The Six Five Pod | EP 298: Arm’s Big Bet, OpenAI’s Pivot, and the Real AI Infrastructure Race

 The handpicked topics for this week are:

1. Arm Unveils AGI CPU — First-Ever In-House Chip, Co-Developed with Meta: Arm steps into direct silicon production with its AGI CPU, raising questions about vertical integration, ecosystem neutrality, and how this move reshapes competition across the data center landscape (The Decode)
2. OpenAI Kills Sora and Doubles Down on Enterprise: OpenAI pivots away from experimental consumer products to focus on enterprise adoption, signaling a sharper push toward monetization and long-term business sustainability (The Decode)
3. Tesla, SpaceX, and xAI Introduce Terafab: A new manufacturing and compute initiative highlights the growing importance of vertically integrated infrastructure in scaling AI and advanced technologies (The Decode)
4. AI Data Center Moratorium Act: House senators introduced the AI Data Center Moratorium Act, proposing a federal pause on new AI data center construction until comprehensive regulations are established (The Decode)
5. RSAC 2026 Signals the Rise of Agentic AI Security: This year’s conference underscores a shift toward securing autonomous systems, with agentic AI emerging as a new frontier in cybersecurity strategy (The Decode)
6. The Flip: Can Tesla Actually Build a Semiconductor Fab? Or Is Terafab a $10 Billion Fantasy? The debate centers on whether Tesla can successfully build its own semiconductor fabrication facility to produce custom AI chips for its autonomous vehicles and Optimus robot fleets (The Flip)
7. Intel & AMD CPU Shortage Causes Stock Surges, Along with Dell & HPE Gains: A global shortage of Intel and AMD CPUs sent both stocks surging on March 25 (Bulls and Bears)
8. Qualcomm Is Downgraded From Outperform to Market Perform: Bernstein downgraded Qualcomm on March 26 from Outperform to Market Perform, cutting the price target from $175 to $140 with the pointed note that “investors can buy actual AI winners instead” (Bulls and Bears)
9. NVIDIA Remains Rangebound Despite Strong Demand Signals: Even with continued demand for GPUs, market performance reflects uncertainty around valuation, supply, and future growth expectations (Bulls and Bears)

For a deeper dive into each topic, please click on the provided links. Subscribe to our YouTube Channel so you never miss an episode.

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 reference share prices, but nothing discussed should be taken as investment advice. We are not investment advisors.

Patrick Moorhead:
Welcome back to the Six Five Pod. We made it 298 episodes. I wanna thank everybody out there. Daniel, we're gonna have a huge celebration on 298, right? Maybe a protein pig out at the well downtown, maybe some creatine and some peps to add. What do you think, Bestie?

Daniel Newman: 

298, I think we gotta persevere for two more episodes.

Patrick Moorhead: 

Yeah, and what are we going to do? Let's pre-plan the celebration because you and I, we always like celebrate our- 

Daniel Newman: 

Yeah, we're big, you know, victory laps. Generally, I have the company on my birthday, everyone has to actually individually write me a memo. Yeah. Oh, you know, wonderful. It is to work with me. I do things like that because there's just never enough opportunity to celebrate oneself. Right. But yeah, 300 is a lot. That's like five years, six years, six, six years. Almost every Friday, dude. That's that's that's an accomplishment. That's an accomplishment.

Patrick Moorhead: 

You know, I can't imagine doing this with anybody else. If something ever happened to you, you know, I would just pretty much have to retire. Hang on one second. I'm going to get the control room to zoom me in a little bit, because you're big and I'm small. There we go.

Daniel Newman: 

Actually, the setups were even, everybody, just so you know. That's not an illusion.

Patrick Moorhead: 

Yeah, just for everybody, when we do our live stuff, Daniel, we have to put him in the far seat because I'm just so much smaller than he is.

Daniel Newman: 

Director of frame mugging.

Patrick Moorhead: 

Anyways, I'm cortisol maxing this morning. Uh, it's so funny. Uh, I've got an adrenal issue and, uh, I got to take a bunch of, I got to peak my cortisol. You want to do cortisol maxing, take some of this, you guys, man, you will be on fire. Man. I'm looking down on my watch. It looks like I'm having a workout right now, but I'm really good about what that means.

Daniel Newman: 

Tell everybody out there. Cause I don't think everybody does it.

Patrick Moorhead: 

Yeah, so cortisol maxing, but the fun thing on all the socials are maxing something, right? And, uh, people talk about cortisol maxing typically as it relates to stress. Uh, but with me and I've, I've had, uh, three tests on my cortisol, cortisone interaction and, uh, You need to spike it man in the morning, right, then that's just normal. So you have a slight drop off and it used to be about 3pm. I would be tired, lethargic, and then I got my act together on health, but I still Had it had a weird curve. I was at the bottom of normal, right? I was just at the hairy edge. So I cortisol max in In the morning and this show is not sponsored by adrenocort. I just want to share that with with everybody But yeah, hopefully it makes me sleep better at night. I don't wake up at 3 a.m. Oh, it's not working very well.

