The Six Five Pod | EP 254: Unpacking GTC: Nvidia’s AI Dominance and the Hyperscaler Challenge

The handpicked topics for this week include:

‍

1. NVIDIA’s GTC 2025: Taking AI beyond the hype, Jensen Huang’s vision for enterprise AI is promising. The rapidly expanding hardware pipeline is setting the stage for a dramatic market shift. But can these companies deliver?
2. AI Chip Showdown: NVIDIA’s dominance isn’t going unchallenged. Hyperscalers are building their own chips, sparking a debate on whether the AI chip landscape is about to get a lot more competitive.
3. Google’s Wiz Deal: Risk vs. Reward? A $32 billion gamble raises crucial questions. Will regulators allow it? And does this move signal a new era of tech M&A, or is it a regulatory minefield?
4. MAG7 in the Hot Seat: Market uncertainty is hitting the tech giants hard, MAG7 are looking a little less shiny ✨Is the time to buy NOW? Or is this just the beginning of a larger correction
5. SoftBank’s AI Play: The Ampere acquisition hints at a strategic power move. Is SoftBank building an AI empire, and what does it mean for the future of data center tech?

‍

Watch the full episode above, and be sure to subscribe to The Six Five Pod so you never miss an episode.

Daniel Newman: Don’t you love when we ramble on? We’re so entertaining.

Patrick Moorhead: Are you not entertained?

Daniel Newman: You’re not entertained? We should get that one as a clip and then turn it into a gif and then we can laugh at myself especially because I’m looking so good today in my glasses

Daniel Newman: Hey, everyone, welcome back. It is episode 254. Moorhead, I forgot which number we were on. It’s episode 254. Welcome back to another week of The Six Five podcast new format. Week number two. We’re gonna get better and better at this thing. Patrick, I may or may not be in an undisclosed location on vacation, doing a terrible job of resting, relaxing, and taking in the libations. I’m going for the super moderator rhyme thing to start off. How are you doing this week, man? You’ve been everywhere.

Patrick Moorhead: You know, I was just in one place. I was in San Jose, out at GTC, Woodstock, the Super Bowl of AI whatever we want to call it. But, yeah, I was feeling pretty bad for you, Daniel. Especially your picture yesterday of you in a pool. The pool of tears falling on your back. You were lamenting your situation. But I just want to let you know, bestie, I was praying for you at that moment. Okay.

Daniel Newman: Yeah,. I appreciate you doing that. I don’t know if people really fully understood the metaphor, but the attempt of taking vacation as a founder in the final weeks of a quarter, when you’re in the process of building businesses, when most of your executive team is all on spring break, when you’re one of the largest events, the most influential events going on. And then, of course, I just couldn’t stay away. I mean, you were on broadcast. Like, I had to keep the TV off to, you know, to break my fomo because every channel I turned on, you were on every single one of them. It was like the Moorhead show, but I had to do a couple myself. Couldn’t stay away either. And, you know, it. It was GTC. I booked this trip, like, six months ago. I didn’t even know the dates, Pat. When I found out I was gonna miss it because I. I didn’t know. But, like, everybody wants to go see AI Jesus.

Patrick Moorhead: Yeah, they really do. But. But, Dan, you got to keep in mind, though, like, everybody does need some time off, and if you’re not going to take kind of the Christmas Hanukkah holiday, then this is the next best because everybody’s kids are off. Right. I noticed my email. My email volume went down significantly.

Daniel Newman: Yeah. You know, everybody needs a little time away. I heard them say, feel like breaking into Peter Cetera, you know.

Patrick Moorhead: Hey, here’s the great news. Your two joint ventures that we have together, Six Five Media and Signal 65. I was not on vacation and our presidents were not on vacation. So the wheel, the wheel turns, baby.

Daniel Newman: Yeah, they were crushing it. But yeah, we’ve got a really big week. Welcome back, everyone. Like I said, we appreciate you tuning in. I hope you’ll be part of this community and subscribe. Join us for all the shows. We look forward to getting more and more feedback as we continue to refine this effort. As a reminder, you know, the format of the past, the first 253 episodes. 252 episodes. I’m gonna get there. Where we used to do six topics for 25 minutes each and we’d go on 15 minute long monologues. We know that was great and everybody loved that. But we’re going to try to modernize this show a little bit. So the way this one breaks down is effectively we’ve got this first section, it’s called the Decode. It’s a little bit like the older format, but we move a little quicker. We float back and forth a little bit more than we used to, than, you know, in the past. And then we go to this segment called the Flip. And the Flip is fun because the Flip is basically where Pat and I take a side of an argument, not necessarily the side that we personally have, but we just basically flip a coin and one of us gets one side and one of us gets the other side. And we got a really interesting topic today. And then of course at the end, we do our bulls and bears section. I. I desperately wanted Pat to do some stock picking, but he fought with me about it. So we’re just going to do bulls and bears. I’m kidding, buddy. I know you.

Patrick Moorhead: True. Now, you do speak the truth.

