The Six Five Pod | EP 296: GTC Expectations, Copilot’s AI Shift, and Apple’s Low-CapEx AI Bet

The handpicked topics for this week are:

1. GTC and the Shift to Heterogeneous Compute: NVIDIA heads into GTC with growing pressure to articulate a broader heterogeneous compute strategy. Pat & Dan discuss CPUs, GPUs, inference accelerators, and how future AI workloads will increasingly span training, pre-fill, decode, and agentic workflows.
   
   
2. Why the CPU Is Back in the AI Conversation: As agentic AI expands, CPU demand is moving back into focus. The hosts discuss why CPU-to-GPU ratios are tightening, why this matters for infrastructure planning, and how AI compute is becoming more diversified.
   
   
3. Microsoft’s Anthropic-Powered Copilot Shift: Microsoft is leaning harder into model optionality by integrating Anthropic into Copilot workflows. The bigger takeaway is not model preference alone, but Microsoft’s distribution advantage, governance layer, and ability to bundle AI functionality directly into the enterprise productivity stack.
   
   
4. The Semantic Layer and the Future of Enterprise Software: Rather than replacing core enterprise systems overnight, AI is increasingly being layered on top of existing platforms. The discussion highlights how enterprise software may evolve through AI wrappers, orchestration, and semantic interfaces rather than a complete replacement.
   
   
5. Adobe’s CEO Transition and the AI Narrative Gap: Adobe posted strong numbers, but investor skepticism remains. The conversation centers on whether Adobe failed to clearly articulate its AI upside, whether the market simply remains unconvinced, and why leadership change may reflect the need for a different kind of AI-era storytelling.
   
   
6. Grid Underutilization and the Energy Debate: Google, Tesla, and others are backing a lobbying effort focused on grid underutilization. The hosts unpack why this matters for hyperscalers, data center growth, and the broader push to use energy infrastructure more intelligently before simply adding more capacity.
   
   
7. The Flip: Has Apple Found a Way to Win Without Massive AI CapEx?
   In this week’s debate, Patrick argues Apple may have found a differentiated path by focusing on device-level inference, silicon efficiency, and distribution across its installed base. Daniel pushes back, arguing that if intelligence becomes platform agnostic, Apple risks becoming just another hardware endpoint with a limited moat at the AI layer.
   
   
8. HPE’s Networking-Led Enterprise AI Positioning: HPE’s results reinforced the strategic value of networking as a differentiation layer. The Juniper acquisition, enterprise focus, and higher-margin infrastructure strategy continue to distinguish HPE from peers chasing hyperscale AI server volumes.
   
   
9. Oracle’s OCI Momentum and the Fungibility of Compute: Oracle delivered the “show me” quarter investors wanted, driven by OCI growth and backlog expansion. The hosts argue that if Oracle builds the capacity, the compute will get used, regardless of which model provider ultimately occupies it.
   
   
10. Adobe, TSMC, and ASML as Signals of the Next AI Buildout: From Adobe’s forward-looking market pressure to TSMC’s continued growth and ASML’s move into advanced packaging, the conversation closes on what these indicators say about demand durability, capital discipline, and the future of semiconductor infrastructure.
    
    

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

IMPORTANT: Note to Uploader: Libsyn AUDIO EMBED - added by uploader.

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.

Paul Moorhead: 

Boom! Six buy pod is back. Episode 296. It's been one heck of a busy week. A lot of new stuff going on. My workout was so arduous, I couldn't even make it into the office this morning. Daniel, how are you doing, bestie? I missed you this week, 

‍

Daniel Newman:

Buddy! You didn't miss me. You're not gonna have to miss me next week, though.

‍

Paul Moorhead:

No, no, we're basically going to be all over GTC just doing that thing. It is the Super Bowl of tech. We are going to dive into that one of our topics and we're going to have the whole six five crew. They're my gosh. I think we have six. Six folks. They're great.

‍

Daniel Newman

Six folks I think we have. Yeah. Well, if something like that, I think collectively and right, the six five we're there. More insights there. Feature them their signal65. They're there's going to be a it's going to be a big week. I mean this might be the biggest tech event in the world at this point. Like it might be the biggest. Yeah I'm sorry buddy, I'm a little distracted. I'm building some agents to to do some stuff right now, so, like. 

‍

Paul Moorhead: 

Yeah, we should talk about some agents, Daniel. Like, what have you. What have you built lately that you can talk about? 

‍

Daniel Newman:

Um, well, first of all, and I haven't really publicized this yet, but I've built this really, really killer autonomous Substack strategy. And now I've got a three time a week Substack that I'm building, and I'm just refining this thing on the on the regular. This one's been really fun because it's like what I'm trying to do, and I'm using Substack as like a play toy for me. It's not my blog, it's not a Twitter where I have a huge following. It's somewhere that I'm kind of net new, and I'm really trying to figure out how to dial in my voice, like how to create something that's, you know, really can look across everything I do in the week my TV appearances, my podcasts, my social media content and can write something.

‍

And then I'm really trying to like, build it to like where I read it and I'm like, Holy shit, that sounds exactly like me. And I'm having a lot of fun with it because, like, it's not right yet. Like it's slow. Like there's parts of it where I'm like, dude, this is amazing. This sounds like exactly like me. And then there's parts of it I'm like, going back.

‍

And by the way, I'm having, like, fights with perplexity and Claude, like, I'm telling them like yesterday, I don't know if you saw my tweet where I. I started yelling at Claude.

I'm like, what is wrong? Like like, you know, and then and then the other one, I think I, you know, I think I just called something and built trash like, this is trash. And it was so funny because it apologized back. Like, apologize back to me. Like we're like, yeah, we're sorry. It is really flat and boring like the PowerPoint had built.

