Anthropic's S-1 Opens the Books, Oracle Goes Heterogeneous, and the Real AI Capital Race

The handpicked topics for this week are:

🔹 Anthropic’s S-1 Puts Frontier AI Economics Under Public Scrutiny: Patrick Moorhead reads most of the roughly $80 billion first-half loss as accounting, with revenue doubling sequentially, but still found a $2 trillion valuation hard to support. Daniel Newman places it at about 20x run-rate revenue, half of Palantir's multiple, and said ETR data suggests enterprises are adding open-weight models alongside frontier ones, not replacing them. (The Decode)

🔹 Oracle Puts OpenAI Inside Its Application Stack and Goes Heterogeneous on Compute: Daniel says Oracle is embedding OpenAI models in its database and applications and adding TPUs and Cerebras to a stack that was mostly NVIDIA. Pat notes Oracle capped new capital at $20 billion through 2030 and cut upfront capital on a one-gigawatt deal from $37 billion to $6 billion using customer-supplied hardware. (The Decode)

🔹 Google's Gemini 4 Carbon Leads on Capability While Access and Cost Decide the Model Race: Patrick says Carbon leads on capability but remains hard to access. Signal65's new Pinnacle coding test shows the top four models overlapping within confidence intervals, so cost per pass is the spread that matters, and Daniel adds that no single best model exists anymore. (The Decode)

🔹 Intel's Classiq Agreement Targets a Full-Stack Quantum Business: Newman defines Classiq as an EDA layer for quantum that gives Intel a neutral software position before any qubit technology wins, with the deal expected to close early next year. Moorhead says it turns Intel's spin qubits, cryo-control chips, and SDK into a business, and flagged its Nervana and Habana history as the execution risk. (The Decode)

🔹 The Flip: Neocloud Consolidation Would Be a Financing Reset, Not Proof That AI Demand Has Collapsed: In a simulated debate with assigned sides, Daniel Newman argues consolidation is a financing reset, citing Azure up 48%, Google Cloud up 103%, and AWS up 39%. Patrick Moorhead argues the squeeze is real because demand pays off only when assets stay fully utilized and Neocloud utilization data is not public. Both ultimately agreed demand persists. (The Flip)

🔹 Microsoft's Azure Growth Strengthens the Case for AI Monetization: Moorhead calls Microsoft the best of the week's earnings, with Azure accelerating and engineering freeing capacity from underused CPUs. Flat commercial RPO was the soft spot, and the stock re-rated about 7% before the broader market pulled it down. (Bulls and Bears)

🔹 Google Cloud Surges, but Alphabet's Cash Demands Complicate the Earnings Story: Google Cloud grew 103% to $30.7 billion while ads grew 12%, raising the question of whether cloud can fund its capex without a faster-growing ads business. (Bulls and Bears)

🔹 ServiceNow's Selloff Tests Whether the Outlook Is Weaker or Merely Mistimed: ServiceNow beat on revenue but was hit on current RPO after management said the prior quarter pulled in revenue, and AI contract value slipped slightly. The sharp stock drop came with a re-rating of only about 2%. (Bulls and Bears)

🔹 Amazon's AWS Growth Accelerates While the Infrastructure Buildout Consumes Cash: AWS grew 39%, its fastest in five years, while roughly $220 billion of expected capex pushed Amazon's trailing free cash flow to negative $42 billion. Daniel comments that this week's cracks were in commerce, search, and PCs, not AI. (Bulls and Bears)

🔹 Intel's Stronger Quarter Gives Way to a Weaker PC-Driven Outlook: Moorhead calls it a great quarter with a weak guide driven by a PC slowdown and substrate constraints, with the stock down 27% as of this recording. Newman says the story ran ahead of execution, and the turnaround is a multi-quarter effort. (Bulls and Bears)

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

Daniel Newman:

I don't know, man. I don't know if anybody's going to believe we know what we're talking about. This week, man is canceled 6th time since its reception 2nd time this month. We're having a absolute meltdown as we record this thing. Patrick is literally worth, like. 10 bucks at this point, that's all he's got left. The good news is he looks great. You are catching him preamble before he has to go on. He's gonna get his sword in one hand, his shield in the other hand, and he's gonna try to figure out how to defend this AI trade with Intel. We're back though, everybody, for, what is this episode? We are in episode 325. 325 times you have tuned into us, if you are one of our loyal listeners. And if you're not, you should ask yourself why, because you should be here each and every week, because this is the best signal in the market.

‍

Patrick Moorhead:

Yeah, it is. I'm trying to figure out, you know, which balcony should I jump off of, the 19th floor, that's my office, or the 39th, that's my condo.