Daniel Newman: I don’t want to call you out on it, but you know what’d be really funny is I’m. I talk all this noise like I know what I’m doing. You should have done it just because you could have kicked my ass and then, you know, you’d have been in the end and like, look at this clown over here making calls. Wrong calls. Although I got a lot of right calls out there. But anyways, we got a great show today. Just a reminder for everybody, this shows for information entertainment purposes. Please don’t take anything we say as financial advice. I feel like I got to say that every once in a while because these are big moments. The markets are crazy. And you know, more and more I hear people looking for our opinions on these things. And I just don’t want anyone to spend money as a byproduct of what we say. Or maybe I do, but I don’t. You know how that works. So. But we’re going to dive deep into GTC and then we’ve got a few other topics. A massive acquisition deal, Google and Wiz, we got several others there. There’s some EU regulation updates. We’re going to try to get to that as well. And then of course, I’ll surprise you guys a little bit on the flip. I’m not going to tell you what it is right now, but you won’t be surprised when we actually get there. Anyway, Pat, let’s just dive right into the big topic of the week. You know, we joke. The godfather of AI. AI Jesus. He took the stage. Let’s start with the themes. Give me the quick rundown of the big themes you took away from. What was it, like a 14 hour long keynote? It was like the Joe Rogan keynote podcast or whatever.

Patrick Moorhead: Yeah, I think this was one of Jensen’s better keynotes, to be honest. I mean there were literally no surprises. I did a social media post saying what I thought was going to be announced and pretty much that’s everything that was announced and kind of the way it was described. What I’m going to wait for in the market section is I’m going to look through the lens, not for markets, what the market wanted to hear, but, but what its customers and its ecosystem wanted to hear. And that was again, this all focuses on downstream benefits, right? All this infrastructure you can play in. But if consumer companies and, and companies that support businesses can’t monetize that and make incremental profit dollars, then what’s the point? I guess you could say share shift when you look at some of the consumer services and stuff like that. But, that’s what this is all about. And interestingly enough, this is exactly what Jensen hit on, right? It was less.

Daniel Newman: Well, don’t get me wrong, there was a lot of tech and science fare, but there was a lot of business, right? It came down to revenue per token. Right. And what’s the efficiency and the money that you can make on that? And I think he did a pretty good job on the hyperscaler data center point. He really brought it with a lot of additions, like a new data platform for the enterprise and you know, robots were all over the place. And I think he made a pretty good case for that. The partner reaction was any reaction, any gauge, then I think he hit it. One final thing on the top line. Like if I look at the messages that Jensen was trying to get across and I think he did a much better job in the Q & A with us than he did in the keynote itself. First of all, Nvidia is in the infrastructure business, right? It’s not a product company. Right. And he reiterated, hey, so I laid out a four year roadmap that all my competitors know what I’m going to do. And his thought was, hey, when you’re an infrastructure company and you have an ecosystem that has to build with you have to do this. And while he said like no other technology company ever has, I can dispute that there’s a lot of enterprise companies that give long term roadmaps. Second, his comment was that Nvidia is an AI factory, right. And it’s all about revenue. And as for the hey, do we skip Grace Blackwell and just, just wait for Vera Rubin? He’s like hey, quote unquote. And nobody says, guess what? I think I’m going to save my revenues for three years from now. Final point on what he wants Nvidia to be. And that’s a foundational company. And to me, I view that as foundational tech. And he said every company, every industry is at this show and therefore we can’t have surprises. Getting back to his earlier comment of being an infrastructure company. So those are my, you know, kind of initial, kind of high level thoughts of this didn’t move the markets, but we’ll talk about that later.

Daniel Newman: Yeah, my thematics don’t really have a lot of changes from the CES keynote. I think his mission is the same. I think his roadmap is progressing. I think there’s this very interesting sort of decoupling that’s going on from Nvidia and its biggest customers. And we’ll talk about that maybe even in the flip we’ll get to that a little bit more. But like I said, I really saw, you know, I’ll kind of hit four real quick high level theme bullets that I saw. One is, you know, very, very important to him to get beyond the data center to make sure it was very clear when I say beyond the hyperscale data center so you saw move into the private cloud. That was a really big thing for him. Two was painting a picture of what the next opportunities are then outside of even data center, enterprise, private cloud. And that was really leaning back into, you know, the autonomous vehicles, the GM win, robotics, humanoid, and painting the picture how big this is. Third was sort of tokenomics in terms of the DeepSeek moment. You know, he really did slam the stupidity of the media and the stupidity of a lot of the pundits that just by the way, I can’t fathom how long this DeepSeek thing has lived incorrectly with so many smart people out there trying to correct it. But I think he did a good job of sort of talking about the scale of compute and the fact that as more and more people use this stuff, like even DeepSeek, as it got used more, it got slower because it needed more compute. And then the final thing was, you know, I really do think that he wanted to make sure the tech leadership was clear. So even if there are people coming up behind him, he wanted to make sure that the four year roadmap showed that they’re so far ahead of this stuff that’s going to come in behind it that that’s fine. They can be the second, third and fourth to the plate, but they’re not, they’re not even in the game when it comes to the kind of full stack AI software, hardware frameworks, libraries, development and so Nvidia is thinking ahead while everyone else is just chasing them. That was kind of the theme I really saw here. Let’s jump in a little bit though because you know, you heard me kind of talk about that enterprise theme. You know, Jensen was running around the floor again doing the Dell thing, Dell, Dell, Dell, Dell. Michael shared that. And then he was over at HPE doing HPE, HPE, HPE and then I think he ran over to Cisco, saw a good picture of him over at Cisco. You know he is the ultimate cheerleader but you know he’s got his whole keynote wave passed quickly on the hyperscalers. And then he talked about CoreWeave and he talked about Cisco and he talked about HPE and Dell. What are your thoughts on that kind of framing?

Patrick Moorhead: Jensen did not mention any hyperscaler by name. And I think there’s something there Daniel.

Daniel Newman: 300 billion dude of CAPEX and not even a, not even a show.

