The Six Five Pod | EP 279: Government Shutdown Blues: Tech Impact and Economic Uncertainty

On this episode of The Six Five Pod, hosts Patrick Moorhead and Daniel Newman discuss the tech news stories that made headlines this week. The handpicked topics for this week are:

1. AI Developments and Industry News: Recent AI deals going down, including Meta's $14 Billion infrastructure deal with CoreWeave and another deal between CoreWeave and  OpenAI totaling $6 Billion.

‍

2. Hyperscaler Strategies and Investments: Hosts analyze Meta's investments in AI infrastructure and capacity expansion and discuss hyperscalers' efforts to build their own AI chips versus using merchant solutions.

‍

3. Microsoft's Commercial Focus: Microsoft appoints Judson Altoff as commercial CEO to emphasize AI in business applications. This move is interpreted as Microsoft's strategy to lead in enterprise software transformation.

‍

4. Chip Manufacturing and Policy: A discussion on the feasibility of achieving 50% local chip manufacturing in the US and an analysis of the administration's efforts to increase domestic semiconductor production.

‍

5. Custom vs. Merchant AI Chips Debate: Patrick argues for going all-in on merchant infrastructure for time-to-market advantages. Daniel counters, advocating for hyperscalers to develop custom chips for long-term control and profitability.

‍

6. Economic and Political Landscape: A discussion on the recent US government shutdown and its immediate impacts. Hosts analyze the recent job market data, including its reliability and the potential effects of AI on employment.

‍

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.

‍

Or listen to the audio here:

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: About three or four years ago, I snuck a tracker in your backpack.

‍

Daniel Newman: You know what let me see. I did. I pulled it out just the other day.

‍

Patrick Moorhead: Found it. Oh, my God.

‍

Daniel Newman: Took it out. I knew air tags were good for something. Apple's biggest innovation is a stalker tag that they gave everybody in the last decade. So, yeah, I'm glad to hear they're finally going to cancel those. Those goggles. What a dumb thing. I know they're not actually canceling them, but, you know, maybe they can get around to being as cool as Zuckerberg.

Daniel Newman: So we did the full revamp of Pat Moorhead and the entire introductory part of that video. He is thin, svelte and muscular, but fat Dan still exists. I look at that now and I'm like, holy crap, I was such a pig. Anyway, it's 279, man. You know, it's funny, like, you're dragging through these. These decades of pods now, and it's like, 279, 278. It's like these numbers are really high, but it takes a long time. Like, you think of the work it's taken. You know, we started this thing in like, what, 2019.

‍

Patrick Moorhead: Yeah, a long time ago.

‍

Daniel Newman: Yeah, we've done this pretty consistently every week. Okay. We've done it pretty consistently every week for what, five years? Six years. Six years, and we're still only 279 freaking episodes in.

‍

Patrick Moorhead: I know. Well, I mean, if you add the. You add the. The ones that we do with guests, there's a lot more. I can't even imagine how much we've done.

‍

Daniel Newman: Yeah, that's crazy. But we off, we roll. We're crushing it. You're crushing it. I'm doing my best, hold myself together. It's an exciting day for me. You know why? Because about one o'clock, I'm gonna split out. I'm gonna drive up to Waco, Texas. I'm going to go spend the weekend today and tomorrow with my daughter at her sorority dad's weekend up at Baylor. Gonna be a good dad. I'm gonna go do a little pretend to hit a golf ball because I haven't done much of that. And then I'm gonna go and sit at a table with a bunch of people I probably don't want to talk to at some dinner. But then she and I are gonna go to the game. We're gonna wake up tomorrow morning, 5:30am we're gonna hit the gym, and there's a 9am or 10am Baylor game. I'm going to the football game. I'm gonna do this, dude, I'm gonna do this right? It's. It's good to be back. But before we jump into this show, we got a great docket this week. We always have a great docket, even when there's nothing to talk about. You and I are great at finding things to talk about. And thanks to AI, there's really not been a time. But how are you doing? Have I seen you this week? Have I run into you? Have we seen each other?

‍

Patrick Moorhead: Yeah. That was so weird. Daniel. Daniel and I, you know, we used to spend a whole lot more time together, particularly during, I guess, four or five years ago when we weren't able to travel. And you were actually coming down to your office down here. But it was fun. Yesterday I'm at this event at a hotel right next to my office, and I'm coming up the escalator, and I saw my bestie, Daniel, and had no idea we were going to be there. And we had a very. We had a very similar run last week when we showed up on the same plane, we're in the same city, and we didn't even plan that. I mean, Daniel, or are we just the beginning of the breakup? Dan?

‍

Daniel Newman: Yeah. I forgot to tell you the exact precise moment that I would be within a certain proximity range and your Dan-dar went off and you had to be near me, and so it drove a gravitational pull for you to come to an event. You didn't have a meeting with the CEO. It was really just about the fact that you sensed that I was near you and you needed to be near me. But it was pretty funny to come down and run into each other.

‍

Patrick Moorhead: I can't believe we don't have one. I mean, are we really good friends that we don't have? Fine friends. Are we connected?

‍

Daniel Newman: I don't even think I have that on. I'm paranoid about, you know, being tracked that closely.