‍

Daniel Newman:

It feels like the momentum you could gain on the extra 20 floors would really take care of any risk of feeling anything.

‍

Patrick Moorhead:

I think terminal velocity, you hit within about 1,000 feet? I don't know. I'll ask. I'll ask. Flash that.

‍

Daniel Newman:

Yeah. Let's, you know, keep our suicidal thoughts to ourselves for the rest of the show. He is kidding everyone. You know, as we'll always say, don't invest money you can't afford to lose because it is the casino, as our president once said, no crying in the casino. But there's still a lot going on in tech. Like this week, has it not been like a forever week? Like it's kind of wild because we've actually gone through both the kind of a euphoria of we're back, and then we went right back to the absolute demise, doom, end of days that seemed to follow what was a pretty abysmal Intel print. The others, I think, are more nuanced. And by the way, I'd say abysmal guide. Actually, the print was OK. But yeah, there's still a lot to talk about this week. So our show does not just sit and complain about the world's ends and dooms. We do like to cover what actually happened. And there's a lot of things happening worth noting. There's been acquisitions. There's been S1 filings. There's been events. So we're going to hit it all. Pat, I know we're going to keep this thing moving. And while I tend to run long, I like do this. Let's just get right to the decode and start off and we'll talk about what happened this week. All right, this is probably the topic that deserves the most time and attention, but the world waited on the Anthropic S-1 filing because Michael Burry was certain it was going to show that they had burned like 500 billion in cash and had no path to making money. I know. Yeah. So I want your take on it though. Like, like, like that was the biggest thing that happened this week until the next biggest thing happened.

‍

Patrick Moorhead:

Yeah. I mean, there's so many ways to break this down and let's talk, you know, the headline loss is mostly accounting, right? First half net loss, 80 billion, 72 billion was the change in FMV and convertible notes, derivatives and safes. So operating loss for the half was 693 against two bills. But you are looking at being cash flow positive soon and also the operating swing, right? Because Q2 26, right, was 11.6, implies like 4.7 billion in Q1. And Q2 was almost a billion dollars of operating income. and an applied Q2 loss of 1.7 bill. Revenue more than doubled sequentially while the company moved into an operating profit. I think that is pretty darn good. The growth numbers are pretty extreme. But that revenue retention was 900%, which is crazy. Well, actually, that excludes pro and max subscribers. And they did say in the S1, in addition to destroying the planet, that the growth may not be sustainable. Where I sit on here, NetDaniel, is glass half full, glass half empty. But also, we must take a realistic look at whose lunch money Anthropic is going to steal, or is it going to be some new market? And then you have to ask yourself, well, where is all that money going to have to come from? So where I end up on this, Daniel, is I find it hard to imagine the company is worth $2 trillion. And I think even 1.5 trillion is somewhat of a stretch to me unless I can clearly see whose lunch money they're going to steal. And where is that money gonna come from across consumer business and government?

‍

Daniel Newman:

Yeah, I will give the defensible argument on valuation, but I won't say that I necessarily agree with that. Palantir trades at close to 40 times revenue. Anthropic at two and an exiting run rate of the year of 100 means it's trading at about 20. So at about half of the you know, multiple that is being put on Anthropic, or sorry, being put on Palantir is being put on Anthropic. So that's kind of how you get to like, okay, that could make sense if it continues to grow at these rates. The deceleration question, because it had an incredible growth in the first half, and I think there was some concern that there was some slowing. You do have law of large numbers. I mean, no company has ever grown to this size in this time. This is an unprecedented growth rate. And it's never happened before. So this is something that the market has to kind of digest and say, like, is this achievable over the long term? They said it themselves. It may not be. We have token prices eroding in the sense that, like, what people are actually paying per token is less. And of course, we have demand surging. So how many tokens people are using is going up. But it does mean that people have to use a lot more AI to continue to generate revenues and run rate growth at this rate. And then you have real concerns about open weight competition. those models are pretty good. I do think that we're seeing though that people are hanging on to frontier intelligence. They just want to use the best. And despite the fact that like there's kind of this thought that maybe they won't and they'll start using second class models or N minus one models, whatever we want to call them to do more things. I'm wondering if that really does live in a small subset of sort of your tinkers and nerds and not necessarily inside the enterprise. And our ATR data is showing so much propensity to continue buying of large enterprises, both within OpenAI and Anthropic. So that's optimistic. Last thing I'll say is it was one of the most depressing, sort of, it was incredibly depressing to read some of their disclaimers, some of the warnings. I mean, it literally did say, to your point about destroying the planet, They also talked about demise of humanity and basically existential threat to people. You know, I think I said something on Twitter along the lines of excellent as one besides the whole end of day shit. Yeah, like, but like, actually numerically and on a business standpoint, Pat, like. this was really positive. Like there've been so many kind of doomers and AI perma bears that have said like, this would just absolutely be garbage. This company has no path to making profits. And it is printing multiple dollars of revenue inbound for every dollar it is spending on compute. And that efficiency is growing by orders of magnitude in more compute. And I think that's an important note. These labs can make money. They're showing that they can, even at this growth and scale, become profitable. But Pat, I'll finish on your thought. Does that warrant $2 trillion of valuation? And where does that come from? Because the market, there's only so much liquidity in the market. So raising that money at that valuation is going to come out of other things. But I think this one is a blockbuster. I just think it's so popular and so interesting, and people are going to want to be near it. I'll probably stay away from it on IPO day. I'll probably let some of the dust settle.