Patrick Moorhead: Yeah, but who, who did get rolled out here? It was Dell, it was Cisco. HPE, Lenovo, just like you said. And this ties into this again, this theme that I’m going to take a little mini victory lap here. I said early on that enterprise AI was not going to be dominated by the hyperscalers for the very reason that cloud, we’re 15 years into the public cloud and 80% of our data is still, is still on prem. Right? So this notion of hybrid AI cloud is a reality. And who will enable this? The Dells, the Cisco’s, the HPEs and the Lenovo’s. It’s interesting. Two thoughts here. CEOs of every one of these companies were there inactive with the exception of Lenovo with YY. I’m sure he had a good reason for not being there. Maybe it was spring break as well.

Daniel Newman: There he is! On a boat!

Patrick Moorhead: Cisco has made a real resurgence, a surge into AI that is quite impressive. You know Dan, you and I got the download with Chuck a MWC and their executive staff. But it is impressive to see them jump on this. Quite frankly Nvidia had to go to Cisco to get help for enterprise networking. Like nobody wants to stand up two different flavors of front end networking. They just don’t. And even backend networking. So that’s really good to see. And by the way, I was pretty excited. I was actually quoted in the Cisco press release with Jensen and Chuck about what they did there. So that was kind of fun. Now in an editorial statement, I hate to even, you know, say who I think is closest to Nvidia right now. It’s Dell. It seems to be a Dell love fest. Last year it was a Dell love fest, this year it was a Dell love fest. And I think first and foremost it’s end to end, right? It’s PCs to PCs to data center infrastructure and the infrastructure, Enterprise Edge. It’s storage, it’s servers, it’s PCs. And obviously you know that founder led, right? And the scale that Dell operates at, you know, Dell made this announcement that said that they had 2000 customers already within a year from them announcing their Nvidia AI stack. So yeah, no hyperscaler was mentioned by name on stage.

Daniel Newman: Yeah and we’ll talk more about at least I have an opinion on that. But the interesting thing too though is kind of you called out networking, Cisco. We’re going to see the infrastructure approach to networking all this AI especially as consumption grows, enterprises buy in data sort of lives closer to compute. Companies are trying to figure out how to implement AI and deal with kind of legacy and technical debt and on prem. We’ve kind of forgotten. And you just did the cloud thing. But co-packaged optics was something you and I both kind of did in our preamble. We said this will be big. We’ve seen more memory moving close to the compute Pat. Now we’re seeing, you know, a lot of conversations about bringing more capacity for networking in the system. Did you get enough there? Did you think we got enough? Did you hear enough? Did what you expected?

Patrick Moorhead: Yeah. So it’s interesting, I wasn’t just reading Morgan Stanley reports, that I know were calling that. I had conversations with no less than 12 people on this who were asking hey what should we do? Is this good, is this bad? And, and stuff like that. It’s funny I had said that CPO was for Rubin and then the press release indicates that no, this is for Blackwell but Jensen slide says CPO is for Rubin. So trying to figure out what gives here. But I think the company did a fine job articulating the benefits of it versus copper. And, by the way just for the layperson,right, you’ve got the scale up network that directly connects GPUs, you have the back end network which connects GPU clusters and you’ve got the front end network that presents all of the AI goodness to the rest of the enterprise, right. Connect it to all the different, you know, run AI applications. Then you’ve got the segmentation of scale up versus scale out as well. And scale out is basically the front end of the backend network combined. So I think it did a good job articulating the benefits versus copper which is literally saving megawatts of power. Right. I thought that was pretty good. And then they made a secondary case for the power savings and the density improvement you can get from doing co packaged optics to the switch. And you know there’s a panic in the run up to this that oh my gosh, anybody who’s involved in co-packaged optic pluggable modules is going out of business. Well, I think Jensen did a really good job, you know, talking about you know, the coherence, the cornings, the Lumentums of the of the world, even Foxconn and TSMC who are participating with this, that you know, you need lasers, you need transceivers, you need prism lens assemblies, everything else. And you know what wasn’t discussed though Daniel, unsurprisingly is hey, what are the potential downside of doing CPO? Right. Hey, your CPO fails, does your entire switch fail? You have to replace that with a motherboard replacement. What kind of modularity is in there? And also what was discussed in the research was CPO to the GPU. That I thought was fascinating. You know, we’ve heard from companies like Marvell who did a January announcement on CPO to the XPU where you know, you essentially it seems more, more reliable but also a lot quicker and higher bandwidth. So you know, our team and your team, like Matt Kimball for scale up and Will Townsend for scale out is going to be covering this.

Daniel Newman: Yeah, it’s interesting Pat. This is kind of the next. First we brought memory, now we’re bringing the networking closer to the system and it’s basically, it brings efficiency and it definitely, definitely enables scale up at scale up at scale. Right is how I’d like to say it. But it’s going to bring a lot of optimizations on energy and efficiency and connectivity. Again, I still think out of the rack and they’re going to see a lot more copper, but I think in the rack there’s just a lot of arguments that the optics just delivers greater efficiency. Let’s sort of, you know, I guess the last thing on this topic because this is kind of an extended special edition. We got a lot of other stuff to get to, but it is the roadmap you, you dug into the four years ahead. You know, we’ve got Rubin, Feynman, we’ve got Blackwell Ultra, we’ve got this kind of GB200 might get retired or early due to the complexity. GB300. And of course Dell’s first out as always. Kind of. Your point is, I just think Dell is the all in on Nvidia company more than anyone else. Not that others like HP aren’t, but they just seem to jump first. And I think that Jensen really appreciates that about them and you’re seeing that in his sort of cheerleading that you, you, you know, the, the Dell, Dell, Dell. I mean, God, I mean Michael just has to be just thrilled. But you know, the ability to sort of post this roadmap. You know, historically speaking, I mean Pat Gelsinger did a roadmap like this and everybody puked on it, you know.