Daniel Newman: Is it? Oh, man. Well, I'm wishing you all the best with your stress.

Patrick Moorhead: 

Anyways.  It was a big travel week for both of us. I attended three events and a business meeting in New York, dropped in on the Intel vPro launch with Panther Lake, did the HP Imagine and then met you up on the West Coast. with for RSA, the security conference. And yeah, I did a red eye in between, got up at 3 a.m. on Monday, and I'm feeling it, Bestie. I put my HRV, I posted it on X, and it's probably about 10% of yours.

Daniel Newman: 

Well, you should work on HRV maxing this weekend, but hopefully you have some downtime, right?

Patrick Moorhead: 

Oh, man. Yeah. Yeah. So much downtime, right?

Daniel Newman: 

When are you and your agents going to start? When are you and your agents going to take a break? I mean, because here's the thing is you could take the weekend off and still have your agents working.

Patrick Moorhead: 

Yeah. No, it's funny though. I spend all weekend and literally all of my free time How many do I have? Let's get a tally here of how many I have going. No, I literally said completed 1, 2, 3, 4, 5, 6, 7, 8, 9, and I'm working on two more. I've deployed two internally, and I'm about to deploy a third for the company. Built a health app, which I know we're five minutes in here. We've got to move forward, but have fun.

Daniel Newman: 

No, you don't. This is actually more interesting than the next part.

Patrick Moorhead: 

It is. But hey, why don't we move into the first part of the show, the decode, where we decode the latest tech topics, try to separate the signal from the noise. Let's go. Daniel, you were out on the West Coast a little early and you attended the ARM AGI CPU event out there on a pier. I think I was at one of the first Google Cloud events that was on that same pier years ago. What's going on here? Is this a big deal? I mean, they're stock popped. We'll talk about that later, but.

Daniel Newman: Yeah, yeah. I mean, look, this was a story that I think both you and I have been expecting. It's been something that we've been, you know, previewing in various capacities. You know, I've been the most vocal about ARM in terms of its roadmap and its desire to change. And look, the big frame here is that are kind of just blew up its own business model. Now, you can call it something else, and they're calling it something else. And it's not negative. So when you say blow up, that kind of gives us negative connotation. But it's actually just, they've been in this process of progressing from kind of pure royalties to, you know, licensing. And of course, then they move from kind of just IP into subsystems. And now they've got an IP subsystem to full chip, which there's no way you can look at that, though, and say they did not change their business model meaningfully. So what, 35, three and a half decades of basically being an IP, they're now shipping production silicon. So, you know, they've got a really good first partner in Meta. And they've got a, you know what I would say, an impressive customer list to get started. What was it? Meta, OpenAI, Cerebris, SAP. A Cloudflare and a few others. I can't remember them all. Yeah. But a solid list. And, you know, they're putting out something that meets the moment. Pat, you and I have talked a lot over the last few shows about the CPU's moments. Well, this is something that Arm has a lot of knowledge, a lot of experience, a lot of IP, and it can build. And probably the reason it's stock pop, which is your point, we won't hit that too much here because we'll have a chance to talk more about that later, but is they're, you know, a company that has about a billion dollars a quarter of revenue. And now they're basically saying within five years, they're going to be five times as large. They're going to 15 billion of revenue from this particular product. Interestingly enough, this is one of those push and pulls. And there is a risk to it. You can't ignore that there's some risk. ARM is now competing with its own licensees. And this is one of those things. It's a really delicate balance. I think in many ways ARM is threading this needle. They're working with a small subset of the partners. They're being very clear. They had the walk-ons from Jensen and Matt Murphy and from Charlie Kawas at Broadcom and Google leadership. So they had a lot of people walking out, but I think that's the question people are going to ask, whether it's Qualcomm, whether it's MediaTek, whether it's any of those hyperscalers, are they competing? I think the idea here is there's enough demand to go around. The ones that are going to build their own are likely not going to be looking at this product, but there's plenty of companies that will not build their own CPU silicon optimized for the AI era. And I think this is a great TAM expansion. for ARM, which needs it, because if you're just in CPUs, there's some limitations. And the last thing I'll say is they did tease a trillion dollar TAM. And as far as I know, CPU is not a trillion dollar TAM. And so my thesis holds true. that this isn't the last chip they're going to build. And then there's going to be a chip in the future. Maybe it's some Masa SoftBank collaboration with all the acquired IP over there. But I think there's more coming.