‍

So good. I'm having so much fun. 

‍

Paul Moorhead: 

Yeah. You know, Daniel, the best thing that I built. So I was, uh, uh, drinking coffee with a friend, and he told me you don't actually have a good app because you don't have a database, right? You know, because all the data basically gets stuck in the JSON file. Uh, so, you know, the coolest thing I built, I built a fricking SQL database hosted on AWS. 

I mean, it's it's a it's I'm pretty happy about that. I got all my content in there, all the analyst content. Uh, and then what I'm doing now is I'm changing, um, all of my apps that I built to hit that versus, uh, hitting the web. So I'm actually doing that right now. Uh, perplexity computer just came out on the iOS, so I've got it on my phone.

I've got on my iPad. I've got perplexity iPad on the right of my big computer in front of me, and a display over here on the left. So I'm surrounded by compute. But but, bestie. Um, let's jump in here. So the producers have called my topic first, even though they should call you. Uh, but let me go in and hit the topic of the week.

‍

Let's jump into the decode.

‍

All right, let's dive in. I'm calling my own number, which, you know, Daniel, I love to do. It's kind of up there with, uh, staring at myself on zoom calls. I've turned the video off. So distracting. Are you, uh. Are you there, buddy? 

‍

Daniel Newman: 

I'm still here. I'm still agenting. But you you are. You're. It's your number, dude, so don't don't let me slow you down. I'm building it over here, man. But no, I'm building for this topic. I just want to make sure all my smartness doesn't get missed, and I'm no longer able to think on my own. I just want everyone to know that I know I'm augmented. The singularity is here, buddy. 

‍

Paul Moorhead: 

Uh. It's crazy. I used to have, uh, fixed tabs that were open with all the apps, and now the first thing I go to is just my AI. It's it's absolutely incredible. This whole AI is the UI thing is real. But, hey, let's jump in. It is the Super Bowl of tech. We have GTC that we are attending out in San Jose. Um, Jensen's going to get up there. And one of the things he promised, uh, was a chip that you have never seen before. So here's what I think we're walking into.

‍

I think first and foremost, um, um, Jensen has to articulate the idea of heterogeneous computing, and then how you can do it so much more easily on Nvidia. The crazy part is I debated so many people on this the last three years. People said one GPU for everything, right? And do you remember, you know, first thing it was, hey, something for training, something for inference. And then reinforcement learning set came out is like same GPUs for the same thing. And then XP are are trash right? And then TPU uh hits a home run with Gemini 3.1. Uh, you see Amazon with training and there are million training, uh, deployment uh, out there in, in Rainier. And then people are like, well, wait a second, maybe there's something about this.

‍

And then Nvidia brings out, um, CPS puts it on the roadmap, uh, which is, um, focused on, you know. Just one element. Uh, pre fill of the, the AI, the AI workflow. Uh, and then they go and they buy grok. So if anybody thinks that heterogeneous computing uh, is, is, is on the outs like you missed you missed the boat. I'll do a victory lap on this. Uh, final thing is, uh, CPUs. Right? CPUs, uh, can't do AI on CPUs. By the way, even though Google was doing most of their AI and CPUs, uh, prior, um, to the last 2 or 3 years, um, and then they became famous again. So I think the play that that Nvidia needs to make. So first of all, they need to talk about how grok fits into the roadmap. Uh, I think Jonathan Ross already, you know, he, he has a segment there, um, um, on stage, not necessarily on the main stage. Jensen. But my expectation is they're going to lay out the roadmap of how rockets rockets integrated. A couple of ways to look at that could be a super fast solution that's super, super expensive.

‍

I think the other way they could optimize it for a lower power inference solution. I don't know what the architecture of the latest rocket ship is. And nobody does. In fact, there's only one rocket ship that's out. And I think it was out like five years ago. Uh, making up the the the the year number. Uh, don't fact check me, but imagine a single layer of of of Nvidia software that allows you to go between hard core training, pre fill decode and potentially even some super high performance options, um, out there. Uh, kind of like a, like a Cerebus, uh, type of, uh, play. Uh, we'll we'll definitely see that. And then I also think we will see this constant drumbeat of the next new thing, which is robotics and the, the, the industrial edge. So those are really my kind of my top two. I don't think we're going to see the Nvidia notebook CPU yet.

‍

I think it's I think it's too early at this point, but I am I am looking forward to that. Daniel, what are you what are you expecting? 

‍

Daniel Newman:

Yeah, I mean, you hit on a lot of things. I think the one you and I are totally aligned on. Use the word heterogeneous compute. But look, the CPU is back, baby. And, you know, you saw the meta announcement. There was a lot of focus on the CPU side of their relationship with Nvidia. Slash arm. Um, you know, people sometimes forget, but that's that partnership is still runs deep. It will be interesting to see if Nvidia talks at all. It has been doing some stuff with risk five and beyond. You know I know as it continues to diversify I don't know if that will actually come out but also Intel.

Just all the diversity there like is is Nvidia going to play its cards at all in terms of CPU diversity. Now, I don't know. Could be interesting, but one thing that we are seeing from a modeling standpoint is the volume of CPUs versus GPUs are going to get closer to 1 to 1, and that's going to be driven by the agent, uh, you know, that you mentioned. But like the agent ramp and the speed that that's happening, I can't imagine right now that after multiple years of Jensen saying that the custom chip projects will not succeed, most of them will get canceled, right? That's been his talk track for some time now. We did get some confirmation. Like, for instance, I think last week we talked about meta, and it turns out they're not canceling their projects.

