, maybe 10 or 20% of the market we think that's true will go custom. But the big GPUs, the big rack scale systems, those are going to win. The entire software ecosystem has been optimized for the big merchants and especially for CUDA. So, look, you know, I believe that, uh, that, that there's going to be some custom chips, but it's not going to make a massive dent.

‍

Those headlines, every headline is, though. This is going to take down, uh, Nvidia, the only company that's got half $1 trillion of revenue for this year lined up in terms of bookings. Is Nvidia the entire rest of the market’s not that big if you add it all together.

‍

Patrick Moorhead

All right. Well, let's, uh, jack up the heat here. We're going to, uh, Defcon three here.

‍

Daniel Newman

I want the first one wasn't bad, but, you know, it wasn't spicy enough for me.

‍

Patrick Moorhead

All right, this is clearly TPU v7 and the numbers are there 4X the performance boost. Uh, Google TPU is not just not just for Google, it's now for Anthropic. And oh, by the way, OpenAI also signed a deal as well. So we're kind of going from abstract to very precise. And how can you possibly turn down a 4X improvement in performance?

‍

Daniel Newman

I mean, look, it's pretty simple. Google wants to make sure it secures its way for capacity from TSMC. Why not shill its its product onto a few other customers so they can get a little more demand, fund their efforts, get enough wafer capacity so that they can they can do what they do. It's good business. It's not going to take off at scale. But this again just backs my thesis of everything that can be built will be sold. I don't know if it has as much to do with the performance as it simply has to do with the fact that right now there's just tons of demand for AI chips.

‍

Daniel Newman

You want to get hot? 

‍

Patrick

Let's get hot.

‍

Daniel Newman

That one's actually pretty hot.

‍

Daniel Newman

The color of this thing itself has me a little bit alarmed about whether I should try it.

‍

Patrick Moorhead

For Google to do its own silicon for its own workloads. And then the next wave is Google doing its own silicon for other people's workloads, including OpenAI, including Anthropic, including its enterprise customers. But now you have the company who is absolutely dead last from the mega caps doing its own silicon for for itself. And to me, this just gets us to the point where we're kind of that third wave of customers. The people you would have never expected to do their own silicon would would go do that. And I think this is the the greatest example that custom silicon in three years from a unit volume basis will overtake merchant silicon within three years.

‍

Daniel Newman

Yeah, I think that's kind of a nice idea. But Apple you know, it's probably looking to take margin out. You know Tim Cook that's his best idea of his entire career has been to vertically integrate everything and take out margin. I mean who's going to need this chip for? They can't build any models. They really don't need it. I mean, they build all their own chips. It works great in their consumer and then in their device platforms. I think in their AI chip world, it's a dream and it's a dumb one. So I know they don't like Nvidia, but they should probably use it and maybe build a model that someone actually thinks is any good.

‍

Daniel Newman

It was a good conversation. I think they did it. And by the way, did this prove that we can actually have debates that are a lot shorter and we just like hearing ourselves talk, because I think we both got to the point a lot quicker.

‍

Patrick Moorhead

Either that or we like to eat.

‍

Daniel Newman

Yeah, that that red one though, it doesn’t look healthy. But the first one was really good. I'm still kind of gnawing on that.

‍

Patrick Moorhead

So by the way, now that we're having a real discussion right now, what's happening is at least we know Nvidia is sub segmenting its workloads. Right. Started with CPM which was for decode using Nvidia GPUs for prefilled.

‍

Daniel Newman

Yeah.

‍

Patrick Moorhead

So and then you've got this Grok thing for some very high speed, uh, low HBM type of um, type of scenario, super fast. Right. So, uh, they're reinventing themselves in order to get up against that. What we haven't heard yet is what AMD's going to do. Yeah, right. They very much have the capability to crank something out. They're really good at multi chip SoCs, but we haven't heard a thing out of them yet.

‍

Daniel Newman

How much do they want to potentially cannibalize Helios two. Because I mean this stuff does influence even if they are like optimized for different things. It in some cases, you know, especially like Nvidia. That's part of the reason I think Jensen ran away from it is like it does inference pretty well. Is it good to find that lower cost. But I think everyone's seeing the volume and they're realizing that like we do have to have different economics across the chain.

‍

Now, you know me. I mean, I've been probably one of the most public saying that Xbox are going to be very big. You know, we actually have XRP U growth crossing over GPU growth in 2027, in terms of growth of revenue in the in that in those businesses. Having said that, the growth is coming from a much smaller base.

‍

So it's, you know, the actual total size of market GPUs have come down. But by that point, they're going to be running, you know, that that whole and again, it's not even just GPUs anymore. It's really people have to understand that it's a system. At this point nobody's just buying a GPU anymore. Um, in most cases they're buying these rack scale clusters. They’re networking chips. There's CPU chips from ARM and other platforms. And of course, then there's the the accelerators themselves. So the mechanics of measuring this stuff has gotten very different than when we started, when we were just counting hoppers. Right. I mean, even with the was it the Vera NVL 72? Now we're even talking about two GPUs per package and they say 72. It's actually twice as many GPUs as Blackwell. And by the way, super confusing that that I've asked like eight questions about it. And I still don't have really great answers, but whatever.