Patrick Moorhead: 

Yeah, that was a great, that was a great breakdown. I'm not gonna lie. I mean, essentially, you know, Arm spent 35 years as the arms dealer, and now it's building the army, right? And the question is, whether its best customers just became its best competitors, right? I think that will be a big debate going forward. And just to fill in some of the cracks, Arm, right, has licenses, and then it has IP, right? Arc license, and royalties on IP. And then it builds, it goes the next step, which to actually, it's called CSS, where they're actually testing that IP even down to the manufacturing level. And this was the, I wouldn't call it an obvious next step, but it's interesting. I had a conversation with Simon Seegers years ago. He was the former CEO of Arm. And I'm like, Simon, for as important as you are as a company, you're not making a ton of money. And this was during the big iPhone, Apple craze, even before Arm was even a factor in the data center. Maybe they were doing, some of its customers had IP and storage arrays, but that was about it. And here we are now, CEO Renee Haas has taken it to the next level. And it is a fundamental business shift, as you see. And if you know these chip pricing, right, we're looking in the thousands of dollars, we don't know that yet. But essentially, yeah, I mean, They're the full part of the value chain, and some will move from CSS to buying chips. Some will move down the chain. It depends on how you look at it. And yeah, there will be likely some people leaning more into RISC-V. And the thing that took ARM probably 10 years to build a data center stack and really only got a full head of steam when AWS went super aggressive on Graviton. So I think this is interesting. There's no independent third-party benchmarks on the performance. There's no independent third-party benchmarks on the 300-watt TDP measurements. And I'd like to see those. And we just happen to have a business called Signal 65, who I think would love Love to do that. Love to do that testing.

Daniel Newman: 

Let's go. You heard it here, Rene. Let's go. Yeah.

Patrick Moorhead: 

Hey, let's go. Let's get into the next topic here. So you might have seen that OpenAI killed Sora.

Daniel Newman: 

Doubles their headcount to… Isn't there a video of you benching 405 that it made on Sora?

Patrick Moorhead: 

I think it might just be 315. It's funny. I asked for 315, but I think it gave me four plates on each end. And then the problem was it also gave me three chins in addition to my 315. It made me hugely fat. But yeah, I did four or five fun videos on that. It's funny, I even tried to replicate my week ahead in video, but by that time, it cut off extensions and I was super sad about that. But, uh, yeah, a couple of ways to look at this. It's like, you know, it's really about enterprise, uh, their enterprise pivot and a race against, uh, anthropic and the IPO clock. And it is, it is amazing how open AI looked like it, it absolutely had a headstart that nobody could, um, nobody could get ahead of and then you know google hits with jim and i three point oh and then you and then and then you know the whole phenomena and traffic where i moved most all of my workloads. I'm too and traffic plus perplexity computer using and traffic. So, yeah, with the IPO clock coming down, let's shut down Sora. There was some discussion, not in a public forum, sorry, from OpenAI that said, hey, we're going to focus in on the enterprise or business workflows. But if that's true and they do need to reduce cash flow, you're beeping like crazy there, dude. Sorry, I'm famous. One of the things to shut down would be Sora right which i thought was cool but i don't think it's vital to their ongoing concern i do think that they need they absolutely do need to put out their strategy are they competing with google with advertising and consumer and with apple. are they primarily competing with Microsoft and the other business startups? So we will see.

Daniel Newman: 

Yeah, I mean, my read on Sora was the demand was falling very quickly. This is one of those products that I think it's novelty. But it wasn't necessarily commercial, meaning there was a lot of novelty and building on it. And those that were willing to commit, if you kind of looked at the utilization of it, it had that big spike when it kind of came out and it fell pretty quickly. Most people are not going to build these sophisticated generated videos with it. And I think people are seeing that. And of course, I think Google, you could argue has a bit of a better mousetrap. for it, definitely more available for people from the distribution standpoint, people that use Google with Nano Banana and different things that Google are putting out and just making image and video generation more democratizing it. So I think OpenAI has to focus. Right now, they're kind of being lapped in a number of different areas. Code generation, you could argue in model efficacy, quality. I saw a number that Anthropic is growing 30% week on week in revenue. OpenAI has got to answer that. If they want to do an IPO right now, they've got to get back to winning. And I think doing too many things is often a recipe for not winning.

Patrick Moorhead: 

Good stuff, man. Let's move forward. Next topic. And we're just going to do a crop dusting on this one, because we're going to debate this one in a flip later. But Tesla, SpaceX, XAI, Elon Musk, introduce the new TeraFab.

Daniel Newman: 

Yeah, like I said, I'll just keep it tight. Maybe your AI notes are saying the same thing my notes are saying. L-O-L-O-L-O-L. You know, joint fab project, Tesla, SpaceX, XAI. And it's very specifically here in town, Fab. It's here in Austin. Um, it's they're targeting two nanometer, which is not trivial by the way. This is not, uh, do we even have, we don't have two nanometer here yet. So they want to, you know, of course say they, you know, think they can do this quickly. A hundred thousand wafer starts scaling to a million. Uh, they're, they're saying at a 20 to 25 billion costs. Um, nobody seems to believe that number, by the way, just noteworthy. Um, And basically they're saying most of the targeting of what they're going to create is for their orbital AI satellites. So it's not a, they're not looking to do build a leading edge fab to do chips for Nvidia or AMD or, or build our AGI CPUs. And so, you know, I think that's kind of, that's kind of the headline. I think you and I might argue this later, whether this is a thing or not a thing, and you maybe can subtly hear what my thought is about.