‍

Patrick Moorhead:

Yeah, we saw what happened with SpaceX and the first people that went in there. And by the way, the ETR data is great on what people are currently doing. I think it'd be interesting to see the intention data around that. Because I got to tell you, I don't talk to a single CIO that is now mega emboldened on what they want to do with open source models at this point.

‍

Daniel Newman:

I will pull that up and I'll come back to it when I pull it up, because I'm going to run it in my little MCP sitting on top of Claw that looks at all my intent data. I'll circle back and I'll rub it in your face when I get the chance. By the way, I think it's And you and I and what we're building, we definitely believe you can build tokens at much more optimized pricing for these enterprises, no doubt. But it is work, building these harnesses, getting these models to perform at the level that's expected, dealing with some of the idiosyncrasies of these open weight models. You need more skill, you just need more skill. And this is where the FTEs and the service providers that are in the middle, that's why Accenture's growing. There's more skill than an enterprise. So an enterprise just wants to turn on AI, you turn on OpenAI, you turn on Anthropics, it just works outside of the box, in the app, in the heart, everything is kind of there for you.

‍

Patrick Moorhead:

By the way, there's also nothing to say that Anthropic can't roll models on-prem either, right? And in fact, Google's already doing that with Gemini on a full stack. So that could be the offset to all of this.

‍

Daniel Newman:

Yeah, we'll see. I mean, didn't, yeah, I was looking yesterday at the pinnacle benchmarks, like, didn't the new Google model, Argon or something, like, didn't it top our benchmarks this week?

‍

Patrick Moorhead:

Yeah, because we needed a new name, it's called Carbon.

‍

Daniel Newman:

All right, so Oracle this week had AI World. I made it out there for a day. I attended a few sessions, and then I also spent some time. They had their investor day. A lot going on there, but kind of the 2 things that I thought were really noteworthy. Is new partnerships in terms of inside open inside of their enterprise. Application stack really kind of deeply embedding. OpenAI, Sam showed up, you know, with a security mesh, with governance around it, that you basically now can operate inside all your Oracle Enterprise, you know, database and applications and fusion and get the benefits of the OpenAI models inside. I thought that was probably one of the most interesting announcements, but the real show stealer at OCI was heterogeneous compute. Uh, you know, Oracle's been slower to sort of commit to diversifying. They have not had a internal silicon project. that anyone knows about to build their own AI chips. And they were largely NVIDIA with a little AMD sprinkled through out more on the CPU side. But now they're basically saying we're going to play the long game here. We're going heterogeneous. We're going to add, first of all, TPU, which is just another proof point that the The hyperscaler pat are going to become merchant silicon providers. We heard it from Amazon this week in earnings. More training is going out. TPUs are finding its way to more clouds. And then they also are going to add Cerberus. Did we say Cerebrus or Cerberus? I hear ways, Daniel. I do. I'm a cerebrus guy. I've always said cerebrus, but, like, I've heard, like, you know, it's like, when I hear those guys say no video on TV, I start to wonder if I've been saying it wrong all these years. By the way, we're not it's definitely not in a video just in case you hear a guy go on TV and talk about it. You probably might not want to listen to that person. But they're adding cerebrus and. Of course, jalapeno is in the plans with the deepening partnership with open so I thought that was actually the thing that really showed me the most intent for growth. They did also in their investor day show some very optimistic. you know, roadmaps as it relates to getting data centers from zero to built out over nine months. They talked about, I think in 29, I'm just doing this from memory, but in 29, I think they will start to, the debt cash flow number starts to turn. They basically are saying there's only maybe one more debt raise they're going to have to do in 27 to support the growth. I mean, look, this has been the balance sheet that everybody's been paying attention to. It's in the hyperscale space. It's been one of the most challenged. You could call it ugly. The credit swaps on these things, the spreads are just getting massive. There's concern about existential risk to this company. I really do think they know what they're doing. It doesn't mean that this is the moment that you would want to buy in. I'm not going to We don't really do that here anyway. But if they can achieve their targets, their $600 plus billion of backlog. And basically, my big hook with them, Pat, is open AI survives. Open AI survives and does well. Oracle is going to do well. So I'll leave it there. And I'll let you take this one for a bit.