‍

Patrick Moorhead: Yeah.

‍

Daniel Newman: But, you know, it's only because I'm sometimes in these bunkers making really awesome deals that I can't tell anybody about, but, man, I've got some really cool brewing, you know?

Patrick Moorhead: You don't even have to turn it on. About three or four years ago, I snuck a tracker in your backpack.

Daniel Newman: You know what I. Let me see. I did. I pulled it out just the other day.

‍

Patrick Moorhead: Found it. Oh, my God.

‍

Daniel Newman: Took it out. I knew air tags were good for something. Apple's biggest innovation is a stalker tag that they gave everybody in the last decade. So, yeah, I'm glad to hear they're finally going to cancel those. Those goggles. What a dumb thing. I know they're not actually canceling them, but, you know, maybe they can get around to being as cool as Zuckerberg. All right, man, listen, we've got a great show coming up. In surprise news, there were some AI Deals made this week. There's some potential foundry deals. Intel Stock is up 50% since DJT. President DJT went in. He's making the US some money. The US is making money. Open AI can't miss a week's Microsoft digging into commercials, going bigger, making investments, hiring more people, or maybe hiring less people, but promoting some people. You know, Secretary Lutnick, is he, Is he, Is he. Is he asking too much of US chip manufacturing? Is he slowing down deals in the Middle East? I don't know. There's all this going on and we got a couple other things we might get to depending on how time permits. And you know, the show, Pat, we break down the decode, we go into the flip, we simulate a debate which typically you win. And I put my tail between my legs and I crawl out of here sadly into my next meeting. And then we go into the markets and you know, that's where I like to shine. Anyway, I'm a stonk guy, as you may have texted me at some point this week. God, you love stocks. I do love stocks a lot with that, Pat.

‍

Patrick Moorhead: I mean, if I have to like to fall over one of your, one of your, your investment stocks, your posts, you know, the rocket ship to the moon market. I don't even know what market means, and I'm a freaking boomer, you know.

‍

Daniel Newman: Well, I mean, that's just me being all bold and making predictions, calling what I think's gonna happen market like market like, put it on your books, okay? Hold it, bookmark it, and then come back to me because all these, all my haters, which I have accumulated a nice number at this point, you know, when I'm actually wrong, at some point in the future, come back and show me where I was wrong. I mean, this is like the best ever diary of my opinions. And like, you want to.

‍

Daniel Newman: But what has happened over the last year is I've been right a lot. And so I'm asking people to mark those too, because I'm. I'd like them to do my victory laps. I gotta stop doing my own, you know, like let everyone else do our victory laps for us. But yeah, we got a lot going. Hey, why don't we dig in, Pat? Because I do want to get through this docket and there are a number of things that I think everybody out there is going to want to hear about. So let's decode. Okay, so you and I talking of markets and stocks, you know, I think you tweeted something basically like not the last customer I would expect, but close to the last customer you'd expect. And here I am all bullish, basically saying they're going to use Intel because they have to, because Intel's been made like Joe Pesci and Goodfellas right at the end. Is it going to be a bloodbath or is it going to be a heavenly partnership between AMD and Intel Foundry?

‍

Patrick Moorhead: Yeah, so these two are bitter rivals. In fact a lot of Intel's challenges, you know, were self-inflicted. But the other one is you've got AMD that's just crushing it on execution, in particular. And the company went from it not being able to execute with sus SUS architectures to absolutely crushing it left and right and being the number two, you know, a distant number two in the GPU market, but number two. So there's a lot of reasons, more reasons that AMD wouldn't do this even if Intel actually had it. I mean first of all there's not an independent board there to dish out the number of packages or the number of wafers. Okay. So you know, the thought of getting a rug pulled out there, meaning you know, Intel, you know, wants to reduce the amount of wafers or packages for AMD. That's a huge risk. Particularly that literally every CPU, every x86 CPU that AMD sells is one that Intel is not selling. So for a lot of reasons, there is a lot of hatred. I mean go back and look at the history. I was a customer of both Intel and AMD and I worked at AMD and there were a lot of really nasty things that went on between the two companies. So and there's a lot of history that, that, that, that goes into, into that. Now let me tell you where a lower risk deal could be. Okay? It could be on packaging, right. Where there is a commitment for packaging. And the other is this idea of an IO die. An IO die is typically it connects the chiplets and heck global foundries used to do the IO die for, for AMD does not require leading edge technology and I would say it's less strategic, it doesn't need to be on bleeding edge. And that could make sense to me. I do think that AMD is going to have to make some sort of move somehow to show that it's doing more on US soil aside from running more wafers through TSMC in Arizona. So yeah, kind of all both sides out of my mouth it on this one where you know, I think they would be One of the least likely. But if they do, here is where I would expect them to do it.