Patrick Moorhead: 

Yeah, I mean, I think you laid it out exactly right. We're going to debate it, you know, a lot later. It's just, it's a big thing to, it's a big thing to tackle. You know, there's only really, you know, three foundries left out there in the world. And Elon Musk creating another one, it's going to be, it's going to be an interesting thing to watch. We will debate that soon. All right, let's get into a fun, fun topic here. You've got AOC and Bernie Sanders basically want to introduce the AI Data Center Moratorium Act. I think we should call it the Doomer Act. They want to wait until we have comprehensive regulation out there impacting energy and water consumption, consumer electricity protections, labor and civil rights safeguards and AI safety standards. So I don't know if you could pile more slop into this. One thing that's already been done is The hyperscalers already came in and said, hey, we will eat any increase in potential power generation costs that are out there. They also addressed water as well. And there's a group out there that is led by Elon Musk which is technology to help the power companies be more efficient. So for instance, they don't store any electricity in batteries. They build everything to peak where they could be generating less energy, storing it, and potentially giving that energy, distributing it somewhere else. There's also certain technologies that They could use to be more efficient as well. Listen, I just think this is the dumbest idea out there. It's just more let's get more taxes. Let's slow stuff down. It's almost like. They just, this is a new way for these two to monetize themselves as candidates. I really, you know, I can't imagine us slowing down and letting the Chinese, I mean, the Chinese are gonna blow us away in robotics and that form of AI and self-driving cars, by the way. And everyone's got the best software out there. Yeah. So do we really want to let them win on the data center front? Anyways, this is a scary political thing because it's the dumbest idea. And I've seen a lot of dumbest ideas go through because the general population is looking at the short term, which I get. But I feel like the short term has already been taken care of. which is power generation and water subsidies.

Daniel Newman: 

Yeah, this is dumb. We don't have to talk about it anymore. This is like next level dumb. This isn't the first time, by the way. This is a kind of a rehash because these two clowns have been calling for this for a while. They would literally tank the economy. I mean, that's their goal right now. Keep the economy shitty because they got an election to win at the midterm. Yeah, why not set us back in this particular category, too? And then we can go back to using, I don't know, calculators and pens for tests. And we can probably just get AI out of all schools, too. We should not embrace it.

Patrick Moorhead: 

I think we should go back to candlelight at night. Candlelights, gaslighters.

Daniel Newman: 

I mean, they are gaslighters.

Patrick Moorhead: 

Yeah, or rolling brownouts like, like in the third world, third world countries. And I think we should stop, you know, using all, I think we should go back to horse and carriage as well. That would, that would help and get rid of elevators.

Daniel Newman: 

All the things that we used to use to do calculations, because it just totally slipped my mind, like calculators, like,

Patrick Moorhead: 

You know, go back to Abacus. Abacus. I want to call it Chisholm bop. You remember that? Yeah, Chisholm bop is doing stuff on your fingers. It's basically advanced finger counting. Nice, nice. Anyways, okay, let's move on to a much more like that topic is nonsense. The topic is nonsense, but it's like, it's so nonsensical. It's hard to even discuss.

Daniel Newman: 

We gotta give it some air just to see how dumb it is.

Patrick Moorhead: 

So Daniel, we were both in San Francisco for RSA. I think I was there for a whopping 13 hours. It was good though. We made it count. RSA 2026. Wear and listen to our videos as they come out. We're not getting access control anymore. It's agent control.

Daniel Newman: 

Yeah, I mean, agent security was like the shipping moment, right? You know, these were, this was the, you know, PowerPoint, now it's shipping. I mean, this was everywhere, every company, new startups, legacies, platforms, disruptors, all in the space, all doing something with agents. We saw CrowdStrike. We saw Databricks enter in a more significant way. Wiz deepened, now Google, by the way. Congratulations. We didn't really hit that one too deeply because we hit it in the beginning, but that happened in case anybody missed it. But there was also a pretty interesting story this week. Did you see that LightLLM story? Yeah. Where basically there was like a vibe coded package that was basically harvesting SSH keys. Again, I'm not a security expert, but when I read about this thing, it basically only got found because this thing was so compute intensive that it basically blew up some security teams system. And when they blew it up, they basically figured out what it was and got to it before this thing spread. It would have been brutal. It's funny. I saw a headline this morning that said cyber stocks are down because the next generation of anthropic models are going to be so basically bad, so good that they're going to create so much more security risk. And I said, shouldn't that mean cyber security should go up? You know, like if it's more and more needed. But, you know, you had some great conversations with Veeam. We talked to Commvault. We talked to Tanium. You know, the agentic pivot in terms of applications are going to create this massive agentic need for security to manage it. I think security is hot. I think security is red hot. I think companies are going to have to think more and think faster. When people like you and me can build products and ship them, we sure as hell need someone to secure them. So I'm long security here. I mean, I know the market only goes down these days. That was only because you entered right about the time that that happened. So we call it the path.