So Jensen is jumping ahead, kind of showing how much more he’s going, you know, from HBM3 to HBM4. He’s showing again the co-packaged optics, the bigger dies, the, you know, more GPUs per. I mean eventually we’re gonna have a box this big, Pat, with like 800 GPUs in it. I mean, it’s so cool how fast this stuff is moving. Having said that, you know, what could go wrong? You know, in my opinion, the biggest thing is you saw with Blackwell, the complexity from Hopper to Blackwell is so exponential that he’s had problems that they just didn’t have in the previous generations. And this is something that happens to semiconductor companies. This is something that’s happened to others like Intel, where you do get that aggressive and then mistakes get made. Now, the world kind of has seen this perfection of Nvidia over the last several years, and now everyone’s just like, there’s no mistakes. But then Blackwell sort of exposed a little bit. It got a little squishy. And by the way, NVIDIA doesn’t do everything. You got foundry partners, you got, you know, your Synopsys, you got your IP partners, you got compute CPU being brought in. You know, you got all the cores, the tensor cores being brought, you know, and it’s not all just Nvidia. It’s a big ecosystem, as you see in these semiconductor slides. You know, my opinion, Pat, is the biggest risk of putting out this timeline is the market’s formulating its valuation of the company. Companies are doing their planning and their road maps, and one year from generation to generation is incredibly aggressive, you know, and I guess he kind of gave himself 18 months from Rubin to Feynman, but I mean, that’s still really, really tight. I give him a ton of credit for being so aggressive, a ton of credit for being so open with the roadmap, bringing the ecosystem along with them, Pat. Proving that they have technology superiority. But what do you think could go wrong with having so much out there and such an aggressive roadmap?

Patrick Moorhead: He’s tripling his rack power like everything can go wrong. But I think he’s really trying to get ahead of it. Right. In a very public way. And, you know, whether it’s Schneider, whether it’s, you know, anybody who does cooling, anybody who does. Vertiv. Yeah. Who I had never heard of until the past couple years, right? Getting them aligned on building for the next generation. Boy, Daniel, wouldn’t it be smart if they built this generation’s data centers to tackle 2028 problems, right? Notice for Feynman, there was no, you know, there was no platform defined, right? So who knows? Maybe that’s a gigawatt per gigawatt. Yeah, kilowatts. Gigawatt. So we just don’t know. And as much as I know, I like to say that, you know, this is the Nvidia show and they’re gonna pull like everybody with and they’re going to be the only one making money. If you are a core provider in the Nvidia ecosystem, you need to get on board and you need to make money. And there’s probably 500 different variables that are required to pull this thing out. Quite frankly, some of them have not even been invented yet. Right. Every one of these power solutions from Vertiv are essentially custom. Right. Because it’s a constantly moving target. This notion of having a standard for a rack is a joke. That used to be the case for standard workloads, but today it’s off the table. We are installing Grace Blackwell into our Signal 65 data center. And we have to consider now, I was talking to Ryan about, you know, he’s going to have to have a whole new power system for Blackwell Ultra and Rubin, that stuff’s not cheap.

Daniel Newman: Yeah, it’s interesting how that’s proliferating. You know, we kind of went back and forth. Small nuclear reactors and we heard about that for a while. Then it got kind of quiet. And I think this has more to do with sort of the market’s exuberance around it. But yeah, the complexity here, Pat, like a lot of what, you know, we went and did some of the tours at Dell. This was non NDA stuff, stuff we could share. They were talking to us about how basically all the design work they’re doing for rack cooling isn’t going to work from this generation to Rubin. So everything that’s been designed to make this generation work will have to be done again. Lots of CapX Pat, which is good for spending, but it’s not necessarily, you know, as scalable as the previous generation of computing. It was a lot more interchangeable, at least in the last several generations of it. All right, we get a lot of Nvidia. We’re going to come back to some things here. There was some other big news though this week, Pat. I’m guessing you were asked about it at least a few of your interviews. I came out and basically said 90% chance this Google Wiz deal does not happen.

Patrick Moorhead: That’s strong, man. That’s very, very strong. Like, why do you think that?

Daniel Newman: And I was super. Yeah. And, trust me, I heard about it a little bit. But you know, I was looking at the new administration and first of all, there’s kind of this belief that Trump is friends with everybody in big tech. Yes, all the big tech CEOs came down to Mar a Lago and kissed the ring. But you remember Trump’s rhetoric. He was not happy with these companies. The sort of ad mafia that, you know, did all the blocking of pandemic related stories. Various news about him and you know, over the last few years. And Trump’s a guy that holds a grudge for some time now. Having said that, it’s not just him, it’s the new appointees to the FTC who have kind of said that, you know, they’re going to look into this thing. And I want to say before I say why I don’t think it’s going to happen, I actually love the deal. After, our conversation with the founder of Palo Alto and how he talked about the cloud being where security takes place. Multi cloud is sort of the future. Having a multi cloud solution like Wiz within the Google cloud portfolio, super powerful play. Having said that, it’s a 30 billion plus dollar deal. So one of the largest deals right now you got Juniper trying to tie up its deal with HPE. It’s stuck in regulation. That deal to me is less invasive to companies that are not a massive global threat. As like, you know, you don’t see HPE getting sued by every regulatory body around the world for its practices like Google, literally.