‍

Daniel Newman: Yeah, My thesis remains the same. It has very little to do with what AMD wants to do and it has everything to do with what's going to be expected. With Intel now being basically a nationalized foundry for the United States of America, I do understand that the investment is not necessarily coming with those particular strings, but there's pressure through direct and then there's pressure through indirect channels. And the indirect channels here are pretty palpable. The US Government did not make this investment in intel to have it be a, you know, a trailing, failing or unutilized resource. The expectation, and we'll talk about this later with Secretary Lutnick, with President Donald Trump, is that the US will return to some sort of manufacturing leadership. So this isn't an if thing for me, it's a when thing. And so I think when you hear that they're having a conversation, you know, this could be a conversation about something they'll do together in 2029. I mean, this is not necessarily anything to do. And people, it's like their brains explode about the idea that it's like, oh, they're having a conversation, they're going to announce something tomorrow. It's like, look for them to make these moves and to actually tape out for different foundries and plan to possibly use a multi foundry strategy. I heard one CTO recently say it's, you know, companies love talking about the multi supplier, multi foundry approach, but very few want to do it. And the reason is the risk. The risk, the cost, you know, the consistency. There's all kinds of reasons that they don't. But in the end, I think it's going to be policy driven. I don't think it's actually technologically driven. And I think policy is going to force and this is going to put some strain on companies because if intel cannot execute and does not have a parody parity in their product, they're going to force good companies to build inferior products. Kind of like, you know, when Apple was trying to put out two iPhones, one with an intel chip and one with a Qualcomm chip. And it literally became like a meaningful deprecation to get the intel version. You remember that that period of time we could end up in the same type of situation here where there's going to be AMD chips at Intel and AMD chips coming from tsmc. The other thing though is I do think any of the smaller supply, smaller vendors to tsmc, the Ones that have a small relative percentage of their yield are thinking multi foundry strategies or have to be because if you look at the demand around Nvidia, like I don't know if TSMC can build capacity fast enough to support what we have as these multi year forecasts every week, these deals, the amount of size, I mean I think it was yesterday I shared that One of the ODMs for Nvidia said that they have now visibility for demand through 2027. So two full years of sales forecast demand and they literally can't stand up factories fast enough to assemble these servers. And so all of this is going to take time. So if intel is going to have some capacity, I think some of these, some of these fabulous companies are going to look to take that advantage of that capacity because I think you've seen this before Pat, but Apple, Nvidia and then like Qualcomm and then it's like by size of how much capacity they're going to get. And so some of these ones that are a little smaller and AMD is not small by any means but it's a little smaller have to figure out how do they make sure that they never get constrained or they reduce the risk of being constrained. So those are my reasons. I think it's going to happen for all the obvious reasons that shouldn't happen. I also agree with those like me, I seriously doubt Lisa Sue's like oh my God, I can't wait to use Intel Foundry. I just think the pressure from the policy side is going to create demand. Whether or not any of these fabulous companies really want to partner with Intel Foundry. It's where I stand on it.

‍

Patrick Moorhead: Yeah, you brought up a good topic too that I put out a snarky tweet on. Shocking. And that's this, you know, oh my gosh, people are talking to other companies. Let me be very clear. Every single one of TSMC's largest customers has talked with Intel Foundry. Okay. And, and not because oh there's all this pressure but that's just the way that real business works and this is the way that, that, that, that this happens. Everybody wants to know what potential suppliers are doing and they can course correct and, and make changes. So, this whole idea, I mean and also you've got investment bankers and people leaking information because they want something to become true. And you know, maybe those longs on a certain stock will leak something or people will leak something on something they want to short so they can make more money. So just be very cautious out there, you know, about, you know, everything you're reading in the, out there, the.

Daniel Newman: Pumps and the dumps, I mean, listen, there's a lot of pumping, there's a lot of dumping. There's, you know, the, the longs and the shorts will go to a lot of lengths to create the kind of news that moves the markets. Right. But the bottom line, Pat, is this could be something five or 10 years in the future. I mean, that is the kind of thinking these companies have to be doing in a. In a cycle like this. So anyway, speaking of a cycle like this, you know, Meta has been busy, by the way, so is Core Weave. Meta announced a $14 billion infrastructure deal with Meta. And then of course, the week before, I'm pretty sure OpenAI announced a $6 billion deal with Coreweave. So there's, first of all, this capacity to play in my stock tweeting. You'll see me talk a lot about a couple of names that I've gotten to really like. Is this, like, Cypher Mining and all these mining companies turning into HPC companies. All of their capacity is being absorbed. So Corey is the same, except Corey doesn't actually have the energy or the real estate. So Core Weave is a. You know, they are. They're obviously building out the data center capacity for GPUs, the bare metal. But clearly all these hyperscalers are going to use these third parties and basically take advantage of any and all capacity they can build for them. So that's super interesting because Meta has also talked about their, what, 1.1 gigawatt D deal in. Was it in. In New Orleans? Not New Orleans, in Louisiana. Down there in the swamp. You know, they're investing hugely. Although it's been kind of funny, Pat, about Meta. Like, Meta went through this period where they were hiring people for 100 million. They were building a super intelligence lab. And over the last few weeks, whether it's been, you know, the SORA announcements, the open AI investments, the Nvidia deals, like, feels like Meta's gone a little quiet, you know, on some of their. What's going on with Llama? How competitive are their models? So they're spending, they're building, they're expanding. The one thing I always like about Meta is they have a real business with real profits that actually uses the technology. And that's what I keep going back to, is that is their strength. Their kind of open source superintelligence lab is a little more nascent to me. Like, how is it doing you spend a couple billion dollars hiring these researchers, you know, and by the way, not everything's going to happen in a week. Like there could be amazing things going on that we don't know about that are going to come out and completely, completely break our brains and break the market. So just keep that in mind. And then, you know, just another note is, you know, I think, I think we should debate this later. We'll probably have to debate something about this later. But you know, Meta, despite you know, adding capacity with core weave, which is likely meaning they're more Nvidia, using more Nvidia because that's kind of what Core Weave does is you know, they bought a startup Rivos to expand their efforts. You know, there are four generations in MTIA. They've got a lot of ambitions to build their own. We've heard about what is it like 6, 000 rack scale systems at their Broadcom partner infrastructure. But every one of these hyperscalers pad is making more and more investments to do some of their own. I mean, let me ask you this, like in terms of the semi parks, I know you love this even more than I do. Like they need to be, you know, they need making these investments. What do you think about this Ribbon Steel?