Patrick Moorhead: 

My finger shakes every time I'm almost about to connect AI to my ERP system that I have in place. I mean, it's like every contract, every dollar amount, every bank account, it's like, man, I don't know if this is such a good idea. But anyways, I'm about to do it. I did upgrade my ERP last week to see what happens. Yeah, the one thing I want to hit on here, I'm glad you hit the light LLM. That was good. It is interesting, and I think you pointed this out really well in the podcast that we did during the show, which is this whole you know, dynamic tension between what you want to get done and security and just how that's always been the conversation. So a lot of stuff has changed. You know, my my week ahead in tech video, I said it was the most consequential RSAC in years. It is. And I think regarding agents, but a lot of the themes are exactly the same. I just don't see the fundamental differences in the macro conversations. Maybe the only one would be systems of action and what agents bring in here. A shout out to Cisco on its agent security. I think it's architecturally distinct. There's zero trust access for AI agents, is adding these task-based permissions, right? When you're about to wire Daniel money, or you're going to wire, your agent's going to wire my agent money. And it's a agent runtime SDK. And I think it's, you know, you build it across AWS Bedrock Agent Core, Google Vertex Agent Builder, Azure AI Foundry, and LangChain. I think that's pretty cool. Shout out to G2 and the folks at Cisco to continue to crank that out.

Daniel Newman: 

Good shout out, good shout out.

Patrick Moorhead: 

Okay, I would like to go to the flip.

Daniel Newman: 

Are you sure you're ready to suffer another defeat after your already cortisol maxing?

Patrick Moorhead: 

Listen, what's, you know, challenges just make you stronger. And that's why I'm so strong, because I get punched in the face or kicked in the groin. And I just get right back up and come back at it.

Daniel Newman: 

Maxing and then win maxing.

Patrick Moorhead: 

Kidding me?

Daniel Newman: 

 got to show me up, dude. Oh, come on. Yes, this is going to be part of the show, the weekly gun show. All right. All right. We just lost half our viewers with with the viewers. You know, we've got 298 episodes. You know, if you haven't kind of caught on to this, you know, this insanity, you're not going to catch it.

Patrick Moorhead: 

So, hey, we are going to debate, can Tesla actually build a semiconductor fab or is TerraFab a 10 billion dollar fantasy? Let's dive in. Daniel, you got the four. Let it happen. You are four. Vertical integration is the only way to guarantee AI chip supply at Tesla's scale.

Daniel Newman: 

Yeah, yeah. I think, you know, at some point you admitted you got the Musk chip design ability wrong. And I think you're going to get the manufacturing call wrong as well.

Patrick Moorhead: 

I do call out when I get my shots wrong, Daniel.

Daniel Newman: 

You do, and I'm giving you credit for that before I defeat you in this debate. First of all, you know, Musk understands what every leader in this space needs to understand and that the supply constraints, they're real and they're going to get worse. You know, we're talking a trillion dollars in semi revenues in 2026. And, you know, there's going to be a massive gap in foundry capacity for the next half a decade. And I think it could be more. So with every hyperscaler design there on silicon, you know, we know that whether it's FSD, it's Dojo and now satellites. All these companies need specialized architectures and silicon, and they're going to have a hard time being prioritized when NVIDIA and Apple are always going to be front of the line. Now, could it be Intel? I don't know. That's a possibility. But, you know, people are underestimating Elon Musk. And I think the reason they're underestimating is because they confuse his ability to do something with his timelines in which he does things. Reusable rockets, gigafactories, Starlink is at scale. I don't know about you, but Starlink's gonna be on every airplane for a number of airlines in the very, very near future, replacing that go-go garbage. Everything he said he's gonna do, he's gotten done. So basically what happens is he says he's gonna do something, pundits say he can't, and then ultimately he does it. The only thing that's always right from the pundits is they're always right that he won't do it in the timeline he says he's going to do it. And the other thing is, like I said, is he's not trying to be TSMC. He's not trying to build a fab that's going to be able to build every leading edge chip. He's very specifically focusing on his two fabs to do, I think it's a total of two different things, and each one will do one of them. So that specificity gives him the advantage where he's going to be very, very focused. And then, of course, you know, it's going to be his only path to his trillion dollar pay packages. He has to get this right for Tesla to hit its numbers, especially with Optimus robots. And then, of course, with everything else he wants to produce, he is going to need more capacity than any of these third party fives are ever going to be able to offer. So. Look, here's what I will say ahead of time. He's not gonna do it on his time. He will not execute this. I don't think he can get what he needs from ASML. I don't think he's gonna be able to get the people, the talent. He's gonna need the right talent in there. And we've seen how hard it's been for a company like Intel to find that person and TSMC sure as heck isn't gonna share their talent. I don't know where that talent's gonna come from, but you know what he does really well? He finds it. And critics will say that it's going to cost more. He doesn't care. He could personally afford this. The man's worth nearly a trillion dollars. He's going to show that with his SpaceX IPO that's coming out next. So look, this is going to happen. He's going to get it right. This is how he continues to maintain his moat and his advantage. And he's going to do what others wish they could do, which is execute and control their own destiny. Just ask Sam Altman.