‍

Patrick Moorhead:

Yeah, it was a great breakdown. I just kind of add in some stuff. Not only are they trying to be the Swiss Army knife of accelerated compute, it also looks like they're trying to be the Swiss Army knife of data platforms as well, even though they've got a massive data platform of their own, right? They did a deal with Snowflake. to bring them in and against, again, you know, Oracle likely views Snowflake like Satan, which is a direct competitor, but now you can use Oracle credits against Snowflake. They truly are becoming more of a, I would say, closer to I can't tell if they're closer to AWS or getting closer to Microsoft, because Microsoft has their own SaaS play and so does Oracle, but really becoming this Swiss army knife for chips and data platforms. And then if you look at their agentic play, pulling that all together, Their agentic play runs all the way through Oracle's systems of record and not through a chat window. And I think that's a big differentiator. On the funding, I attended their investor day last year, and they did a tremendous job outlining exactly a lot of discipline that goes into it. And they hit the FUD right up front. And again, I'm not Babe Ruthing this, but I did like that Oracle capped new capital at $20 billion through 2030, half equity, half debt. And they guided free positive FCF in the second half of 28 and above $30 billion. in 2029. So I think that that is good discipline. And the other thing is, they laid out like on a one gigawatt deal, customer supplied hardware as opposed to them buying it. You see the side deals that people are cutting with AMD, NVIDIA, Intel, and the rest. Oracle's upfront capital moves to $6 billion from $37 billion. But they're trading revenue for returns and the market liked it. The market, it was up 11 percent. before the overall market collapsed. By the way, a comment on the heterogeneous compute, I think where you can sell every ounce of compute that you can make, I like that strategy. Daniel, I saw your post about how quickly they could bounce off that and between that, I don't know if that works in steady state, right? Where buying... You mean the fungibility?

‍

Daniel Newman:

Yeah.

‍

Patrick Moorhead:

Well, steady state, once you get to where supply opens up a little bit, where the more you buy, the more you save, you can get a volume purchase agreements. So, but who knows? I don't know. The stock market says that AI is over today. So, maybe it happens soon.

‍

Daniel Newman:

It's done. Nobody's going to use it anymore. So real quick, Pat, on the pervasion and interest on, you know, in Anthropic. So we talked to 1,112 enterprises in October 26. Anthropic adoption coming, adding it 16.5%, we're planning to start paying it and 63.9% are increasing spend with anthropic, 16.9% plan to be flat with anthropic and 2.2% we're decreasing. Only 0.4% replace. Yeah. Well, open AI was a little lower, 8.7% adoption, 52% increasing spend, 30% flat, 5%, and they had a 3.5% replacement. So a higher number of enterprises are planning to bail, but like, it's almost negligible attrition. So maybe to your point about the open source, open weight, what's probably happening is it's an and strategy, not an or strategy. They're probably not ripping it out, but they're gonna, add it to for certain workloads where they can do much better economically. I don't know. That's what the data says. It's just data, folks. Don't listen to it. Trust my anecdotes. Anyway. All right. Let's take the next one, Pat. Gemini. Okay. Carbon. Let's talk about this is our little model section and did a pretty important update from the world, the land of signal 65 pinnacle.

‍

Patrick Moorhead:

Yeah, for sure. You know, another week, another 10 models out here. So, uh, leading it as carbon, right. Leads on capability, but, you know, you can't get access to it, uh, right now. And, you know, quite frankly, that, that has been a little frustration, even on, on my company side, getting access to Google's, uh, latest grade, uh, greatest, but it does certainly look like, uh, Google is, is back and, and they're on top. I don't know if they brought it out early to buttress their not so great or not so well received earnings, but it's out there. And you've got, you know, a chat should be Tara comes out as competing on price and computer use, not peak coding, which is cool. With all these websites shutting down, shutting down agent access, they can get access through your browser with Kua. So this is very important. And yeah, Signal 65. announced yesterday its brand new coding test. You need to check this out, and hopefully we'll put this in here. But essentially, the test has models build software in unfamiliar code base, which is unique, and then grades the results by running it against unseen samples and counts failed attempts in the cost per pass job. Again, a Signal 65 pinnacle First, I mean, hey, we come in later than the market. We've got to bring something very, very new here. So confidence intervals are about 30 points wide and the top four overlap. So the spread that matters is cost at this point when you check that out. But we should probably bring Ryan on at some point to break that down.