‍

Patrick Moorhead: Yeah, so I think we're gonna, yeah.

‍

Daniel Newman: Don'T, don't debate me in advance because we'll hit this on the flip. But just curious what you think.

Patrick Moorhead: First of all, just, just real quickly, you know the reason that all of these hyperscalers are, are going with people like Core weave is they have the space, they've got the power and they've got the GPUs. I, I don't understand the core weave long term business model doesn't seem very sticky to me. But if they can't make money on this deal, something like this with Meta, they're never going to make money and they're just going to wither away and die. But I think that's okay for people like Meta who may not need them in the future or for that matter, Azure. Yeah, and buying chip startup Revos. I did some digging a little bit on this one with some of my sources and it's not necessarily that Meta loves this architecture of Rebos. The claims to build a GPU like architecture that's CUDA compatible. I almost vomit in my mouth when I hear Cuda compatible. It's just nearly impossible for anything to be CUDA compatible. And if it were that easy, AMD would do that. I do think the architecture is interesting, though. It uses a bunch of RISC V cores that essentially, you know, it's not actually a GPU, but it has hardened ASICs, plus the CPU. Right. Which. Which seems to be one of the. One of the winning architectures for an accelerator.

‍

Patrick Moorhead: What I dug up is, this is an Aqua Hire. It's to accelerate MTIA, and they needed more horsepower to be able to do that. And, you know, Meta is just, you know, rocking on custom silicon. You just hear less about it. But they're on their fourth generation of MTIA. They're having challenges with them. With their bigger MTIA and, you know, to do. To do real training. The first trip was streaming. The second second chip, I believe, was suggestion engines. And Meta is just one big suggestion engine. Suggests ads, suggests friends, and I suggest content. So.

‍

Daniel Newman: Aqua Hire recommenders, buddy. Yeah, I recommend peptides. I recommend following on Twitter.

Patrick Moorhead: Yes.

‍

Daniel Newman: I recommend watching Patrick's weekly what's new in tech videos.

‍

Patrick Moorhead: Thanks, Bestie.

‍

Daniel Newman: I'm here for you.

‍

Patrick Moorhead: Honestly, you haven't watched a single one, and that's okay.

‍

Daniel Newman: No, I do whatever the recommender thinks. I want to see them. I don't, but I actually get all of them on every channel. By the way, every recommender believes. Pat, I want to watch your videos because I'm a super fan.

Patrick Moorhead: You sure that's not paid? Paid video?

‍

Daniel Newman: Are you doing paid? I don't know. Yeah.

‍

Patrick Moorhead: Are we unpaid?

‍

Daniel Newman: I mean, you got to do a little paid. I mean, I don't know. Some people, Some. Some, you know, low IQ people in the past have thought that somehow paid is. Is a version of cheating.

‍

Patrick Moorhead: Very low iq.

Daniel Newman: Next time a company promotes themselves at a Super Bowl, I'll be like, that's not real. It's not real. Doesn't count.

‍

Patrick Moorhead: Not good.

‍

Daniel Newman: Only organic.

‍

Patrick Moorhead: Yeah.

‍

Daniel Newman: By the way, you got a pretty big organic reach.

‍

Patrick Moorhead: Yeah. I mean, Twitter and LinkedIn. Even though as much as I despise LinkedIn, I mean, you're the same. What's so funny, though? I can get a banger on like 100, 200,000 people on LinkedIn, and I just have no idea where it's coming from. And the first thing that pops in my head is, oh, they, you know, somebody must have posted on a SharePoint site. But, you know, the biggest concentration for a post might be 10% from a certain company debunks that. And then some of the slop I'll put on. On X. Right. You know, I'll post the most thoughtful post and you'll get like 12 people who will watch it.

‍

Daniel Newman: You know, it's brutal. Like I said, I've given up on being thoughtful on X. It is a pure posting site and once you own that, you'll get way better results.

‍

Patrick Moorhead: Your follower count is impressive. I mean, they like the slope, they like the moon, they like the rocket ship.