Patrick Moorhead:

I mean, Daniel, given enough time, I think I could figure out something. So for you to say, yeah, yeah, he says something and delivers it 10 years late. I mean, I think we could all do that. But no, I mean, listen, fundamentally, chip design and chip manufacturer, different businesses entirely. I mean, I think he even thinks he's going to recreate ASML and you know, applied materials and somebody like that, good luck. And EUV machines are $200 million each, Daniel, as I think you know. And then high in AUV or $370 million and ASML can produce, I don't know, five to six globally. So no, no SML, KLA, applied material orders from Teslas have been disclosed. And yes, I know, I know it's early, but remember when chips are going to come out, you would have had to have put your orders in by now. And then it's, then it's the, you know, we talk about the IP. conversation which is national security you have to have fab ip in the developed in the country we don't even know who's tune animator process technology i think the only one who licenses what they do is is ibm at this point and i don't know maybe intel gets into the licensing game maybe you know sf2 by the way Samsung partnered with IBM on SF2. So yeah, so we have no idea. And quite frankly, you want to talk about a devastating precedent, it's Dojo, right? Teased in 2019. announced in 2021, powered on in 23, disbanded in 2025, described as evolutionary, a dead end, and restarted in January 26 and 20 years left to start a competing company. So, you know, this is what we could. And let's talk about Bernstein. our buddy at Bernstein. I don't know if this is Stacy or not. Yeah, it actually is. Bernstein thinks a full TerraFab would require five to $13 trillion in cumulative capex, slightly below Musk's 20 to 25 billion. Morgan Stanley, says $35 to $45 billion, which is half as much as Musk has put out there. Lest we forget, TSMC took nine years from Arizona announcement to two nanometer production. I could keep going down the list, Daniel, but I'd rather not to. I don't want to embarrass anybody who may have argued the other side of this. I am concluded. I am finished. It is finished. Okay. What do you really think?

Daniel Newman: 

I think it's going to be really, really hard and it's improbable, but he is one that I would never say never in terms of getting something done. I think, yes, I think your argument, the best argument you made was that if you give it 10 years, anybody can do something. I think that was the best part of your argument. But I do think if he can do it in 50%, even 100% longer timeline than he's suggesting, he's going to set himself up for a large advantage in the long term. And so I think he'll try, I think he'll try. The 5 to 13 trillion, that's interesting, my lights just went out. 5 to 13 trillion is a really large number though. That one I'm trying to figure out how you get from 45 billion to 5 trillion.

Patrick Moorhead: I think what it is, is you're building a brand new type of technology that does stuff like etching. You're essentially having to, it's what the Chinese are doing. They're trying to work out ASML. It has to include something like that. You're trying to replace applied materials. You're trying to replace KLA. That's all I can tell you. Yeah.

Daniel Newman: 

Seems like a big, big number. I don't think he's going to go for that.

Patrick Moorhead: Good conversation. So, hey, let's shift into bulls and bears. I do want to take a production pause. I asked twice to put Dell and HPE as the same topic because they're basically the same topic. So, I'm going to call an audible. So, Daniel, I'm going to introduce Dell and HPE, okay? And then do we want to do the Qualcomm conversation? Yeah, we can. The downgrade? Yeah. Yeah. Okay. All right. So that's what we're going to do. That's what we're going to do. All right. Welcome to the Bulls and Bears, where we dive into the latest earnings reports, FOMC discussions, pretty much anything that drives tech stocks. Let's dive in. All right, Daniel. A lot of numbers were down this week, but two that I wouldn't say peaked, but we saw a lot of action around was AMD and Intel. Is this the CPU chip shortage rally we're seeing?

Daniel Newman: 

I think it's two things. I think both are beneficiaries of the topic of last week, the demise, which seems to be spreading lawsuits now coming out against super micro. And when all that revenue potentially in the risk off of some of its supply chain potentially takes place, where's that revenue going to go? Well, directionally Dell seems like it would be the largest winner there. And HPE would definitely have the opportunity to capitalize on that as well. And then, of course, to your point, these are two companies that sell a lot of servers in the data center, CPU data center servers as well. And we're seeing shortages, right, Pat? I think we see numbers now that were one to two weeks, now six months. the CPU renaissance is here for real. And so there's the upstream beneficiaries and the downstream beneficiaries upstream. You saw the arm pop on, you know, it's numbers. And we talked about that, but you know, this is going to be good for Intel. If it can get a supply chain in order, it's going to be good for AMD, which has been taking share in that space for a long time. But the OEMs have an opportunity here and they didn't make the list, but I would imagine Lenovo can benefit from this too. The two things together created a pretty nice tailwind for these two companies. And so while the market at large is cratering, our friends at Dell and HPE are celebrating all-time highs, which is kind of wild, considering the AI trade has been pretty risk-off.