‍

I think you said that, by the way. So I think there's a little tiny victory lap here we need to do because people were out there saying they're going to cancel. It's like, no they're not, but they're going to buy Google and they're going to buy because they need compute now and they don't have what they need now. And clearly with the avocado drop. They don't. But anyways, back to Nvidia. They're going to have to solve for having something that's an accelerator. Like that's very inference. Uh, efficient. Um, that can go alongside the, the, the large platforms because they're where they do get beat is on these very specific workloads being done in massive volume where you're seeing, you know, a TPU or a different custom chip delivering better inference economics. And so having a kind of a something that sits between its big GPUs and the CPU that uses the grok, uh, memory, you know, Sram technology and the heavy memory on chip, I think that's going to be part of the story. And I think everybody's, uh, kind of waiting on that. The other thing I think would be interesting is Jensen always goes down the whole path of the portfolio. I'm still super interested in humanoids and robotics and where they talk. They're I think they had 6 billion in revenue last quarter that they put into that bucket. Which, by the way, is, I think, bigger than AMD's entire datacenter business. That's not a knock, but I'm just saying, like, it seems like a small number because it's small in comparison to the overall Nvidia core, but that's a massive number.

‍

Um, and so, you know, that's a trillion plus dollar type market, uh, Tam, coming into the future in near future. But I think needs to be, you know, who's going to win there, who's going to win there, who has permission to win there. Um, and I think he's going to talk more about that. So there's a couple things that are on on my mind there for uh, for GTC, it's going to be a big week.

‍

Paul Moorhead: 

Yeah. Hey, quick fact check on myself. Uh, Jensen said a chip that will surprise the world. So …

‍

Daniel Newman: 

What did you say? Sorry. What did you say? I forgot … 

‍

Paul Moorhead:

Something out there. Uh, but but. Yeah. Jensen. Also, I think on. Is it day three added an hour and a half hour and a half for agents. That's going to be interesting. We're going to hear about the five layer cake. Jensen penned a blog or his open claw assistant did. And then finally there's a financial analyst, uh uh, Q&A on uh, on the 17th that, uh, we may or may not be invited to. Hey, let's jump in, jump into the next topic here. Uh, with all of this back and forth, remember, it was open.

‍

AI was unassailable. Uh, and then it was, oh, my gosh, uh, Gemini 3.0 is really good. TPU is amazing. And then Claude comes in, I saw something and said, hey, if you're still using open AI, you're a boomer. Uh, or if you have to ask what Claude is, you're a boomer. Uh, so so so get there. And then Microsoft comes out, uh, essentially with wave three of Copilot bundling integration with anthropic.

‍

Daniel, what what did they do and why does it matter? 

‍

Daniel Newman: 

Well, look, I mean straightforward. They're taking advantage of you. How do you say this? Nicely, Pat? Yeah. We're taking advantage of the fact that co work is what copilot should have been. And what they're doing now is they're saying, look, we're Microsoft. We can build but we can distribute. And so you want to use co work and not Copilot. Be our guest. We're going to figure out a way to make money on this. We're going to bake it in and have it be supported within our tools. And you can have the best of both worlds, enterprise CIOs and those that are saying, we're not leaving the Microsoft platform anytime soon. Microsoft Platform will continue to be the core productivity suite in the enterprise. But the tool. So instead of having to go over and use co work and then drag things back into Microsoft, just have it work inside of Microsoft. And Microsoft is basically doing what I think it does best. And that's delivers optionality.

‍

The optionality is here is here for these companies enterprises? And I have had this conversation countless times on the show. They're not moving that quickly. They're not going to be throwing away Oracle and shutting off Salesforce and getting rid of Microsoft just because new tools work. Too much security risk, too much cost commitment. By the way. You know, these contracts are mostly multi-year, so why not give them the ability to add the features they want, which is the wrapper of AI. And that's by the way, Pat, this is a little bit of the trend line I think I see with a lot of the enterprise applications is like, this is what's going to happen.

You even hear about like anthropic partnering with all the PE firms now to basically allow anthropic to be put on top of the software companies they bought, because what's happening is we're wrapping these kind of known, working, secure, you know, tools with AI that people enjoy using. It's the it's the semantic layer.

‍

Yeah. You said this all along. And like even as you and I have been absolutely blown away by the the tools that are being built. You know, I think I sent this in our little group chat last night. I said, look, I built like ten things in perplexity, and every one of them is broken by the way they, they, they fix quickly, like you go back and you say like, go find what's wrong and fix it.

But these things are like the most amazing proof of concepts that have ever been built. Um, getting them to production, I think. It's not saying never. It's just saying not yet. And so what's happening right now is we're saying, look, I want that experience, the perplexity experience. I want the anthropic experience, but I want it. And my existing tools that are already connected to all my APIs and all my databases build it that way. And Microsoft probably is the best company on the planet in terms of being fit to distribute this. 

‍

Paul Newman: 

Yeah, Microsoft talk for a while, talk for a while about diversity of models. But you always had to ask the wait is it is it really is.

‍

Is that really the case? Uh, because they're close aligned with OpenAI and this deal with, uh, anthropic basically copilot could work with anthropic. And boy, if you've if you've used the, uh, the cloud plugin for PowerPoint and Excel, it is absolutely mind blowing. And Daniel, you nailed it. I mean, I was sitting there thinking, how did Microsoft not get this with, uh, with copilot, right? Um, and and but but here we are. Here's something we didn't talk about though, Daniel. And that's E-7 is the bundling play. And E7 is essentially an enterprise license that it sells, uh, the largest enterprises. Right? Uh, paid copilot seats are around 15 million, uh, in total, up 100, 160%, 160% year on year. Uh, good usage, but right now, that's 3% of the 450 million m3, 65, uh, seat folks. So, Daniel, they're plugging this. They're essentially plugging this in, and I see it as a very similar play as. As we saw with teams. Right. Remember everybody like my gosh teams versus slack versus Google Chat. Right. And then Microsoft just they bundled it in there.