‍

Daniel Newman: Yeah. You know what, the one thing is, because I actually do have outside of the companies we advise and I don't play with them as an investor, they love to know what you're doing. So I've started to share a lot more of this stuff so that I don't have any conflicts and that I can invest in and that I am investing in. And that's the stuff that everyone gets excited about because it's like, you know, it's one thing that you have an opinion on a company, it's another thing when you're actually putting your money where your mouth is. And so anyway, that, and you know, some of the random Nvidia posting because I just, you know, I continue to be the guy that just doesn't believe any of the nonsense, the China stuff, the, the, you know, all that. Like, I just, I, I, I might be wrong. I mean, there is a chance I will be wrong, but so far it's not been the case. And so we'll see where I land. By the way, everybody out there, if you are checking out gtc, Pat's going to be the super host this year of GTC DC. Sitting down with Mr. Gerstner. Very cool. Someone I got to meet at GCC last year. Very, very cool. I actually rode a little bus over to the convention center and it was just him and me.

‍

Patrick Moorhead: My gosh.

‍

Daniel Newman: Yeah, I was, I felt, you know, pretty starstruck, you know, meeting smart people. Like I told you, man, it's all about who you hang out with. It's all about you can't fly with the eagles when you're hanging with the turkeys. So I just try to get myself near smart people and then hope it's like osmosis.

‍

Patrick Moorhead: Yeah.

‍

Daniel Newman: All right, man, let's move on. Let's hit the next top OpenAI. Pat, have you created a video yet with yourself? High fiving me, buddy.

‍

Patrick Moorhead: I really am. I love new tech, I love the demos. But here has been the meme of the new video generators. It goes kind of like this. This is going to change everything. Okay. Oh my gosh. And then three weeks later, people forget about it and then nobody really Ends up, ends up using it maybe except for professionals or meme generators. Bigfoot videos on YouTube. Okay. I'm pretty sure, you know, I see some of the slopes that are getting recommended to me on YouTube and you know, I listen. It's not because I'm a boomer, actually I'm not. I'm Gen X. But when I see something that I know is not real and it's being passed off as real, I just, I reject, I reject that. But looking forward, right. It puts the, potentially puts the strain on companies like I would say you know, Adobe for small creators, right. Or they're trying to do videos. The bar for me and my workflow is going to be where it can replace you and I in this podcast. We throw the topics in there. It pulls in history, our snarkasm, our belief systems and then you and I can just go in and, and you know, we'll be sitting in the hot tub and our video will be, will.

‍

Daniel Newman: Be, will be generating yeah you know, these kinds of. Every week something new pops up. Sora, you know, I will have some fun with it. We will probably try to create some six, five, you know, Sora clips to start sporting with people. I do think, like I said, there's kind of like everything was kind of what I said in the AMD intel thing. There's, like two time horizons that people need to be able to parse through that. They're both like the near term. This is a gizmo, it's a gadget, it's a toy, it's a, it's nothing. It's something to make GIFs with. The future of GIFs are like Sora enabled with us in them. In the longer term, the fact that you could insert yourself into a video, write a script and create something pretty cool and it uses and trains on all your. Like that could end up being pretty awesome also by the way, it could be super nefarious. Like there's so many things like I was thinking about this, I was talking to the Wall Street Journal about this yesterday and I'm like, there's two great opportunities. It's like one is like you can basically absolutely create this amazing thing. And the other thing is people can create this absolutely very realistic looking stuff that puts us in compromising situations like the ultimate blackmail tool. But then in the future the person that's going to have that stuff happen to them, Pat is going to end up being like that's not real. That was AI. And then people that actually really do it are going to be like no, that was AI. Even though that was a real video of them, we are going to have such a problem. Like right now you can kind of watch these and be like well why are you growing a third arm? After you give the high five you see another arm popping out the back. But like at some point they will be good and it will be right and it'll be really hard to discern between what is real and what is not. And that I don't know, this might be the bit of Gen Xer in me but like it does scare me a little bit. Like I am a little nervous about how this could be used inappropriately, nefariously, negatively. Bad actors freaking, you know, used to create really looking images that can open your face, you know, unlock your face on your phone. I mean, you know.

‍

Patrick Moorhead: Well, you know I, what I've been pondering is will people put a premium on human content that they know is human? You know, kind of like a, you know, an RDA recommended daily, daily allowance type of thing. And I think like you know how there people will pay extra for handcrafted goods and handcrafted, you know, custom homes and stuff like that. I think there will be a market for human content. And you know Daniel, we've had the debate on even our own industry, right? What are you, which I think you know, we agree on some points, we disagree on other points but you know the human nature there, you know there still is value and I think a premium on doing things that AI can't do very well.