‍

Patrick Moorhead

I think we both knew that this would happen. I mean, there would be different silicon doing different things. You hit the efficiency levels, the amount of power that need to take. And you're talking about tens of billions of dollars in data center. Something. Something had to break. And it did. Alright, Daniel, um, I appreciate you letting me beat you, but it is time to meet. It is time to go.

‍

Daniel Newman

Go ahead and send in the comments who you think won those debates? It's D A N I E L  N E W M A N In case anyone has spelling issues.

Patrick Moorhead
Are you sure it's not B I C E P S?

Daniel Newman
Oh, okay, I had to translate that.

‍

Patrick Moorhead

All right. Let's dive into the next segment. And that's Bulls and Bears. What you're basically talking about all things market that hit on the week. Let's dive in.

‍

Patrick Moorhead

Alright, Dan, because we love chip makers and talking about designers, we have to talk about TSMC here. What happened with their earnings? You know you think we have a lot of action here in the United States. Each US company like four times Taiwanese companies have to come out with a monthly estimates. So here we are. Was this a fourth. Was this a quarterly update? Was this a monthly update? Why should we care?

Daniel Newman

The least surprising result to start the next wave of tech earnings would be TSMC blowing it out of the water. And you hear a, you know, CEO and you hear their CFO come out. And actually, you know, this is a company that runs on being very secretive. They do not share much. They do not talk specifically about their customers.

‍

They don't talk about specific products that are being built. So you can't ask them about Blackwell's and Vera's, um, you know, Helios or any other systems or specific chips within those systems that are being built. But, um, they were very optimistic. You know, first of all, they raised CapEx substantially, meaning they're going to be building faster, building capacity faster.

‍

You know, they control the packaging right now. They are the king of chaos. Um, and right now, as we're building out this, this massive AI wave, uh, you're seeing N2 scale more quickly than N3 did, and you're seeing almost 80% now being on their advanced nodes. You're seeing, uh, what? Taiwan, USA made a deal $250 billion. It's government backed now. It wasn't TSMC. Taiwan made a deal that will be a large beneficiary to TSMC, because the Taiwanese government is going to back them, spending 250 billion to accelerate their roadmap in the USA because, you know, N3 is not here yet. And two. Isn't there only one N4 in the US? But their guide was a massive beat.

‍

Um, look for the bubble bears that are out there that think that AI is going to come to some screeching halt. What they heard in the earnings call from TSMC, as there were at least one quarter away from a bad quarter. This was not a bad quarter. And what this means for earnings overall is that Nvidia's going to have a good quarter. AMD is going to have a good quarter likely, but less certain, Apple is going to have a good quarter. Broadcom will have a good quarter. But Pat, the momentum is here because the whole story of the quarter gets told by what TSMC tells. And yes those monthly reports are are great. But this was the big quarterly real deal Holyfield report. And guess what. AI is still AI is still not being canceled.

‍

Patrick Moorhead

By the way, I agree with your macro thesis here, but I think we're looking at it the wrong way. Just because TSMC builds more chips doesn't always mean that equates all the way down the line to the very end. I think what it said to me. What was most interesting to me was not how they did in the quarter, but what they talked about CapEx.

‍

They said they were going to be rapidly increasing their CapEx, which, by the way, if you did the math of what all of the chip announcements, all the chip deals, that that was a foregone, uh foregone conclusion. The other interesting thing that I saw is they mentioned Intel, right? Which very rarely do they ever talk about their their competitors.

‍

They basically said they acknowledge them, they're not afraid of them. And just because people are plowing money into it doesn't mean they're going to be successful. And, you know, I was kind of thinking for them to even bring up Intel, uh, it must be on their radar.

‍

Daniel Newman

Yeah, Actually, that's a positive.

‍

Patrick Moorhead

You can't ignore that. And don't you know, in these these rumors about Apple not being able to get not being at the front of the line? Um that's evident. Nvidia becomes the new best customer. Um, by the way AMD is a top three customer. But people very rarely talk about it. And it's very simple. Uh, sure, they have GPUs, but they also have a lot of CPUs, a lot of client CPUs.

‍

They've got 25% market share in NPC's and close to 40% market share and overall servers, which the volumes are a lot higher. But anyways, those were some of the really interesting things that I saw. And by the way I'm not a bubble bear. I just still want to see some more, uh, downstream people buying. I know it's going to happen. I'm using it. I see it, uh, in my teeny tiny company that I have, uh, and what it's enabled to do. So it is coming. It's just enabling that for consumer and enterprises. And probably the longest term will be enterprise.

‍

Daniel Newman

Yeah I just look at what I forget the, the apps and the businesses and clearly the rerating of SaaS companies. I mean, you're seeing their forward revenue trading multiples cut in half. Even the ones that I think are going to survive have been completely rerated. But what AI has been able to do in coding, yeah, in a short period of time is going to rewrite the entire app and productivity ecosystem.