‍

Right. So essentially they're bundling that in. The other thing um, you know you talked about apps working, which is 100% true. But the other thing is the governance layer, right. That is an underappreciated moat that that Microsoft, um, has here. Uh, final comment and just I'm still doing research on here.

‍

Uh, work IQ right. Um, Microsoft, you know, I think thinks it's the data advantage, outlook team, SharePoint OneDrive, Excel. But here's the thing. I can integrate all of that information, I think, directly into Claude. Right. I can I can put one drive. I can put my my outlook. I can put SharePoint. So I'm not I'm not fully certain.

‍

Maybe it's a more efficient, but I need to go off and, uh, and do more, um, uh, do more of there. But I think, you know, I think Microsoft has simplified it. Right? Remember, it used to be a $30 per seat copilot add on, and they repackage the entire productivity stack, uh, around governance being the mode. So wow, a lot of changes.

‍

Daniel. What's going to come up next week? I can't I literally I, I can't wait to see … 

‍

Daniel Newman: 

I'm having fun. I'm having fun right now. 

‍

Paul Moorhead: 

I'm having so much fun. Do you know I feel like so I used to I used to be a PC gamer, believe it or not. Uh, and then I had kids. Right. Um, and I feel like I spend less time on social media and and most of my time on perplexity computer and Claude. Right. It's just it's I mean, it's everything. And then just when I think it can't do something, I figure a way, um, uh, to to make it happen. A good example. It's not sexy, but it's like I'm pulling in, uh, six, five media content into the our content database. But I want to make sure that your quotes don't get attributed to me, because my quotes are a lot more intelligent. They're more accurate. 

‍

Daniel Newman: 

Um, they're missing rockets. They have no rockets. They're lame. 

‍

Paul Moorhead: 

Exactly. Anyways, I figured out how to do that. And, you know, I was getting, um, I was getting, uh, rate limited. Uh, on pulling an expose. Okay. Five bucks a month. Go get the X API, pull it in, sign up, and boom. Problem fixed. Oh, that was the other thing my friend said is I didn't have enough APIs in my in my system.

‍

So I started paying money to, uh, to get, uh, quick, uh. By the way. Uh, one of the topics maybe coming up is, you know, we all fell in love with MCP, but, uh, it looks like it's falling out of favor for, uh, clicks and APIs. Anyways. Fun topic. Let's move to the next topic, Daniel.

‍

Daniel Newman: 

Did you just did you just subtly just blow by the fact that MC might be dead? Is that what you just subtly just threw that? I mean, you're just gonna drop that? 

‍

Paul Moorhead: 

That's the meme. I hope not, because they just set up an MCP server, um, for streaming. Yeah, yeah, but I don't know. What do you prefer, Daniel? Rest APIs or or MCP or CLI. What do you think? I kind of prefer letting the very, very intelligent coding agent decide what infrastructure and what tools and what connectors I use, because they're all they're they're all smarter than me. So, um, you're going to tell me your CTO for a second?

‍

No, that'd be lame. I mean, he's super smart, but, like, why would I ask him? Would I have to wait for a response? I won't wait for a response. I don't even like to wait for Claude. Like I get ticked off. You know, Dario, buy some freaking GPUs, dude. Seriously? Seriously, man. No. Uh, put some, uh, Cerebus in there. Sorry. Do something. Yeah, something. Okay. Big news. Came out yesterday. I think I was on the plane coming back from Washington, DC, and, um. Adobe CEO announces his departure. You know, 18 years, it was a legendary run. Uh, for sure. And I would say probably one of the best tenures in all of SaaS would be.

‍

I mean, Benioff is, you know, has got to be has got to be number one, right? Uh, in there. And but listen, um, the market voted that they didn't buy into Adobe's AI future. And, um, here are some here are some facts. Is that enterprises want licensed imagery. Uh, they want a workflow to create, uh, amazing stuff inside of their companies. Uh, Adobe also expanded into what I'll call a CRM, uh, customer or CDP, uh, which was a, a good, a good add on. But in the end, it was all about AI and the investors. If you look at at the SaaS apocalypse, I mean, they got hit early on. I think the one thing that the company I hope will do differently in the future.

‍

Um, which, by the way, their financials are great, right? I mean, like $26 billion, uh, of of of of IRR. Right. I mean, he, uh, he built a machine, so but they didn't clearly articulate two things. They either didn't clearly articulate the big benefits that they were getting from AI financially, or they just didn't have it. Um, well, I guess there's a third option that, uh, Wall Street just wasn't impressed. Dan, what's your take? 

‍

Daniel Newman:

Yeah. Look, I mean, you said it, right? It's the end of an era. 18 years, 100th earnings call. Earnings call. And he's stepping down on a record quarter. And yet his stock dropped 8% I think. Are we talking about this later. So I don't want to do too much of the numbers right. We have this later in the in the Golden Bear. So let's keep it on on on what's kind of happening. This is uncertainty at its finest. This is the fact that while numbers look good, growth looks good. It's an uncertain future. And I'm guessing, given the performance, he felt it was the right time to maybe bring someone in. They're going to have to bring someone in that, you know, maybe from a lab, maybe a number or two person from a for a. anthropic, a perplexity, a like I think what's going to start to happen is you're going to start to see kind of out with the old and in with the new. It's time to bring executives in that are sort of born on this AI labs frontier model era that can really articulate the fact that this can't be vibe coded. And by the way, that's the work that has to be done, because this is one of those particular areas where people are like, I don't know, um, you know, have you used Netlify yet? Um, we've been yeah. So, you know, we've been we've been building like, you know, our wireframes of on our AI platform in the future of it.