I think it’s another topic on the docket. We’re going to talk about more digital regulations against Google now. Again, a great company brings great innovation. We know Europe is the ultimate speed bump toll of the world. They love to sue and find big tech companies from the US as a way to bring revenue in as opposed to driving their own innovation. And I’m steadfast about that. But that means again, this deal getting done. First of all, can it get through the US right now with a sort of bitter Trump that’s still a little bit upset about how he was treated by all these big tech ad companies. Second, Google’s already being looked at under a number of different regulatories. They’re trying to break the company up, Pat. I mean, you know, they’re going to split them up and let them do $30 billion deals in some of the biggest, most exciting security companies. So that’s out there. Most of the world has various cases against Google right now, including Europe. They’re all over Europe. China is going to be complicated. If they need China. I mean, Google’s banned in most of China, so I don’t know if they’re going to need it. I just think the regulatory environment is really tricky right now and I haven’t seen deals of this size get through. So while I like the technological aspects of the deal, I just think it’s a big aggressive try right now. And this could be another kind of Figma one that, or an ARM Nvidia one that we spent a lot of time talking about and there’s a lot of reasons to like the deal but just can’t get done. I don’t know. Are you the opposite? Do you think I’m too bearish here?

Patrick Moorhead: So I just want to hit before I dive in that, just a quick, quick blurb here. You know, Daniel, I’ve been talking about hyperscale, excuse me, hybrid multi cloud fabrics for a long time and the idea of having security as a hybrid multi cloud fabric is not new. Cisco has it where they can secure your data, whether it’s on-prem or in the public cloud. And they have network fabric as well. So this is just that and it’s a positive thing. Enterprises are sick of buying oh, a solution I can only run for the public cloud enterprise SaaS or I can only run it on-prem with this. So I really like it from that point of view. Trying to guess what will happen with regulation is not easy. I’m not going to take the shot on this one. But I will give you another point of view which is if, if you look at the way that most antitrust laws are written, right. It cannot put you into a monopoly position. And if I look at market share of what this would provide to Google and when I think of the Wiz, the Wiz is being combined with Google Cloud, which is the number three player. And then if you look at, you know, their market share in the security market, you know, I don’t think it breaks any of these rules. But, with that said, there’s a potential challenge there because it’s Google and what it’s doing on the consumer side. But I think if regulators look at it as a Google Cloud type of entity, it’s going to be a lot easier for them to get it across the line.

Daniel Newman: Yeah, I mean, maybe if they split the company and then do the deal. I mean, I’m being a little, you know, provocative here because I don’t want them to split the company. I think it’ll break the experiences that are created across Google that have a lot of value to people. But I mean, just as a note, like even just this week, Pat, and you know, I won’t spend a lot of time, we won’t spend a lot of time on this. But you know, Google was hit with two charges of breaching landmark EU rules just this week. They’re ordered to, you know what they say. And then Apple was as well. And again, same thing. But Google could be fined 10% of annual sales by Europe. And again, we know that’s not going to happen. The EU is going to take a couple billion dollars off Google and then move on. It’s like a speed bump process and they do this every so often. But remember, this kind of deal has to get through regulations everywhere. It’s not just in the U.S. so you’ve got a complex environment. I agree with you. But I also don’t think Juniper and HPE break any true monopoly laws either.

Patrick Moorhead: Listen, I think the HPE Juniper thing, I think their competitors got the jump on them. Right?

Daniel Newman: Sure.

Patrick Moorhead: I think that’s, that’s what happened. Very similarly, what happened with the VMware Broadcom.

Daniel Newman: Yeah, no, I mean there’s, there’s a lot of politics that happen within these. And to your point, why you maybe don’t want to take the shot. I’m being a little bit aggressive here, but I just think people are overestimating the Trump administration’s willingness to support Big Tech M&A. And so I hope they, in a lot of ways, like, I hope I’m wrong. Like, you know, in most cases I kind of want to be right so I can do the victory laps. But in this case, I actually want to see Big Tech M&A because Apple desperately needs to acquire some sort of AI company because they cannot stop sucking at it. By the way, this isn’t really a topic for the show, but I just want to call it out on the air. I found it hilarious that Apple took its chief of its Vision Pro product to put on top of its Siri products. So they basically took it. In the spirit of trying to show that they understand how bad their AI products are and see how bad Siri is, they take a guy that’s running another part of the portfolio that’s also wildly underperforming and give Siri to that person. So now they can have two products suck under one person.I can’t get my head around this, Pat. I know this isn’t the topic of the day, but good lord. I mean, any thoughts on that? Did you read about it?

Patrick Moorhead: No, I did, First thing I thought was Vision Pro is a loser. And AI is a loser. You know, you put two together, what are you going to get? Now I’ve had to convince myself it was an important thing. You know, Vision Pro was a very difficult, multivariate, multi-year problem that needed to be solved with a lot of new inventions. And so if I had to talk myself into it, then, you know, we could, we could do that.