‍

Daniel Newman:

Yeah, I mean, look, what did Ryan say? The best enterprise model is not necessarily the best coding model. There is no single best model anymore. Exactly. So there is a best model for the job and selfishly for everyone out there, if you haven't checked out. And by the way, all this data is out there for you. So check out signal65.com slash pinnacle. It is the ultimate benchmark. for all kinds of different use cases. And that was signed, trust me, bro. All right, let's hit the final decode topic. This was kind of an interesting one to even talk about now, because I feel like talking about Intel's ambitious long-term vision versus what it just did on earnings, like almost deserves, like, I should be slapped for talking about its quantum ambitions. But at the same time, we know companies, despite, stock market tantrums have to keep operating for the longterm. They can't just make decisions based on how the market might react to them or whether they're gonna sell shares. Intel wants to be part of what comes next. And so I think he's starting to run more of his cadence playbook and doing it with quantum. He wants to own the automation layer. He's trying to kind of stay neutral on the hardware a bit. And what a different, what a better way to do that than getting into the software side. If you kind of look at how, what Classic is, and I want to emphasize the Q, it might just be Classic, by the way. It's kind of like an EDA for quantum. Abstracts hardware-specific circuit designs, the way Cadence and Synopsys does this for chip designs, and obviously the former Cadence CEO is probably a good person to do this. You know, Intel, you know, it does own silicon spin qubits. So they are in that space. So they've been in this and it's been around the business for a while. To me, it's a fairly offensive move. They're getting ahead of a trend. They're trying to buy a kind of a seat to be where they expect things to be in the future. Because until fault tolerance arrives, you know, we don't really know which qubits are gonna win. So this is gonna close sometime in the beginning of next year. And I don't expect any issues with it. It doesn't feel like there's any, going to be any regulatory battle here. The only, this is another one that just interestingly enough, I don't know if you saw that, Pat, but I guess Lipbu is an investor in this one. And obviously, Um, that puts it on everything though, like, he legitimately is just, he's just so diversified to his credit. So, by the way, there is no disclosure on the terms. I think it's an interesting long-term bet. You know, it's just, I still think, and we'll talk about this later, but I still think the Intel bet, though, is all foundry. Like, this one, to me, is just, it's interesting, it's exciting, quantum's coming, but there's so much happening right now, and now, after that print, execution's gonna be the focus for a while.

‍

Patrick Moorhead:

Yeah, a lot to talk about here, but it is tough to talk about in the backdrop of that earnings. But again, Intel does have to invest in the next bets. Otherwise, it's going to find itself in the same place as it is with data center AI accelerators. So first off, Intel has been doing quantum for a while. And what this acquisition does is really pulls them into what a business would look like. They've got spin qubits, horseridge cryo-controlled chips, and its own quantum SDK, but never built a business around them. And it's funny, I don't know whether to call them classic or classic. If I do the Q or not, do I do the Q? Well, when I was I was on with Intel's CTO Pushkar and also near the CEO, and they used the word classic. So that I recognized I had been saying it wrong for a while. But no, this turned us into more of a business than it is a technology buy. I think the key part is is is neutrality, right, where they want to be the glue that's feeding into things like, you know, above IBM's Qiskit and NVIDIA's CUDAQ. The big investors, I mean, Qualcomm is an investor, IMQ is an investor. There were a lot of, NVIDIA is an investor. So, it's interesting if they can do both. It's very similar to Qualcomm's modular in that making this more of an industry-wide investment than just Intel. But that is difficult. As far as Intel acquisitions go, you know, history is the risk to manage if, you know, whether it was Nirvana or Havana. Those didn't become lasting businesses, but they're staffing this differently, right? Near is in charge as opposed to having it sucked into the data center business on its own. And I think this is a, a smart move, again, very similar to the CEO of Modular at Qualcomm reporting directly into Christiano. So makes a lot of sense, but again, I'm not Babe Ruthing this. I think the valuation was at 2.3 billion. That's what Reuters had said. So, you know, congratulations to the founders and Lipu. How much did you own? I didn't do a work for equity deal on this one, unfortunately.

‍

Daniel Newman:

Damn it. You never do the ones that are like 10 Xers. You only do the ones that are one-tenthers. I don't know.

‍

Patrick Moorhead:

Grok was pretty good.

‍

Daniel Newman:

Except for that part that we're still waiting on.

‍

Patrick Moorhead:

Yeah. Yeah. Maybe that'll be an episode for another show. How we got shorted on that one.

‍

Daniel Newman:

So you did, it could have bought a horse for you, Pat, another horse.

‍

Patrick Moorhead:

Are you trying to make me cry? I already lost my ass in the market. Like, you know, remind me of how much I lose on horses.