‍

Daniel Newman: Agree with that. So my read on the whole situation is that ultimately everything is changing, everything will be efficient. But I do think some human things, the question is, is it like going to be combustion engine cars where it becomes exotic, unique and premium but very small part of the market in the longer term or does it become truly a premium with a real market segment nation and scale and growth Meaning we're actually the demand curve starts to swing back to actually wanting more realistic human content. And also by the way, I think there's a difference between like this kind of stuff or you're watching like a video. Like are you going to want an AI actress? Like do you see the thing about the AI actress that got its own agent in Hollywood now that's like, like that versus like you know, like if you use prompts all the time. I do too. Like the stuff it's writing is getting to the point where it actually really sounds like me. So it's like is there a Premium to have me actually sit and type out a note? Or is it okay if it's me via an AI that allows me to do 5,000 times more volume? I don't know. We're going to find out pretty soon though, Pat. I mean you and I, you know, if we're, if you're CS holding signs, you know, will, you know, we'll pundit for food or whatever because you know, it all ends really abruptly, you know, which side of that argument ended up being correct. Anyways, meanwhile. All right, let's hit a couple more things. Microsoft doubles down, triples down. I'm going to move a little faster because I know we've been dragging a bit. But Microsoft points Judson Altoff as their commercial CEO, which to me is like, hey, we all know this consumer stuff. The OpenAI ChatGPT, all these tools have taken off in AI, but everybody's asking when do we get more value from AI in the business? It's taking time. A small percentage of data is hitting. But you've seen big Azure growth. But again, are these applications consumer, are these applications for business? I think the enterprise opportunity, the consumption, the implementation, the utilization, the productivity, the efficiency that needs to go into business from AI is going to be the next big proving ground. And so Judson's been one of the top executives for a long time, but he's basically going to take the reins, Pat. He's going to pull together and really put that emphasis. And I mean Microsoft's run to all time highs. Clearly the market's buying its story. Azure's had great growth. My read is, look, right now it's about telling the right story. The right story is Microsoft wants to be on the right side of this enterprise software transformation of this software industrial complex transition that we're in. And it looks like putting someone to make sure they eat, drink, sleep and breathe. This particular item is a good decision and I wouldn't be surprised to start seeing more of this from some of the other hyperscalers, some of the other software companies because this is going to be a double down on getting value from AI if we want to keep this bull market running in a bullish direction.

‍

Patrick Moorhead: Yeah. Daniel, conversation you and I had, I think a week ago, I can't believe how many things are going on at the same time. Even at a high level, it's hard to keep track of what's going on and what's important. I mean one of the, the biggest examples of this, you know, Microsoft does the biggest data center announcement they've ever done and the most powerful in the entire industry in Wisconsin the same day you've got the Nvidia and Intel announcement that happens and the Microsoft Wisconsin data center announcement gets drowned out. I think what we're seeing here is these big companies recognize that things are moving so quickly and there are large players they're competing with but also smaller players they're competing with and they need to work faster. Right. And to me putting a CEO in charge of, you know, putting Judson as CEO of the commercial market and products and operations and while engineering doesn't, doesn't report into him, I am told that he gets final decision making on products that should move better and also better understanding what your customers want and what they're willing to pay money for. I think that will speed up and then you've got Satya that will spend less time doing that and more time invested into the bear tech technology kind of like a founder and also the gigantic capex investments that the company is making. So this is a classic divide and conquer and Sati is still the CEO of the company but he has a co CEO that works for him but is empowered to make the decision. The buck stops with Judson on this. I like what I see.

‍

Daniel Newman: Yeah, I, I think that's a good assessment. So let's hit one more item Pat on the policy front just really quickly. The 50 of chips locally thing, I mean is that, is that just like moonshot at this point?

‍

Patrick Moorhead: Listen, 20 in 2030 looks to be a moonshot particularly when you look at the growth in AI and I doubt you know we will hit that and that was the original chips act. So you know, 50% by X time frame. Yeah, it is a moonshot. You know in fact it's not a moonshot, it's a Jupiter shot. Right. We're going to put you know land people on the, you know, the horrible planet that's Jupiter that kills everything that sits on top of it and takes a decade to, to get there. So yeah, but I do see this is strictly from a business and policy standpoint. I think the administration has tried all these different ways to get more investment in the United States and I think they've finally woken up to the danger that having this concentration inside of Taiwan is, and it's not just the oh, China invades Taiwan and cuts us off from, from everything. It's the, any negotiation that Taiwan has with us, they have the advantage in any type of negotiation even if they remain a, a sovereign, a Sovereign. A sovereign country. The winners here. I mean, intel has to be a winner if this is real. Okay? And this to me looks like, you know, a continued negotiation, but it looks like the administration has more clarity around what they really need, which is 50%.

‍

Daniel Newman: Got it. So, Pat, there's a bunch of other things. So much in a week, but you and I actually have to do real work a little bit today. I know all you wish we could just sit around and, and pontificate all day, but I want to do a flip with you. I want to hit this XPU GPU thing. I mean, look, it was like Broadcom is going to do chips for open AI and then Nvidia comes in with 100 billion and now people are like, well, maybe they won't do the chips. I mean, can they get that much investment from Nvidia? I mean, technically, yeah, no, maybe. And now MTIA and they're buying companies to advance, but they're also working more with core weave. And of course they want to get faster capacity. So should these hyperscalers just wave the white flag and go all in on merchant infrastructure? Let's go to the flip. Okay, sorry, I quickly distracted myself. Pat, you are the go. So the question here is, I mean, is it time to wear the. I think you're saying yes, wave the white flag and just go all in on merchants. Go all in Nvidia, all in amd. Scrap your Broadcom project. Scrap your custom Marvell. It's Nvidia time, baby.