Daniel Newman: So we’ve got to keep moving. There’s so much decode to cover. We’re gonna get back to Nvidia. Okay, I’m not gonna say Nvidia anymore. If I keep saying it wrong on purpose, eventually I’m gonna go on here and say it wrong on accident, and then I’m gonna embarrass myself like I embarrassed the CES CEA guy for saying it wrong 15 times at the CES keynote. Let’s get to the flip. Will Nvidia continue to dominate AI chips in the face of Amazon, Google, Microsoft, Meta, ByteDance, OpenAI, Broadcom, Marvell? Everybody seems to be doing or helping arm supporting a custom chip. How is this gonna play out, Pat? Will Nvidia continue to dominate or will they run into real competition? Let’s flip a coin and see which one you get, Pat.

Patrick Moorhead: Oh, my gosh.

Daniel Newman: You got the four, man. You gotta, you’re gonna have to make the argument that this is not really competition.

Patrick Moorhead: Dan. I mean, first of all, historically, Nvidia really hasn’t had any competition from anybody on this topic. You know, I know we talk a lot about A6, we talk about AMD, but what we’ve squeaked out what, 5% market share out of that, and Nvidia has 90, 95% of that. A lot, a lot of reasons for that. There’s the hardware, which I know everybody talks about. First of all, it’s not even a GPU, right? It’s a programmable piece of silicon, right? There’s no rasterizers on there for graphics. There’s nothing you can do to make it a GPU. So I think even the moniker that this thing, you know, is a graphics card is just wrong. And then, you know, secondly, if I look at the hardware, it’s not just the chip, right? It’s not just bringing up a GPGPU chip and calling it a day, right? It needs a full system, it’s a full rack approach, which I think pretty much everybody agreed to at the conference. I haven’t heard anybody say that this is not a rack scale challenge at this point. And you know, if, if it takes a village to pull every bit together. I remember when OpenCL was going to decimate Cuda because it was an industry standard. And the problem is when, when the Nvidia gear is operating at 10 times the speed of an industry standard, people just, they just blow across it. And then, you know, if you look at the, what this could mean to hardware in the future, look at the, you know, we talk about this combined R and D. If you look at Nvidia, the cash that they will be able to invest in future R and D into the hardware will likely surpass everybody who’s doing A6 and, and also AMD. And I didn’t finish my thought on the platform piece, but you have to do the compute, which is the CPU, the GPU, scale up network and scale out network to pull this thing off.

But hey, let’s go into the software and it’s called Cuda, right? People have been trying to get out from under CUDA for a long time. And the hyperscalers for, let’s say a managed service and the hyperscalers for their own enterprise SaaS have been able to not use CUDA. I wouldn’t even say effectively because let’s not forget that Nvidia still has 95% share out there. But if you look at the enterprise, where it’s not just Cuda, right? Where Nvidia is standing up an entire enterprise AI stack with models, with frameworks and just look at the companies who have, who have gotten behind NIMs. Pretty much every enterprise SaaS company has gotten behind NIMs. All of the great companies, right, like ServiceNow, SAP and Salesforce and Adobe, they all have NIMs. And guess what? You can’t run a NIM on anything but Nvidia hardware. So yeah, Dan, Nvidia will continue to dominate AI chips for the foreseeable future until Intel screws up, right with execution. Kind of like, you know, pulls an Intel of, of some sort.

Daniel Newman: Yeah. So here’s why you’re wrong, Pat. That was everything and that was how everything was. But when you take, you know, six to seven companies that have trillions of dollars in assets and you basically erode their margins and you tell them that they have to buy and you take away all their, their power of control, you, you risk their long term prosperity because you start to tell them how they can operate their business, inevitably you’re going to start to see some whiplash. We already saw Apple kind of say no from the beginning. Now if their AI didn’t suck so bad, they would be a great testimony for what this could be. But what you have heard is a lineup now of CEOs getting up on stage. Whether it’s been Sam Altman, whether it’s been Sundar Pichai, whether it’s been Andy Jassy and all of them now are touting their own wares of building their own GPUs. Quote unquote TPUS quote unquote XPUS. But here’s how it’s actually going to break down. And the good news is it’s not zero sum. And the good news is it doesn’t mean Nvidia just goes away. But what it does mean is this is every one of these mega hyperscale cloud providers has internal workloads and external workloads. Their internal workloads are things like search or generative text or recommendation engines. And all these companies are going to increasingly and more rapidly move to using their own in house silicon because they can do it much cheaper. They can build their full stack. They have the software, programming and engineering in house. They can abstract away from Cuda, build their own libraries and get more control over their world.

The thing about that though is that’s a large percentage of the total consumption of AI. You’re thinking Meta, all their, all their social platforms, you’re thinking about Google and all their search platforms, you’re thinking about, you know, potentially down the line it could be Microsoft and all of its tools for office and productivity. That’s where I think it is heading. Having said that, you know where it lands. You got a trillion dollars of AI chips by the end of the decade according to Morgan Stanley. You got half a trillion according to Lisa Su. And I bet you it lands somewhere in between. What I end up seeing happening here is you see something like the 95% go down to like 70% so the good news is, and this is probably the worst part of my flip, because I’m not, I’m kind of arguing your side just a little here Pat, is you will see their dominance erode as a percentage of market share. But as we head to a trillion, I still think there’s a whole lot of upside for everybody in this particular space, Pat. Broadcom certainly is going to be the most disruptive. I think ARM is going to come in and get more and more in the way and become competition to Nvidia over time. And if Marvell can get generation to generation success, they are in the game too. But these hyperscalers, they don’t want anything to do with giving all their margin away, all their control away and being completely beholden to limited capacity and a company that is going to be working aggressively to kind of control the distribution of its market leading innovations. They want to compete. So therefore their dominance is going to be up against the other biggest companies in the world who are all gunning after them. It was clear this week at GTC, even in Jensen’s lack of mentioning these companies. I think the, the, the bromance, the alignments of these companies, it’s coming to an end. There’s a codependence, but it’s not out of love, it’s out of necessity. All right, so let’s go on to our last segment here. Bulls and bears. Pat, you know we talked about this a little bit earlier, but I just want to get your quick take. We’re going to do these a little faster for everybody out there. Don’t you love when we ramble on? We’re so entertaining.