‍

Daniel Newman:

No, I don't ever want to see you cry, buddy. I just want you to remind yourself that you've had all this money to lose. Some people never have it.

‍

Patrick Moorhead:

You know, it never makes me feel better because it's never good enough. No, we're insatiable.

‍

Daniel Newman:

Hopefully you all are too. I hope you are.

‍

Patrick Moorhead:

I think a lot of people who listen to this are probably somewhat insatiable. Yeah. Right. Or they have their own podcasts and and they're not listening to us.

‍

Daniel Newman:

or what Alexis Ohanian called like the dead internet. It's not actually humans listening to us, it's just bots. Hey there bots, hope you're enjoying this show. We are also, we welcome our AI overlords to the audience and the community.

‍

Patrick Moorhead:

I sent my first automated email to somebody just based on a yes. And I did it through Muse. And it was, you know, pretty good. But at the bottom, I said, make sure you put the disclaimer. This is coming from Pat's Muse bot. It was a family. It was a family thing. So I figured.

‍

Daniel Newman:

I like that you're first deprioritizing human to human interactions with your family, and then you're going to work back into doing it in your business. It's good. That's that's actually how I would do it as well. All right, buddy, let's get to the flip. We got to move, man. I got to get you on TV. You're going to go be famous. Let's go to the flip. We're going to have a simulated debate on an argument that we may or may not agree with our side. But this week, we're going to basically ask the question. We're going to talk about the neoclouds. This was the big debate. We probably would have had it last week if we had done our show. But it was whether neocloud consolidation would be a financing reset rather than proof that AI demand has collapsed. Let's see who's saying it's just a reset and who's going to say that AI demand is it. You had me at hello. Look, this one was easy. I mean, I don't even know if I need to argue it or I can just take the win. Go back, read my Twitter. You'll see I already nailed this one once. It's sold off again. I'm going to nail it again. Look, It's the easy side. This week's numbers made the argument for me. Azure grew 48 percent. Google Cloud grew 103 percent. AWS grew 39 percent, and it was its fastest in five years. Show me a demand collapse. I can't find it. So why did neoclouds all fall? Well, basically, Elon posted that 200-plus neoclouds faced guppy-level death and consolidation. Be honest here. Elon is not a disinterested observer in this debate. SpaceX just raised 40 billion of debt to buy more GPUs, and he's building his own compute. It's the fastest growing, quote unquote, neocloud slash hyperscaler on the planet. And it matters to him where capital goes and, of course, where it doesn't. Jensen also nailed it from the other side and basically said the demand for capital is bigger than the supply of capital, but that's a financing problem, not an AI demand problem. So lenders, by the way, are doing what they should do, right? They're separating crazy speculative bets from capacity and from contracted revenue. This week, IRINN got financed. Last week, Nebius got more financing. And Nebius said it would sell out its 2027 capacity It already can't. And Google added to that by basically saying they're going to have to use more third party capacity this quarter because they literally can't stand up and build enough fast enough. So some of these companies go to zero. Yeah, probably. Bad capital structures, or they will be consolidated, or they will be bought for purely the energy they can provide or the GPUs they still own. But that, again, doesn't signal demand. Our market model has data center CapEx going from about $1.6 trillion this year to about $3 trillion next year. At roughly 40 billion a gigawatt, not every balance sheet will survive that. But remember, 2001, the fiber didn't disappear. The equity did. Somebody bought that fiber cheap and the internet ran on it for 20 years. Same thing here. Consolidation moves GPUs to stronger owners, consolidation, but it doesn't turn them off. Sensational headlines sell. By the way, I tweeted this. Sensational headlines sell, common sense prevails.

‍

Patrick Moorhead:

You know, you are the leader of the one liners, Daniel.

‍

Daniel Newman:

I just don't have the time in my audience. I don't think you deserve all my thoughts for free. You gotta subscribe for that shit.

‍

Patrick Moorhead:

I love that. That's good. Yeah. I, uh, I need to figure that out someday. I'm just like free, free, free, free.

‍

Daniel Newman:

You're really great thoughts. I just shared a one liner of Pat's great thoughts.