‍

And it's like, you know, you used to use Figma or you use Canva, you do these design tools to kind of wireframe stuff. Now when you're when you're doing proofs of concept, they're going beyond like even like what perplexity. If you can actually, in real time develop things connected up to API's, surface real data and in and like you're getting really close to building what looks production in the amount of time. That was a fraction of what a figma and any of these tools would use. So there's work to be done on anything that was kind of like building for that kind of. So I think ultimately this is a case where the numbers don't actually suggest disruption is here. I think I said this last week, um, we're really going to have to watch the RPO because these multi-year subscriptions, like most of them, aren't going to get canceled. But like, is there a is there attrition? Does attrition start happening? But I think for uh, Shantanu and team, they just need a new face. This is like what we said about Apple a while ago. Like Apple might need a new face for the next era, but then again, maybe they've solved that as well.

‍

Paul Moorhead: 

Great stuff man. Hey, let's go into the final topic. And this is a an interesting one. I just I spent a couple of days in Washington, DC. Uh, I spoke at the, uh, Dell Federal uh, symposium, uh, went to an award ceremony for Michael Dell and the CIA. Uh, CIA has a foundation for, um, fallen CIA agents and their families. Uh, Michael was, um, got a huge award there, Patriot Award. And at the very end, he did a mic drop and nobody knew this. So, uh, the foundation had depleted $17 million in funds over the last couple of years. And Michael came in and said literally his last sentence, I'm giving $18 million to, uh, uh, to a pretty amazing, um, but yeah. 

‍

So Google, Tesla and Verus created a lobbying group on grid under utilization. Sounds pretty wild, right? The grid is falling apart. Yet this group is saying that that the grid is being underutilized. Daniel, what in the heck is going on here? It seems counter counterintuitive, but maybe it's not. 

‍

Danile Newman: 

Yeah, I thought this was one of the more underrated stories of the week. You know, it's Google Tesla Veris, which you may or may not know. They're a data center developer. Um, carrier global Renew Homes spark fund. And this utilized coalition is all about basically advocating for better electrical grid utilization. Um, you know, we're not designed for what? What is coming. You know, it was really designed the grid for this kind of brief peak demand spikes. Um, but a lot of our grid is not used, by the way, a lot of the times when we talk about grid capacity, there's kind of like we're at capacity. But at the same time, we're not using that capacity all the time. We're at capacity when these things, you know, uh, spike.

‍

So, you know, a couple things that they're up to. Right now, I think they're they're working in Virginia, which a lot of you have heard about these data center kind of capacity issues in Virginia. Um, and they're basically advocating for some of these policies to start there. So and by the way, they're not you know, they're looking at, um, you know, policies that that drive EV adoption, distributed energy, smart grid technologies and flexible demand response systems. So, you know, Google is going to need grid for its data centers. Tesla needs it for, uh, EV and it's solar. Um, you know, if you look down the chat carrier, it's HVAC, so you kind of. It's interesting because all these companies have a different benefit from it. Um, but in the end, this is all about, you know, bringing hyperscalers, uh, meeting their energy needs.

‍

And then you kind of cascade down there. Um, but in the end, it's like, hey, we got to use the grid more intelligently. We do need to add, but there is an opportunity, uh, to use all the compute all the time. Is my metaphor there, all the compute all the time. The grid could be used more efficiently all the time. So we're not just adding more capacity and continuing continuing to use it inefficiently. And I just want to give a shout out to perplexity for really helping me break that down, you know, very shortly. Um, because like, you know, I hadn't anyone that knows me knows I'm a winger, like I wing it. But this topic I had not read up on.

‍

So in real time there you are, seeing the power of perplexity and my voice being combined to create a beautiful love child.

‍

Paul Moorhead: 

I think you did a great job. The only thing I'll add is, you know, the potential alternatives could be adding batteries, right? And doing stuffs get more efficient. It's going to be interesting to see how people react to this. My guess is it's going to take a lot of education. I mean, having spent two days and I did, you know, talk to senators and congresspeople when I was there about this. And the depth of technical understanding is thin. They have, uh, staffs that that, um, try really hard, but imagine how hard it is for us to keep up with tech. And then you have to have a position that goes across, you know, the entire country and you don't want to do something.

‍

It's not going to get you reelected, I guess. All right, Daniel. I am We are done with the decode. And let's go into a into the flip. Um, this is a I really like this topic, Daniel. Um, I don't I think it still has more juice in it. Um, you know, Apple hasn't invested in CapEx at all, and they seem to be doing actually quite well. Uh, even aligning with other AI companies who are investing hundreds and billions of, of dollars in this. So the question. Has Apple found a way to succeed without investing? Let's flip that coin.

‍

How about that? The answer is yes. And I'm going to tell everybody why Apple has found and say I play right. First and foremost, Apple Silicon by measurements is the best inference hardware on the market for for on device AI. You've got M-series chips, you've got neural engines, high memory bandwidth, unified memory architecture. And don't get me wrong, everybody is is building stuff to compete with it. But Apple always seems to be one step ahead. And it's not just on on phones, it's also on on PCs. And I think perplexity computer on the Mac pretty much proves that the hardware thesis, without Apple needing to create the AI stack, and how much Apple had to do with it, I don't know, but perplexity chose the Mac first. And oh, by the way, Claude Quirk, uh, chose the Mac first as well, even though I'm using it currently on my windows computer. So, um, I think Mac OS, you know, is giving agents deep access to that localized content that a lot of the cloud assistants can't. You know, we always talk about this on device AI stuff.