‍

Yeah, just use Anthropic Cowork. Just use it. Just use, um, Claude Code for a day and you'll just be absolutely blown away at what we can do and the speed of which we're moving. I agree, there's some digestion, some ingestion, some indigestion related to this stuff. But there is a reason these companies are spending.

‍

And the reason the bubble can't pop is as long as these data centers keep going up, as long as they keep being energized, as long as the GPUs keep getting racked in the big economy and the big macro moves very, very quickly. Yeah, The customer is consuming this stuff and spending money on it. There's clearly some unknowns there, but it's just so inconsequential at this point because those big dollars are so much bigger.

‍

And then again, before we even figure that out, the edge stuff's going to hit the humanoids, the physical AI stuff, the autonomy and the robots and the vehicles, which is just going to create another multi-trillion economy. But I think you're going to see it in healthcare. We're going to see it in Fin services. You're going to see it in manufacturing at such tremendous scale. The agents are coming. Yeah. My org chart is going to be like 93% agents and 7% humans by the end of next year.

‍

Patrick Moorhead

Yeah. But I generally agree with everything you said. I, I think it's going to be a choppy curve. It's going to kind of do go up like this. And I'm not saying that's not what you're saying, but that's that's my expectation. Yeah. Moving in. They're going to be fits and starts. And at some point the growth percentage has to decrease. The only question is when and where. I'm totally in on industrial and robotics. And we've got a unique set of challenges for those. And by the way, companies like companies like Qualcomm, who I think are very well positioned in that we agree on.

‍

Daniel Newman

We agree on that for sure.

‍

Patrick Moorhead

Yeah. Hey, let's wrap up. Bulls and bears here. Let's go to the opposite extreme of the AI curve and that is Infosys. Okay.

‍

Daniel Newman

Talk the implementers.

‍

Patrick Moorhead

Exactly. And if you're a large enterprise you're trying to get digital transformation done. Infosys also has an engineering services group. They made a couple acquisitions. So they're actually doing front end and back end ship work for companies with their designs. And they increased their forecast.

‍

And it was they had very strong demand coming from financial services companies, which they're typically tip of the spear when you want to think about enterprise in deploying a technology. So I think that is a good sign for the AI trade because quite frankly, the the GSIs are getting hammered, right. They're getting hammered on the meme that says AI is going to replace what they do. That's the first one. Uh, and the second one, um, is is they're showing up with lackluster earnings. So Infosys did well. And we'll see. Uh, we'll see what Accenture does and see if they can, uh, mimic that.

‍

Daniel Newman

Yeah. I mean, look these are the parts of the AI supply chain that I think are worth watching. A lot of these companies have taken big hits because the expectation is that maybe agents will do everything, but there is still work to be done to build. Yeah. It's just like developers aren't obsolete. They're becoming more engineers and they're becoming more workflow people. But it won't disappear itself. Uh, buddy, I think I think that's the docket I think.

‍

Patrick Moorhead

It is. And I think we're going to just end this just just with a quick, uh, what's your biggest 2026 prediction? Daniel? You probably made it. You probably said it, but can you share it with the audience?

‍

Daniel Newman

Uh, 18.5 inch biceps. Oh, you're talking about tech?

‍

Patrick Moorhead

Yes.

‍

Daniel Newman

Boring. Um. Yeah. I think. Enterprise. I think this is the year that Enterprise AI starts to show up. I think there's an accountability I think we're going to see. We're seeing it already super fast. Getting out the gate with with everything, with, uh, with cloud code and cloud co work that they've been the best at monetizing this.

‍

But look I mean we're at the point now where we need to it's a private moment. So in our little bubble debate that we just had the private moment is AI. We need to show that it can drive productivity and revenue into the average business. And I think that's when people will no longer be able to deny its importance, and it will start to justify all the investment.

‍

Patrick Moorhead

Yeah, mine is somewhat somewhat similar, but a little bit different in that, in that we're going to see, I like to say David Letterman, stupid pet tricks for anybody who remembers that – he brings somebody on and they would do something spectacular. We're going to have spectacular moments. Uh, like, uh, Claude just brought out with vibe coding an application that people seem to like.

‍

Daniel Newman

In ten weeks.

‍

Patrick Moorhead

Ten days. Right? And then we're going to see real ROI from those enterprises, whether it's connecting the front end to the back end. Um, and either either making more money for companies or saving more money. I do think that this is the year where we would start to see any layoffs as well, that are truly accompanying AI, not just, oh, we hired too many people during the pandemic.

‍

Daniel Newman

I guess we'll see. We'll have to come back to this.

‍

Patrick Moorhead

So anyways, thanks everybody for tuning in to this special edition here. 100,000 YouTube subscribers. We want to thank you for what you've done to get us here, and we hope you'll tune in. Subscribe. Tell everybody about the show. And for Pat and Dan, we are signing off. Take care.

‍

Daniel Newman

Bye bye.

‍