Patrick Moorhead: Are you not entertained?

Daniel Newman: You’re not entertained? We should get that one as a clip and then turn it into a gif and then we can laugh at myself especially because I’m looking so good today in my glasses. The Wiz agreement. I kind of alluded to this but like just as a high level. Pat, does this show and, should this 30 plus billion dollar deal spur more excitement within PE and VCS? I mean it’s been a quiet period of time. Is this what they needed to hear to get fired up again to start pouring cash in? I mean AI money’s been going in but does it change anything?

Patrick Moorhead: It’s unclear. Daniel, I think you nailed it in another conversation where you said it’s just, you know, we’re not too sure where the Trump administration sits. I know all the VCs got behind Trump or most of them got behind Trump for the election thinking that he would be this big catalyst for not only GDP growth, but also to loosen the strings on M&A. And, and you know, we haven’t seen that yet. I mean, let’s not forget too, that, you know, VMware was approved under even the prior administration. Right. So I think it just showed that, you know, even if you have, you know, a, you know, hating big tech, FTC and Justice Department, you can still get something through. I think it’s too early to call, but maybe this is the litmus test, Daniel. But my gosh, these things take so much time to get through. But we will see the moves, I think some of the early moves, right. We see, hey, we’re going to investigate you and then, oh, we’re going to sue you to slow this thing down. I think this will be the, the ultimate, one of the arbiters of that. I think the other thing that could help is we see an IPO go well. Right?

Daniel Newman: Yeah, there haven’t been a lot of them. Reddit went pretty well, went way up, kind of came back down. Yeah, I’m with you. I mean, look, the money is pouring in. VCs, they have to, you know, PE and VCS have to put money to work. I mean, that’s the good thing for companies, builders. But obviously when multiples get pulled down, when transactions aren’t happening, when second, third, fourth series are not, you know, ABCDZ’s and F’s and things aren’t turning to IPOs or acquisitions, it starts to create risk in terms of bringing funding into these things. So, yeah, it’s definitely going to be something to watch, Pat. You know, another thing is, you know, I know you went on Bloomberg, talked MAG7, talked Nvidia, you know, it seems to be going down, down, down. I mean, it’s been a rough, I guess, you know, who would expect Trump to start a tariff war and, you know, get us into a fight with every country in the world and try to, you know, bring down inflation in a completely atypical way, which we’ve, we’ve talked about with tariffs on the show, Pat. But like, you know, the Mag 7, are we at a value point? Is this, is this like the time to buy or, you know, I know you’re not a stock picker, but like, what the heck.

Patrick Moorhead: Yeah, so I, I just, I just got off Bloomberg, I did a 7am hit and, you know, I’m going to give you the same story which is, you know, there’s two things going on with Mag 7. The first thing is this overall risk assessment. Interest rates, recession, GDP, consumer confidence, business confidence, tariffs. People aren’t too sure what to do. But I think the more relevant one is I think it’s calling into question the downstream benefits of AI. Right. You know, how do consumer companies make more money, how do businesses make more money with AI and not just protect share and therefore we’re seeing Mag 7 decline and I don’t think Nvidia gave us any more confidence in the future about what the downstream impacts of AI can be. I mean all, all I heard was cheaper cost for tokens with, with Blackwell and, and, and Rubin and then I think all of the, you know, AI edge and robotic stuff that’s a, that’s a five year, that’s a five year poll and I don’t think the market reacts to anything outside 18 to 24 month window. So people are just sitting around. There is a lot of money coming in into equities, interestingly enough, which I’m trying to piece through a Bank of America report that I was looking at that talked about the flow of money into equities is so high, you can imagine the drop of equities if all that money wasn’t coming in. People, mostly wealthy people, are putting money into equities right now.

Daniel Newman: Yeah, I mean look, I, I’m, you know, I’m pounding the table on some of these like Google, I mean it’s just retracted to its lowest forward multiple in a long, long time. And despite my concerns about its ability to get deals done, the company has incredible moats. Great growth, great margins, great business. I worry about Tesla. I mean I love what Elon’s doing and I know it’s unpopular, but I, I don’t love everything about the way and that’s like the thing I always feel like I have to, I have to like contextualize or, or put, you know, put a bit, a bit more insight around that because I don’t love the way things get done. But look, when you run a business and you look at how waste is being managed and you know, I like the fact that he’s trying to suss it out. You know, by the way, I don’t know if you saw X, got a $44 billion valuation this week. It was in the Financial Times, remember everybody was celebrating that he was losing like 30 billion, 35 billion.

Patrick Moorhead: Yeah.