‍

Patrick Moorhead:

Oh, that's, that's good. Thank you so much. So, um, listen, uh, Musk actually has never been wrong. He may have been off on timeframe, but, you know, Musk's math leaves literally no survivors, right? And, you know, it's funny if you look at even the giants, does it make sense to any of you that even the giants are raising outside money? Microsoft said it's going to tap outside funding. Went attractive, leases now and debt later. Oracle capped new capital at $20 billion, half of it debt. And if the best capitalized buyers are borrowing, how do the marginal neoclouds not going to get squeezed in this own name? And demand is not enough if assets are not being 100% utilized. And I know we actually have no data on percent utilization of each NeoCloud. Like how much is sitting there, meaning and not being used? Oracle showed a great example of how they flip between customers and I question if the Neo clouds have that type of capability. I mean, maybe they do, maybe they don't. And, you know, I think IRIN, right, is one borrower and not proof for the sector. And I think your IRIN example shows that financing exists for some, Just not that every borrower can fund new capacity on the terms of that work. I rest my case, Daniel, and I might need to go out and buy a bunch of neoclouds now.

‍

Daniel Newman:

Is demand, though, crushed?

‍

Patrick Moorhead:

I mean, market says it's over, dude. All right.

‍

Daniel Newman:

Well, so I'm not sure you, you, you, you, you, I think you did argue the financing problems. I'm just not sure we actually determined whether or not demand has been crushed, but I think you and I would both agree demand persists. There's a reason everyone's going out to the debt markets and wants to build more because they're going to make money in the future. And they're betting their current cashflow to make more off that cashflow in the future. That's why you use leverage. That is why people use leverage. And that's also how you get smoked when you use leverage, because you're making a bet and you're putting a multiplier on it. And if you're wrong, you're multiplying by wrong. I've done that before. Don't be a degenerate, everybody. All right, buddy, let's get to bulls and bears. So I know we're going to speed around this a little bit, because Pat's got to go be famous, and I want to get over to my TV to watch him. Let's hit the five. We've got Microsoft, Amazon, Intel, ServiceNow, and Google. But how about, let's just, you hit each one, and I'll hit one, and unless there's something that's just pouring that you need to add, let's just, we'll just swap off. We'll cover them all off. Let's start with Microsoft. You can take that one.

‍

Patrick Moorhead:

Yeah, it was all about Azure and they kicked butt. I mean, if I look at all the earnings out there, Microsoft was the best. Multiple beats, good explanation of capital being used, how their engineering group found the capacity, right? Underused CPUs. which brought that in. Azure is accelerating, not decelerating. I think that's incredible. And if you're going to pick on something at all, it's commercial RPO is flat. But again, market liked it. They had a 7% re-rating from all the data that I saw.

‍

Daniel Newman:

Yeah, to your point, Microsoft crushed it. Actually, just right now, I think it's just being dragged with the market. I actually think the performance is quite prolific. All right, let's talk about Alphabet Google. This is an interesting one. It was the beginning of the end, and it's that divergence of the business that is growing, and then the business that has paid for that growth. You know, I'll just quickly touch on this. The cloud business was absolutely amazing, and it further validated the strength of the AI build and the commitment of Google to its cloud business and its full stack infrastructure. 103% growth. Meanwhile, ads shrunk. And there's increasing speculation that that business may be in trouble over the long term. That business has been the financier of a lot of what Google's taken in terms of its various moonshots. And it's been the underwriter of cloud. Now, the good news is cloud margins have grown, its revenues have grown, its profitability has grown. But the question is, can it keep growing at the rate it needs to grow with the capex it needs to spend without the support of a strong Google ads business. Now, ads did grow 12%. So all the doom and gloom, it did grow. So like, let's just say it hasn't started to contract yet. It just isn't growing as much as people want it. And then there's concerns that that growth is coming from maybe some more pressure on ads itself, on the pricing of ads, on the distribution of ads. And so that's gotta be looked into. But largely given that we're AI and cloud focused here, that business crushed it. And, you know, at this rate of growth, they are quickly knocking on the doors of the biggies, you know, 30.7 billion, 103%. That's growing twice as fast as Azure, almost more, I can't even do the math quickly, faster than that against Amazon. You know, one thing of note, by the way, Pat, in our October ETR data, we basically noted that they had the highest net score amongst the different clouds, the biggest year-on-year gains of the three hyperscalers. 39% of our organizations in our September AI product series basically say they use Gemini, plan to stay, and another 18% are planning to evaluate it. So a lot of strength coming out of Google, but that divergence of core business and cloud business has been a stressor. By the way, it seems like there's been a lot of tale of two parts of companies, Amazon, Intel, but I'll let you hit Another company that, God, sold off so hard, ServiceNow.

‍

Patrick Moorhead:

Yeah, I mean, they beat the revenue. They got crushed on CRPO. You live by the sword, die by the sword. There were some discussion in the call that said, hey, we pulled. The reason the prior quarter was so good is we were able to pull some of that revenue in, which is why the current quarter wasn't wasn't as good. They missed on EPS. They, you know, they met on the guidance. And they, you know, they beat on revenue, and they just got slammed. Now, it's hard as you're a business where you are a people or your customers are buying on what your agents are doing and the intensity of the work versus, let's say, seats or cost per seat over time. But even AI, ACV was down a little bit as well. So AI is growing faster than the core. Uh, but it's just not, it's not, it's not a big enough percentage of their business. And I think, uh, I think investors, investors got spooked. Yeah, I think the free cash was, you know, stock was down big time, but the reratings were only down 2%.