Patrick Moorhead: 

Yeah, I mean, it's funny how this started, right? Lisa Sue has been talking about this for a long time, her clients doing AI on CPUs. The market didn't want to hear it, so AMD stopped talking about it, right? But the reality, and then uh, you know, Jensen at, uh, at CES, he starts talking about this and I had a, you know, one-on-one with Ian Buck where, where pretty much we, we just talked about CPUs and, and, and Vera, uh, and then Intel shows up on their earnings talks about three to six month lead times. Uh, and then the market goes absolutely crazy. And then we go to GTC where you've got, uh, you know, Vera standalone rack. For CPUs i had to ask about that it's like you know was this just a simple slide or entire rack of vera not vera rubin but vera CPUs and the short answer was yes i also prior to that had confidential conversations under NDA. uh with uh two of the hyperscalers to do a a drill down on this and and and here we are right everybody everybody's having shortages cpus are uh are are cool again one of the things the cpu to gpu ratio could potentially move to uh one on one to one i think it was eight to one and four to one but we could literally look at a a, a one-to-one and, you know, GPUs aren't a, what I would call a volume business, but they're still in the millions, uh, which could be a, a huge, um, a huge pop, uh, pricing, uh, rumor has pricing is going up and that's not just pass through from TSMC and higher costs. It's that, um, AMD and Intel and, NVIDIA and soon to be ARM have been running hot lots. which are burning a lot more wafers a lot more quickly, accelerating stuff like packaging. I do think that from a stock standpoint, Intel probably disproportionately gets a benefit off of this because their stock, the value of the design company is so far is so far down. Right now, it's the speculation about Foundry that's driving their stock at this point. So, yeah, I mean, net-net, the AI boom broke the CPU supply chain and AMD and Intel are raising price because they can, right? Which I think nobody expected in 2026. Let's move to the next topic, and that is another rock in here. Dell Technologies hits an all-time high on AI server demand, and their stock is rallying also on the back of Supermicro and HPE. HPE is blowing it out. with its stock together. I want to tie these two together. HP at a 9% rally. Did we cover this?

Daniel Newman: 

Did I just cover this? I'm totally confused. Did we get things out of order? Because I think I already talked about this. Guys, just pause this for a second. Cause I talked about it. Unpause. Okay.

Patrick Moorhead: 

And five, four, Daniel, great job covering CPUs and infrastructure demand. CPUs being driven by AI and man, Dell and HPE stuff.

Daniel Newman: 

That was like a four part. That was like a me question, dude. Like four topics in one.

Patrick Moorhead: 

Listen, you're on a roll. Okay. So why don't you pick up Qualcomm? Stacy Raskin from Bernstein downgraded Qualcomm to outperform to market perform price target $175 to $140. What's going on here? Stacy was a huge longtime supporter, definitely a bull on Qualcomm.

Daniel Newman: 

Yeah, it's a pivot, and I think it raises some fair questions. I have a lot of respect for Stacey. Is he wrong here? I think he might be wrong. I think he made the call too soon. That's the thing. I don't necessarily know. Look, everything will be proof into the future. You and I talk about this a lot. Qualcomm has shown that it's able to diversify its business, but where it's diversified so far doesn't necessarily give it large credit in its ability to participate in this AI moment. It's shown, it's teased what it's going to do next with its AI inference racks. And Qualcomm certainly has the skill capability, has made some of the right acquisitions to build low power consumption inference servers. I expect them to capitalize on NPU IP to be able to take advantage of scale there on the compute side. This is one of those stories that hasn't come to fruition, but I think there's an important word that follows that sentence. It hasn't come to fruition yet. Now, if there is a trillion dollars of spend on semiconductors, Qualcomm, and I've said this in my comments, like, look, they get into this space, they don't need to win a lot of market share to really meaningfully move the dial. I think they've got some of these Middle East deals. We know the demand for inference compute is substantial. We know that Qualcomm has the skill and the IP inside to do something in this space. So my read has always been small market share, a couple of key markets, win a number of new billions of dollars while addressing a new TAM, I think it's a play they can do and do well. And that should deliver increased EPS, increased revenue, and it should be enough to show that it's not missing the opportunity to participate in this space. His comments on the Apple thing, it's just to me like that was like three years ago. I don't really understand bringing that up now, but I guess having that transition now being final and moving into production, quote unquote, is gonna be something to watch in the quarters to follow. But they've done a good job in automotive. They're showing some strength in IOT and edge. And like I said, a small piece of the market share. So I'm gonna say too soon, will he ultimately be wrong? My read is that yes, he will ultimately be wrong, but I also think that Qualcomm still has a lot to prove.

Patrick Moorhead: 