‍

Patrick Moorhead: Yes. So for this simulated debate here, they should go all in on merchants because it's all about time to market. Okay? It's not about cost savings, it's about time to market right now. And I think Jensen put some good Jensen math out there that talked about. And I think this can apply to AMD when they get their Rack scale. Rack scale solution out there as well. You're actually better off going with a merchant GPU than even getting an ASIC for free. So the amount of tokens per dollar and the time it's going to take to get that out and ensure, you know, GPUs will consume more power in aggregate. But it gets back to how many tokens per. How much revenue can you derive from that? The other thing is that, you know, look at the, I'll say, you know, put Cuda and RockM out there, the amount of developers who are developing for those, compared to the interfaces that, that go into all of the XPUs, it dwarfs any, any single company's software bench. So the Developer friction is real and every one of those alpha site interviews, that's 100% what they, what they're saying. The other thing is the hit or miss view of it. I would say that you know, even, even for one of the biggest, the, the best XPUs out there probably has a 50 hit rate on whether it's hitting, hitting or not. And a lot of that has to do with this, you know, vicious cycle of AI and the models and oh my gosh, okay LLMs we're there and then we get into multi turn reasoning and then we get into pre training and, and the things that Deep Seek brought to the table. So literally every ASIC dollar could fund more capacity with GPUs, better networks or, or even you know, yeah, new data centers out there because each one of these XPU Designs is between $300 and $500 million out there. And you know, let's just say you buy into my 50% hit rate. There might be a time in the future when the model vacillation and the curve dies down. Like ASICs, XPUs were really good at machine doing object recognition, doing one thing, there might be a time but the time is not now. My final comment is that you, I could see a situation where intel could continue their Gaudi 3 as a merchant. XPU that you could develop on signal 65 has done a lot of good testing on that. That, that shows how much more efficient an XPU is than even a GPU. So let's say you've got amd, you've got intel, maybe you've got Groq as, as, as an option, maybe you have Cerebras, Maybe you have Goudy 4 that intel could create. So yeah, it's time to stop at the custom stuff here folks.

Daniel Newman: All right, that's the, that's the jam Pat. But I disagree. I think it's a terrible idea to discontinue it. You know we have a 583 billion dollar TAM by the end of 2029 it's still going to be about 60 to 70% GPU and merchant off the shelf. But that other 30% or so 100 plus billion dollars is going to be the largest hyperscalers deciding to control their own fate and wanting to make sure that they optimize their supply chains. They never have capacity issues related to being able to get something from merchants and of course being able to harden a piece of silicon to be used for the very specific needs that they have to optimize within their infrastructure. So look, if you don't want to make as much money and you don't want to be as profitable and you don't want to guarantee your future, then, sure, go all in on Merchant. But if you actually want to say, hey, we still want to be worth a few trillion dollars down the line, and yes, it's great that Nvidia is going to be worth 5 trillion very soon, but we want to control our fate. We've got businesses to run, we've got customers to serve, and we've got profits to make in order to return value to our shareholders. You're probably going to go down the XPU route and look, everything is as bad as it'll ever be today. All these things are going to get better. AI is actually going to make chip making better. It's going to optimize these designs, it's going to shorten cycles. You got these companies that are actually helping to do this, to make this happen faster. The Broadcom, the Arms, the synopsis. These companies are helping build and visualize and design and optimize silicon so that they can be taped out faster and brought to market more quickly with better results. The software work has to be done. But heck, everybody thinks Huawei can do it. So you don't think Microsoft can do it? You don't think that, you know, Meta can do it? Of course they can. It's just a matter of time. The Asics we have today are the worst they will ever be. They will continue to get better with every generation. The Google TPU has absolutely proven that you can make a hardened piece of silicon that can be pretty darn flexible and can do pretty amazing things. Why I'm so bullish on Google is they got that done first. They've done it the best. Amazon is following this is my biggest complaint about Microsoft. It's got to get this done sooner than later. And by the way, Nvidia is going to get more than their fair share with close to $400 billion of AI infrastructure revenue by the end of the decade. So I think we can have our cake and eat it too. But these hyperscalers need to control their own fate. And don't, don't, don't cry in your soup bowls, Nvidia bowls, because you guys are going to be just fine as well. So, onward, build your custom chips. Give Pat and I something to talk about for the long haul. I think you and I both agree it's a mix. Yes, in the end.

‍

Patrick Moorhead: Yeah, yeah. And when the sine wave slows down, it's going to make it a lot easier because you're not going to have to crank out a new XPU every year. Right. And you know, the design tools are getting better, but one of the biggest expenses is trying to get a mass set inside of tsmc. Bleeding edge, bleeding edge here and industry is, is investing a ton to, to disconnect themselves or decrease the blast radius of, of Cuda.