‍

Daniel Newman:

Yeah. Yeah. Well, the analysts are terrific at making changes.

‍

Patrick Moorhead:

Well, after three analysts raises, raise their expectations.

‍

Daniel Newman:

I mean, if you actually look at long-term free cash flow projections and the fact that this is a very capex light business that seems to be still growing at a pretty nice clip rule of 57. There's a lot to like in there. But there are all the challenges that you mentioned. So Amazon, another tale of two companies. But this one's going to be interesting, Pat, because AWS has always been the cash cow. So unlike Google, where it was a cash cow that paid for the cloud to stand up, AWS's cloud has paid for every ambition that Amazon has had to build its commerce engines, its advertising engines. So it had that same kind of divergence where cloud did great and the other part of the business didn't hit the expected performance. But in this case, it's kind of a reversal. So now AWS going negative free cash flow will not be able to sort of underwrite as much of the commerce business right now. And so the concern is that with AWS always having sort of been the The financial minder for the rest of the company, where does that money come from? $220 billion of expected spend this year. But as Andy Jassy says, Matt Garman has reiterated, they can't build it or sell it. They can't build it fast enough to meet the expected demand. They didn't only have their fastest growth in five years at 39%. They're saying that this growth rate is expected to be sustained multiple years into the future. So they are either sustaining or even accelerating beyond what's their fastest growth in five years. So the A.I. investments are starting to really turn for the company. But the problem like I said is you know with the 220 billion. and the free cash flow going negative by $42 billion, there is, I guess, some real concerns about how does this work out. Now, these are grade A plus tier credit companies. They can raise. They can dilute. But of course, all those kinds of things impact balance sheet and impact shareholders. So they I don't think there's any risk of existential risk to Amazon being able to continue to build out. I think the growth in the business will will support it. But I do think people are rewriting the stock for a different financial structure. a different balance sheet structure and some different challenges to be able to execute. And then, like I said, on top of it, it's just the execution. You mentioned oracles. In Amazon's case, how quickly all this spending turns to growth and turns to free cash flow. And I think we know we're at least a couple of years out. And does the spending ever really slow? Are we ever going to be allowed to slow down on spending, or will this AI build-out always be this demanding on cash resources, Pat? But look, all these companies confirm the strength of AI. What we're actually seeing, in my opinion, are cracks around it. You're seeing cracks in commerce, cracks in search, cracks in PC's client in some case, like it's not the AI stuff. Everyone's full steam ahead with the AI stuff. All right, we got to get on last one. I think you're ready to go too.

‍

Patrick Moorhead:

Yeah, let's sit until really quick. Great quarter, absolutely horrific guide. Three things hitting at the same time, PC slowdown, substrates, and then a very unfortunate and sad accident at one of Intel's foundries out there. And the stock is literally down, as I'm looking at it right now, it's down 27%. Uh, which is absolutely yeah. Yeah. Well, hopefully you bought it, bought it, uh, when it was 30. So, um, and, and it's interesting. Um, uh, PC is a very known, um, issue and so are substrates. Uh, those are industry issues. Uh, And again, I can't fault a company for a market-inflicted wound, but I do want to see what AMD comes up with. Were they able to get substrates through TSMC? Were they able to move wafers from PC to servers and GPUs? How did they do? And the Israel factory is a horrific accident and not something that will increase. I think what this did, though, is it it popped the bubble of invincibility for Intel, who was trading around Foundry. You know, I would have liked to, you know, if Lipu is going to make a big claim, this would have been the time to do it. Now, they did sell the future with the size of the CPU market and the roadmap and everything. But I think investors were looking for more foundry, more foundry love. So we will see.

‍

Daniel Newman:

Yeah, I'll send you on your way. I have lots of thoughts, but both Pat and I shared a bunch of different tweets. I even wrote 1 that was more than 1 line. If you want to read it, but yeah, look, it's it's this was a story that got way out ahead of execution. So they just have to execute and I think all will be fine, but it is a more than a few quarter story turnaround. All right, everybody, thanks so much for being part of this week. Pat, go crush it on the intertube webs, wherever you are. I hope all the bots watching CNBC enjoy you. Just kidding. Everybody hit subscribe, be part of our community. There's always so much going on here. We appreciate your support. We'll see you all later. Bye-bye.

‍

Patrick Moorhead:

We done.