That was great commentary. This really comes down to, I'm trying to remember the exact words that Stacy used, doing well in a bad neighborhood. I think that was the description that he used. I think that that's code. Wall Street code for, hey, the fundamentals are fine, but the stock still goes down. Right. And what that clearly says is that the longer term automotive and the longer term robotics and the longer term data center is being being discounted. And then, you know, when you you and I had Stacey on the show, we asked him about, hey, what's your time horizon? He said 18 months. So i get it i get it in that eighteen month profile robotics isn't gonna hit in that time frame i think at their at their industry analyst day start their financial analyst a model they rolled out was twenty twenty nine. Right? So this makes sense. So I do think Qualcomm will continue to make progress, particularly in robotics and IoT and things like drones. But again, we'll have to see how this works out. Hey, let's move in to The last topic we're going to cover, we're going to talk about. NVIDIA going sideways for GTC. I mean, if you look at the stock's move, it's not doing great. I mean, even though NVIDIA has a trillion dollar order book on two products over a small period of time, their PE is still sub 22. I think you pointed that out in one of your tweets today. And that implies a 40 to 80% Upside so either entire sell side is wrong or the stock is the most obvious mispricing in all of big tech. So I think NVIDIA's position is really structural, not what I would call cyclical. And GTC confirmed pretty much what I wrote and I think you wrote before the keynote, that Vera Rubin is architecturally complete, the most architecturally complete AI infrastructure announcement of any semiconductor company has made, especially when you bring in LPX and the CPU part of it. And I think I don't think that sell side or institutional folks are necessarily counting this in, but the constraint is really in the end going to be energy, right? We don't have, there's no thesis or no spreadsheet that says that we get to be able to have enough energy for all of these hyperscaler data centers starting in 2028. We're good this year, we're good next year, and maybe there'll be some moon shots You know, there's a company that's taking old jet engines abandoned in the desert and turning those into energy generation. But, you know, we've got a two year lead time on natural gas turbines out there right now. So even if you can find the natural gas like they're doing in Texas, you may not be able to get a hold of them for a long time. Danny, what do you think's going on here? Is this Iran? Is this something weird and different? The entire market's in shambles.

Daniel Newman: 

I think like we have Iran, we have election, we have a fading Trump put on, you know, the economy right now. You have uncertainty around jobs. You have, you know, I just, I don't think this has much to do with NVIDIA as kind of the structural pivot that's going on in the economy. We're in a correction. So, you know, Microsoft's down 35%. You know, we've got major, you know, a lot of growth companies are off 50 or 60. SAS companies are off almost 60 plus right now, by the way, with no actual, with no actual real signal that their businesses are being disrupted. We are in a, you know, in a correction period. The good news is generally out of correction periods, you'll start to see things come back. This is when it gets really scary, but this is when grownups invest. And this is when people, you know, like you and I crawl into our holes. I'm kidding. We wait for it to hit all time highs. And then we start writing, we start buying again. But in all seriousness, like, look, it's very easy to look at them now trading at like sub 15 forward earnings. And the only question you should really ask if you're thinking about like, is this a good price? Is, do you believe the future earnings? If you believe this is real, then over time, this is going to win. I genuinely do not think this will lose over time, but I think you have to buckle up right now because with the Strait of Hormuz is closed and bond yields are going up and Trump can't turn the switch like he did back on Liberation Day, we might have to wait for things in Iran to clear up. I think we'll get there, but I actually think that has more to do with it. NVIDIA is doing amazing. If you own that, you could probably sleep like a baby. I mean, you really should sleep like a baby because they're gonna do a trillion dollars in revenue in the next two years and generate more free cash flow. And by the way, they can do anything they want with that. If they don't like their stock price, they can buy back. They can do like a hundred plus billion dollars of buybacks a year if they want. I mean, they got so much optionality. So a lot of nonsense out there. Be patient, you know, even I'm at that point where it's just maybe not worth looking every day right now. It may not be worth looking.

Patrick Moorhead: 

I mean, it's time for me to get ready to sell more because this is my best strategy.

Daniel Newman: 

Tell me when you sell, because I'll buy.

Patrick Moorhead: 

Let me know when you're about to sell. Yeah, typically my investment strategy is to buy high and sell low. So, and then justify it as getting a tax break, helping me on my taxes. Brilliant. Brilliant. So Daniel, let's shift to, I guess it'd be this week when it would be airing. Are you traveling next week?

Daniel Newman: 

No, I don't think I am.

Patrick Moorhead: 

I'm not going to be in here. I can't even believe it. I'm really excited about that. Yeah, me too. I'm going to be excited. Um, yeah, I, we, we should definitely meet up and, uh, and do, and do, do a workout.

Daniel Newman: 

Yeah. It'd be fun. We should know what I'm saying. Do a show, but I joined a new gym. My son's got a basketball tournament. I'm watching NCAA. Like I don't, I'm not like a big NCAA basketball, but it's, these games are fun to watch. Like I'm having fun watching, you know, so much more inspirational than watching NBA. Um, And I mean, dude, you got to recharge that HRV thing you've got going on that cortisol thing, like you need to you need to chill, maybe maybe do maybe some Doomer's and chill or something. I don't know. But like, totally.

Patrick Moorhead: 

I do know that I can make that happen because I've been through this cycle before. It takes me about 10 days and I'll be at 100%. But I just did too much. Three weeks of travel. It was stupid travel too. Two, three a.m. flights. I'm an idiot. Anyways, not going to do that again. But hey, I want to thank you all for tuning in. Give me all your travel tips because obviously I lost the plot. I want to appreciate you for tuning in. Hit that subscribe button. Tell your friends, families, pets all about the show. If you have any concerns with the show, you know where to find Daniel. If you love the show, you know where to find me. Take care. Have a great week.

‍