‍

And the funny part is Apple really didn't lean into it on the AI side a ton. They did it first, and then they fell on their face with smartphones and they are silently building that capability. And then, hey, well, what about the cloud? I think the Apple Google Gemini partnership fills that gap well right there, partnering with Google for cloud side Siri improvements, they booted out, uh, open open AI want launched. Um, and you know, if you look at the distribution strategy, Apple's got 2 billion devices like it's the ultimate distribution, uh, advantage uh that they have. And and I think the the final the final final point here is Apple didn't have to spend $500 billion over three years to be able to develop that. They use their basically monopoly on devices to delay that. And with the lack of a lot of other manufacturers being able to get consumers excited about this stuff, this is why their stock is rocking. And and, you know, and they're not experiencing something like a I'll call it a device apocalypse.

‍

We haven't seen that yet. 

‍

Daniel Newman:

Well, now let me just tell you, I I'd love to agree with you, but then we'd both be wrong. So I'm going to disagree with you. Um, I like that what Apple just did was a secure its future as nothing more than a commodity, which is what it has been working towards. And its decision not to invest in infrastructure has all but assured its fate. You know, we know for sure that this does not mean Apple has an AI strategy. There's just more technology sitting on top of their endpoints, which may be good, but I don't know if that's that's a moat in itself. So basically if agents become platform agnostic, which you just said, you're using it on a windows device, not an Apple device.

‍

And by the way, are you using it on a phone to which might be a Google device? Um, uh, therefore Apple's, you know, their operating, their operating system advantage is already eroding because the intelligence layer sits above the OS. And if that layer is going to run on any of those things I just mentioned, Apple's just another endpoint. So, basically Apple's been slow. They've been cautious. They've been incremental. And others like Google and now Microsoft even with distribution you know they're doing $99, $49, $100 stacks. Um Google has Gemini. It's deeply embedded. Um, and now Apple has a slightly better Siri, maybe slightly better.

‍

We don't even actually know if it's any better. So perplexity. OpenAI owning the experience and running that on Apple hardware, to me, means nothing more than Apple is just a commodity hardware layer. And frankly, when you're a company that's built its entire success on being special and warranting a high margin, now what you've done is you've created a nightmare scenario for your company. So my opinion is, you know, they've commodity, you know, commoditized AI models. That's not a defensible moat, uh, running other people's models on your hardware. That's not a defensible moat. In the end. You know, Apple is just another endpoint. It's just another commodity with good inference, but is a little bit of good inference enough to really move the needle for the company?

‍

Or is this just another ploy to sell cell iPhones and laptops?

‍

Paul Moorhead: 

Good stuff Daniel. So now we can really talk here. No, um, it is pretty amazing, um, that they were able to to pull this, uh, to pull this off, and they've been silently adding so much capability to their client hardware, especially on the bandwidth side. It's just it's nuts. But still, I think there's a long way to go to get people excited.

‍

And when people think of the amazing things, like everything that you and I talked about and what we're most excited about, um, had nothing to do with the device. Um, I the Citrine article, even though it said some really crazy, insane stuff and they, they published it to cover their shorts. Um, they did make some great points that I thought was an amazing vision, which essentially is, you know, uh, AI on device. You know, imagine it's, um, um, everything you do on your phone is just automated, like, everything you and I are loving to do. Um, like, you know, shopping for groceries and stuff that we can just automate. That's that's on a browser, on our smartphone or between apps on there. I do think. Um, once. I don't know, I don't know what's holding it up at this point.

‍

I have a feeling it's compute power on the on the device. I mean, we saw freaking open claw on on gigantic Mac stacks of Mac minis, right? And the cool stuff you were able to do on there that just needs to make it to the smartphone.

‍

Daniel Newman: 

Yeah, man, I, I think they have the opportunity to sort of be the edge computing layer for AI. Um, clearly all these platforms, you know, we're seeing all this renaissance of compute the CPU and of course, every Mac mini is a little mini data center. I mean, every PC, every notebook is like people need to understand that. Like they have all the heterogeneous computing on these things, but it's limited. And so when you're trying to run like a trillion parameter model, it's not plausible, but as model as model efficiency gets better, as you can do more with 7 billion parameters and you can get more accuracy, like these devices can do a lot locally, and there's a lot of economic advantages.

‍

Companies are going to be looking at that like, oh, you can do all this. We don't have to make all these calls to the cloud and spend all these tokens. They're going to be pushing for that. And if Apple can deliver much better economics, they're on a better UI that people are using love every day. Then of course, the consumer is a whole nother side of this. So it's really interesting. I still don't know if I fully believe it, but I'm beginning to be able to convince myself that this is something that could become an advantage for Apple,

‍

Paul Moorhead: 

I love it. Hey Daniel, let's move into bowls and bears now we've got to. I don't necessarily have to do a speed round, but let's get in and and crush this. Let's dive in.  All right Daniel, hype up first. It just seems like there was nothing that the company could do to satisfy Wall Street. 

‍

Daniel Newman: 

I mean, it had a nice pop. It went up like 9% after hours. I don't know where it's sitting right now. 

‍

Paul Moorhead: 

Oh, I'm talking about the. I'm talking about the past, bestie. Oh, yeah. Yeah, I'm just saying. I mean, let's finish my my intro. But now they had a great earnings. 

‍

Daniel Newman: 

Sorry. I thought we were doing. I didn't know I didn't know you were going to give me color I got bored. What did he say? Um, 18% revenue growth. Uh, 151% year over year networking growth. You got to love that. Data center networking in particular, 382.6%. But that does include a whole company that was acquired against, you know, a ground floor of zero. So, you know, it's important because you and I are truth tellers. You know, we are. Look, look, um, Antonio Neri got it right. Buying juniper, that was a differentiator. He was not going to compete at scale with Dell. Yeah, The data center this is their differentiation layer is being the company with the networking uh infrastructure for the enterprises. And that number looks great.