‍

Daniel Newman: So look, it's a win with, but they all want to have some margin optimization in the future. They're going to have to. All right, Pat, just a few minutes here left on the show. Thanks everybody for joining us today. Let's just hit a couple of things in Bulls and Bears. It's a light week because first of all, we're kind of through the earnings cycle. But just a few more weeks and we'll be back. So that's great. We can do that again. But look, there's only really one piece of Bulls and Bears news this week and you can call it one and a half. But like our government shut down this week. Okay. Democrats refused to pass a spending bill that would keep a budget that would allow the government to remain at least partially funded, if not entirely funded, which by the way, my daughter told me one of her friends was ROTC at Baylor, got a letter this week that her tuition is no longer being paid. So this has hit hard and it's hit fast. A real instance of where this shutdown I guess potentially is impacting the market and impacting people out there. Because like you and I, I don't think we see a lot like, I don't know, stop lights are still on. You know, we're not really because a lot of the things that we think services we get, these are municipal and state services. That's the whole argument of the Trump administration in many ways is that we have too much federal government, but things like that, the bills, the jobs and labor, we didn't get a jobs number today. Not that they're ever right. But what do you think about that?

‍

Patrick Moorhead: I mean, I didn't trust the numbers anyways. So I mean, I don't trust anything at this point coming out of the BLS. But the closest thing we have is ADP, right, with private payrolls. That said we had 32,000 jobs in September, but hiring plans have hit a multi year low. And you know, if I follow your X platform, you are talking a lot about that. And what I'm trying to figure out is if we're deporting so many people and I really don't think we're laying off. I mean, the tech companies are laying off a lot of people. But what is actually going on here? The only direct line that I can find to real employment, not government employment, is interest rates. And I thought that was an interesting thing that Claudia showed a clear line between the ebbing and flowing of interest rates and non governmental jobs that were, that were out here. So I think, you know, we need to, we're paying the price for all the deficit spending during the dark days and you know, we can, you know, debate and argue whether that was the right thing to do, but I'd rather not. But we pumped so much money that was just backed by debt and we're still churning through that money and then we still. I. Even though it doesn't show it in the data, Daniel, the small businesses have to be getting hit by tariffs. I just, I just don't believe that. I don't believe those numbers because I look at small businesses and how they buy. Like do they have the ability to just turn around and oh, I was buying from China before. I'm going to buy from Vietnam, right? I'm going to stand up to a supplier. Right. I'm a grocery store and I buy, you know, I've got 5,000 SKUs inside of the store. And I think this, you know, is having an effect as well.

‍

Daniel Newman: Yeah, I'm going to keep it short and sweet, Pat. Look, I have a really hard time believing any of the numbers. It's our own, you know, confirmation and recency bias that we like to agree with the numbers that serve the purpose that we want and we disagree with the ones that don't or will proclaim they're not legitimate or real. Look, I've been overwhelmingly the one that says AI has been meaningfully going to impact jobs. Maybe not immediately in terms of where it actually is, but companies are changing spending and making decisions and how they're investing into the future and hiring because of it. So I continue to believe that jobs are the real risk. The reason we need to bring down interest is we need to stimulate businesses to keep investing because that investment is what's going to keep the economy turning. We're going to have to figure out if it's uncomfortable. We're going to have to have a real conversation about jobs and about AI at some point. Like that whole rhetoric of every transformation brings more jobs. Maybe that's true for AI in the longer term, but the short term and how fast this has happened has not given humans the ability to keep up with the kind of designing the roles that we will have into the future. And so I think it's, it's, it's a long and short tale. And again it's the people's inability to see both short term timing and long term timing. No, AI hasn't replaced your job yet. Today it might be 50% or 60% but not completely. And that's why they haven't fully replaced things. But trust me, in the boardrooms, they're looking at things and how it can be done and they're making forward looking investments based on that thing. You know, in terms of the overall shutdown path. This government just, the whole thing is such a disaster. It's not a right or left thing, it's a way to spend as much as we can. Both parties are the same. We're only option, and I think President Trump was honest about it, is we have to grow faster than we inflate. And so the bottom line is all policy goes that way. That's why I tend to be so bullish. And you can fight it, you know, and you can be, you know, like they say, bears or shorts sound smart and bulls make money. You know, the bottom line here is that the fight is going to be they're going to keep pouring, keep spending, keep growing, because that's the only option we have. If it actually goes the other way, it's not a little bit of a problem. It's fucking catastrophic. Sorry, beep that out. It is catastrophic. Our economy will implore, implode if we stop growing. So real things to worry about. I think this gets resolved because I think the Democrats realize if they don't resolve it, they're going to get crushed in the next election because for the first time the shutdown is being blamed on them. Historically it's been Republicans that are sort of the ones that have done this kind of tactic and they've taken the heat for it. But right now it's that not enough Democrats are crossing the aisle to keep the government funded and real people are losing jobs right now. And I agree with you about small business. I don't think we know how to document the inflation impacts on these smaller companies. I do think that's probably where inflation is hurting the economy. It's not hurting it the way that all of the panic ins said it would hurt it. Overall, we're still growing. GDP outperformed most expectations as a whole. But again, how much do we trust that? I don't know, Pat. I got to run. I got a call. Started two minutes ago. But I want to thank everybody out there. Great. Friday. Bring your Peptides. Bring the guns. Not the bad kind of guns. These kinds of guns.

‍

Daniel Newman: All right, Pat. Thanks, everybody, for tuning in. Be part of our community here at Six Five. We love you all. We'll see you all very soon. Bye.

‍