Daniel Newman: It’s back. How did he bring it back? He brought it back by basically cutting a ton of costs out of the business. You know, he made it efficient and he ran it and he grew it and now it’s become probably arguably the world’s most important source of information in terms of real time fire hose across the political spectrum. Okay, but having said that, like Tesla, I worry about because basically right now, I mean, half the country just absolutely, I mean, they’re lighting cars and dealerships on fire because they just don’t agree with the politics. And so I don’t condone any of that stuff. But it’s just a signal to me of how far. And obviously if you were someone that was big on green and sustainable, that bought your Tesla under the onus of sort of the green potential of having an electric vehicle, God, you must be in a really tough situation right now, kind of floating in those two head spaces like, you know, what happened? I saw one of the politicians sell his Tesla and bought a Yukon or a Tahoe or something. And I’m thinking you just went from a, you know, a very efficient electric car.

But Pat, across the spectrum, I mean, look, these stocks are super defensive. I know, like over time, Nvidia is not playing out that way. Apple’s very defensive. You know, these companies can grow. They have so much operating leverage that they can grow through the most difficult economic period. So even if we go into recession, it’s really hard to bet against these companies because they won’t stop growing. They just grow a little slower. But you’re talking about revenue growth, margin growth, expansion. And of course, these companies are all implementing AI at a big scale and they’re able to take a ton of cost out. They won’t do it like Tesla. They won’t do it like Musk. That’s just a different style. They won’t be like Hawk. You could say Hawk. Like, there are certain leaders. And by the way, Lip-Bu is looking like he’s going to be one of those kinds of leaders over at Intel, but they can definitely make more efficiency, grow these businesses and stay safe. I like these companies here and now. All right, one last quick one before we shut down the show. Pat, I just want to get your take on the Ampere deal. I mean, this one was. It wasn’t out of left field, but you start to see how all the cards play out here. It’s basically felt to me like did ARM just acquire Ampere, you know, through this weird SoftBank thing. But they left themselves some independence by saying it’s not ARM, it’s Ampere. But we all kind of know ARM’s going to get into design. Like what’s your take on that one?

Patrick Moorhead: Gosh, Daniel, I got so much to say on this one.

Daniel Newman: Let’s keep it to like two minutes because we both got to go, buddy.

Patrick Moorhead: ARM is in the business of soft ip. Ampere is in the business of data center ARM based chips. Did you also see the SoftBank acquisition of Graph Core? Like SoftBank accumulating all of these capabilities to go after the freaking brass ring which is CPU, accelerators, IP and chips. Imagine if this was an OpenAI Stargate play. But on the other side of my mouth I’ll be like none of this stuff takes. None of this stuff is quick, right? If you weren’t working on a stargate accelerator already for open AI. Like how, how is that going to work? Like how many years is SoftBank gonna have to work on that? But, Masayoshi Son is an absolute genius in his success and in his failures. And I think this is going to be really big. And it could be one of these change the industry type moments for, for SoftBank. We don’t know exactly what they’re doing but I’ve got about five scenarios and what I think they could be doing or what I would do if I were him.

Daniel Newman: Absolutely. It’s a Stargate deal. To me all day long there is. This is just one more player which just makes my flip even a little bit more right. They are going after it right now. Nobody wants to be left out of this trillion dollar market opportunity. I don’t care how you slice it up. Half a trillion. A trillion. Remember right now Nvidia with all of its prowess. 95 it’s 100 billion expected this year. There’s another 5 to $900 billion annual revenue tied up to these accelerators right now. Nobody is sitting on the sideline. None of these companies want to sit here, be beholden and just let their margins erode. And, and all be beholden to a single player.

Patrick Moorhead: By the way, one, one, one thing supporting again. Remember the flip is not exactly what.

Patrick Moorhead: But it’s like open AI has one app. Okay. They have one app and I will, bet money that they did not write to Cuda. They write, to lower level types of things.

Daniel Newman: Jabs PYTORCH.

Patrick Moorhead: There’s no way that they’re using CUDA. There’s no way they’re using CUDA libraries or CUDA models. They make the model folks. So, yeah, could, you know, make the case. But, where, SoftBank IP for accelerators, where is it? Who’s creating it?

Daniel Newman: I guess they’re gonna have to put that on their list of things to acquire.

Patrick Moorhead: I mean, they did acquire Graph Core, but I mean, I don’t know. I just, I, I don’t know. And I know that open AI doesn’t have enough people to do what Google has done with TPU. That’s on, on the, you know, on their seventh generation, by the way.

Daniel Newman: That’s why I’m extremely steadfast that the external workloads, the ISVs, the partners, the enterprise data center companies, the big enterprises, and that’s why the big cloud companies will have a big Nvidia lot of capacity because like I said, for internal workloads, it makes a ton of sense. And that’s where market share will erode. But GCP, AWS, all these companies, they just can’t move fast enough and they can’t offer enough software support to enable these developers now in time. With AI, could that change? Absolutely. I just don’t see it changing fast. So, you know, remember, everybody, the flip is Pat and I take a side. It doesn’t mean it’s a side we believe in. So when you see the opinions out there, just remember that we may or may not actually believe what we’re saying, but on the other hand, we might. So anyways, I gotta say goodbye now, Pat. I got real work to do. This has been a lot of fun.

Patrick Moorhead: You’re on vacation! I hope you’re not going into a meeting.

Daniel Newman: I got a couple more meetings this morning and then I’m gonna go to the pool or the beach or the ocean or do whatever for a little bit and I’m gonna drown myself if I don’t come back. If I don’t come back, it’s because I came out of our revenue meeting very, very discouraged. And I just went ahead and I just let myself float into the ocean and with no real way to get home. But seriously, everybody, thanks so much. We really appreciate you tuning in, being part of this show. This is episode 254 of The Six Five. New format, new show, same hosts. We appreciate you all tuning in, but we got to go for now. Bye.

‍