‍

Um, you know, they're doing well with cash, improving the debt situation. Uh, and they've raised guidance substantially on their growth because of networking, showing like 70, 68, 70%. So love that. And of course, they saw a big improvement in their free cash flow. Um, you know, I kind of said it. Look, it's staggering growth of networking. That's the differentiator. They need to pull as much of the shipping on their, uh, GPU and Nvidia forward to start showing growth alongside some of the other companies that are that are growing notably faster, but they've also been selective. He said it like they are not going to chase the same deals as Dell and Lenovo.

‍

You know, they are going to go after that enterprise. They're going to push for that higher margin. So you got to kind of have that patience there. Um, and what was it was 720 bits of gross margin expansion. So clearly this acquisition is is bringing cash to the bottom line, which is really important. 

‍

Paul Moorhead:

Yeah, I mean, I think what they did is they turned up the contrast ratio versus other other plays out there, right? I mean, obviously, if you're not going to participate in the high volume hyperscale AI server volumes, uh, you need to bring in a lot better profit percentage, uh, on that. And, uh, the networking adder, I mean, uh, Antonio looks really smart for Aruba and, and now juniper. So also the, the beta or the risk profile, um, of buying juniper, I think has been put to bed. People forget they had over a year in the clean room to figure out. And it's interesting, this, uh, um, water cooled switch, uh, 1.60, um, got a lot of, um, got a lot of, uh, interesting, interesting play. Uh, let's move on to Oracle. I mean, listen, I think I did the run up show. I think you did the the post show on Yahoo Finance.

‍

But you know, they they had the it was really a show me quarter. Right. And and really only one number mattered and that was OCI growth. Um I had said that they needed to be in the 70% range. I think they hit 75%. Uh, or sorry, 84%. I said they needed to be at 75. They hit 84. And and they did just absolutely crushed it. And then they, uh, they I think they doubled down on, um, they doubled down on CapEx as well, uh, at the same time, which was uh, which was pretty, pretty awesome. Um, that $90 billion 2027 uh, guide, uh, and a TikTok's equity stake, uh, people aren't even modeling, uh, at this at this point. And I think even the biggest debt hawks need to explain how you lose money on $555 billion in contracted backlog. Right. Um, and maybe those are just shorts who know nothing.

‍

Daniel Newman: 

Yeah. I mean, the OpenAI drag is real. So Oracle having such a big part of that backlog being associated. But I've said this on the show and I'll say it weekly because maybe a new listener show up at the fungibility of compute. If they build these data centers someone will use the compute. The argument about the debt bomb and the argument about them basically funding old technology with debt, which was kind of a big argument that came out this week.

‍

That is every data centers challenge right now, every data center that's being built is being built for a certain architecture of mostly Nvidia. But whatever it is they're building and there is a level of flexibility. And I think Clay McGuirk and the team came out and talked about this like we're not building something that absolutely is stuck in time. It's not like we can't add Rubin or Feinman or these features, but obviously they are trying to get these things up and we're seeing in real time the impact of a generation of GPUs right now, because all the you and me glowing and oozing over computer and opus, this is the Blackwell era. We're seeing it in real time. So yes, they just have to be able to update to Vera Rubin and God, I can't imagine what the step function of compute capabilities are going to be when that thing comes to production. Pat. But like they will do it. They've got the growth, they've got the backlog. They've got, you know, clearly people are more than happy to fund their debt because for every small incident they've had with funding issues, they've had no trouble being able to fund the debt.

‍

And they have equity financing that they also haven't even tapped into yet. So look, if if OpenAI absolutely capitulates, there could be a drag, a temporary drag while that fungibility moves to another supplier. But I think that, you know, this is a panicking story. It's not a real story. And of course has a real core business. They don't have the same core business as Google or Microsoft to create the same amount of cash, but they have a real core business. And I mean, the stock would drop 50%, I mean, 50 plus percent from its highs even as its numbers continue to grow. So I think it's been repriced. And I think they think they showed what they needed to show.

‍

Paul Moorhead:

Yeah. Well, a lot of people missed too. I mean, in there, you and I both attended the Investor Day in, uh, in Vegas. And, you know, they did a great job talking about, hey, we don't just put like, $5 billion out there. It's like we're very carefully this is when we have to commit cash. And they also showed the four different ways that they make money on these infrastructure with different margin profiles. Um, they were very open, right. Some people and somebody asked me, why is Oracle so closed? I'm like, they told us everything that financial day. So, uh, go go go read the transcript and look at the slides. Hey, Daniel. Let's follow. Let's close here with Adobe. Uh, Q1 earnings here. We already talked about, uh, CEO uh, stepping down, but hey, uh, what a what a way to go. Uh, you know, out with a ton of beats out there, right? Um, AI first IRR tripled year over year. 

‍

Um, you know, the big elephant in the room, though, is obviously the CEO succession. And anytime you put a board member, um, a, uh, what I'll call a non-operating board member that, you know, is not going to become the CEO. Uh, you know, that that it was swift, right? This wasn't planned. This wasn't, um, you know anything? Uh, 19% non-GAAP EPs growth, uh, outpaced 12% revenue growth. Uh, which means essentially, they've got good operational, uh, leverage. So, you know, I, I think, uh, Think Shantanu. His his exit will be looked mixed and he had an amazing run. Built. Helped build and lead an amazing, uh, an amazing company. Uh, what a good quarter to go. Uh, to go out on. I mean, the market hated it, right? It went down. It'd beat on every metric and the stock went down 7%. Right. Because it can't decide if AI kills Adobe, uh, or it saves Adobe. Right. And the CEO announcement is just leaving at the worst time possible. Uh, without time to find out.

‍

Daniel Newman:

Um, yeah, I totally, totally agree. I think we kind of this was one of these topics where we sort of did it once, and we're doing it a little bit again. So this, you know, great quarter. This is just the perpetual uncertainty of AI eating software. Um, and nothing. I mean, these are great numbers, Pat. And by the way, like I said, it just reiterates that, like if you've ever heard the term like the market is forward looking, this is the instance of where it is forward looking. So sometimes I know you'll text me after an earnings and you'll say like WTF? How is this down on this result? And it's like it's rarely about what you saw. It's often leads into the guide. The guide was even good though, and then what it actually ends up leaning into is people believing the sustainability of whatever track it's on. 

‍

You see the actual curve of Adobe's growth in revenue. And it's like this. It's like the dream of any CEO just up into the right. And yet all the market has done all year is just hammer this company. Um, and I'm guessing if you're short new, you're saying I'm out of ideas. I don't know what to do. I all you know, because because when you by the way, you go to grad school or you, you you come up and you've worked in big companies, had real jobs, that's one of your favorite quotes. Like, all you can control is every quarter deliver the number like deliver the number, deliver the guide, and beat and beat and beat. And when you do that for what, 18 years in a row. Almost every quarter. And you're getting hammered. At some point, you get up and say, look, let someone else try this. And I think that's probably, as you mentioned, swiftly, that's what happened, is to let someone else try a moment.

‍

I'm sure Shannon has made a little money. He can take a break, he'll take a breather. And I'm sure if he wants to come back as a CEO somewhere else, he will. Um, you know what a great career, great quarter Adobe. Um, you're unfortunately in the infinite loop of going to have to prove yourself every single quarter and deal with the rerating that's taken place. But yeah, there we are. There we are. We did it. 

‍

Paul Moorhead: 

Daniel. Daniel, do you want to hit, uh, do you want to hit any more companies, or do you want to you want to call it? 

‍

Daniel Newman: 

Yeah. I mean, look, um, you know, I'll just really, really swiftly mention, you know, TSMC, um, you know, they release monthly, so we don't need to dive too much into this, but it's always worth watching. Talk about an indicator. Um, a couple indicators that came out is there, you know, 30% continued growth, by the way, a little behind 33 that the market wanted. But this company in my opinion. I don't know how you feel, but like the ultimate bellwether is like of the next earnings cycle is always what TSMC shows at the beginning of the earnings cycles.

‍

If TSMC is ripping, it's super indicative of just how much demand is going to flow through flowing down to the chip companies, down to the OEMs, down to like, so when they keep growing and they're planning to spend 45 billion of CapEx, by the way, this year. So they're expanding. They're growing. And one more little note.

‍

I don't even know if it's really bulls and bears. But did you see that ASML is getting into advanced packaging.

‍

Paul Newman: 

Yeah. Hit it one more time? 

‍

Daniel Newman:

ASML I saw this morning I haven't even read the details but just kind of a note. Something everyone could go check out. 

‍

Paule Moorhead: 

Yeah like I, I didn't think I heard. You didn't think I heard you correct. Yeah. I mean they're not going to do the full packaging flow, but I'm interested to see what they're what they're doing. I'd be surprised.

‍

Daniel Newman: 

You know. Yeah, I mean I, I'm reading it on Twitter, so I haven't dug into it yet, but I saw it and I'm like, oh, that's kind of interesting to mention because, yeah, talk about a company with a moat that, by the way, we barely ever talk about like what happens if they, like, ever wanted to like, double prices. I mean, I wonder how they even why they ever even control their prices. So, like, there is no. 

‍

Paul Moorhead:

Well, listen, I, I've. They are a very methodical company that looks long term. They don't want to kill markets. They don't want to kill competition. They're really, really smart. Right. And and the reason that all the big CEOs go there to negotiate in front is because the relationship people. And, and, um, I've heard stories, right, that that talk about even the hyperscalers senior leadership going in there and they essentially get get laughed out of the room. They don't trust them. Right. And these are companies that are worth trillions. Right. So history and method and markets make the biggest difference to. To TSMC I think the best example of that is is what's going on today, which is. If you were just doing financial doing, uh, making decisions based on financials, you would give.

‍

Wafers and do packaging for the highest bidder. Right. They don't do that. Right. They're smart. They know they need a portfolio. They could kill the, uh, PC market, uh, as well. And they. They're they're not doing that. I think we're seeing, for the most part, that same type of discipline and strategic thinking from the memory providers as well.

‍

Because where do you make most of your money? You make it from the GPUs. You could kill the windows PC market. Uh, the smartphone market, but then you're deciding not to. So good point. 

‍

Daniel Newman: 

Good point. There is there's a heart under there, they say. 

‍

Paul Moorhead: 

Yeah, yeah, that's exactly right. 

‍

Daniel Newman: 

You're the host. You got to close this out. 

‍

Paul Moorhead: 

All right, buddy, we are done. I want to thank you for tuning in. It's south by south. We west weekend here. I already kicked it off last night. I attended the, uh, Craft Adventures opening party. Uh, they're opening up an office here in Austin. Subsequently, uh, David Sachs, uh, moved from California, San Francisco to Austin on December 32nd.

‍

Just kidding. Uh, at the very end of December, uh, I'm not the big party guy, but, uh, I may or may not be going to the, uh, the All In podcast, uh, today, and we'll see what happens. Anyways, y'all take care. Hit that subscribe button and, uh, have a great week.

‍