136. NVIDIA Earnings and the AI Boom Gets Bigger

Hello. Welcome to the Cube Pod episode 136. I'm John Furry with Dave Volante for the weekly Cube Pod. Dave, salute position. Hey, >> buddy. >> Salute. >> How you doing, John? >> All right. Oh my god, what a week. >> The AI boom just got bigger, Dave. >> So, it's we're boom town territory here. A lot of picks and shovels being sold in the old model. Now, Nvidia is selling not only picks and shovels, but they're building the railroad, the stations, and the destination on maps. So Nvidia full takeover mode. >> It's all about Nvidia full stack ambition. This week we saw SAS come back with Salesforce and rise of the agents. A lot of action and again I don't know where to start. I mean boom >> strike octa. I mean it's funny John, right? I mean the analysts realized wow we got to raise our estimates. We we were undervaluing Nvidia. I've seen estimates now 400. I mean, first of all, monster monster quarter uh 90 something billion 60 billion in operating cap uh profit up their estimates next quarter 108 billion just you know again and that's not even including any China business right so remember that whole China thing you know over a year ago yeah that's that's real impact everyone's sold out vera rubers are in full production you're starting to see a lot more action and this is I mean to me the theme is we're moving beyond the question is AI I real the question that we need to talk about is who captures the economics of the AI as the architecture expands from chips to systems, systems to agents, agents to the physical world. It's all coming together, Dave. The economics, the boom is happening and and if you look at it, it's the it's the buildout phase. It was happening. It's still happening, but it's moved on to execution and control. You're starting to see, you know, tokens equals profit. We heard that Nvidia remains at the center with their earnings and I put a post out on their next mountain they're going to climb at the edge we'll get into. But they remain at the center. The story is getting much bigger than GPUs. AI factory, networking, inference, agents, robotics, software, economics, security, and political backlash are all converging. This is the AI boom is broadening and maybe not breaking. and Nvidia's numbers crushed it and Salesforce the so-called sac SAS apocalypse >> they just gave the middle finger to all those pundits >> talk about sales >> you know I mean all the pundits saying oh my god SAS is going to replace >> uh this about that was >> that was interesting with uh with Dario but before we get off of Nvidia I mean you wrote a post you laid out sort of I mean 106% growth on 96.2 2 billion. A couple of things really stood out to me. Data center revenue was up 117% like 87 billion. But that was interesting that the hyperscalers revenue is about half of that. And and then everybody else, Neoclouds and Enterprises, but that second cohort is growing significantly faster than the hyperscalers, which is also growing like crazy. And I love the fact that Colette Crest like right off the bat at the call she said we're we're forecasting 70% growth for next year whereas cons consensus was like in the mid40s. So that was a sort of a mic drop moment because remember as soon as they announced the stock went down it went down like two or three%. Yeah. >> And then of course they heard that and then the thing started rocketing. It goes down back down today and people are like oh it's you know double top and you know >> no there's no double top. Look at the look at the fiveyear trend. It's right straight up a little bit of a bump, but >> it's un well they're saying it would fail to get back to whatever 20. >> It was close. I mean, it's going to get there >> operating but there was some interesting stuff going on in the numbers that people are pro maybe, you know, concerned about, but but I I want to dig into it a little bit. Their their operating cash was like 42 billion in the quarter, but their operating cash flows were down as a percentage of revenue and their day sales outstanding. and Colette said they're out to 60 days and people, you know, they're a little concerned about that. I'm not the least concerned about that because what they're doing is they're helping the ecosystem. They're helping the non-hyperscalers who don't have tons of cash. They say, "Hey, we'll push that out a little bit. We have the cash, >> so we're going to extend the terms a little bit." Um, I mean, they're basically a 400 billion dollar company growing at 70%. I mean, what else can you say to that? And the other really powerful takeaway in that call, if people didn't pick it up, was was Jensen and and actually Colette said this first that if you go back to Hopper, it was like monetization was $18 billion per gigawatt for their customers. And then Blackwell, $2 billion per gigawatt. And then Vera Rubin, they're saying, is going to be $40 billion per gigawatt. And then one of the it was I think it was Aaron Rakers my my my business friend Aaron Rakers asked on the call well can we expect that to continue and and Jensen goes great question and of course the answer was yes I mean it's basically the new Moors law and you like you said Reubin's in full production Gro 3 LPX is in production um and inventories are up people are criticizing for that but inventories are up because they're gearing up for the Vera Rubin and so and they're debt is up a little bit but it was a absolute blowout quarter. There is no demand problem. I'm so not worried at this point about circular finance financing competition. AWS who's supposedly competitive with tranium is going to do like like two million GPUs that they're buying from Nvidia. So obviously they need them and I think the the the memory prices obviously are a concern but the number one issue none of that concerns me. The number one issue is still the productive utilization by enterprises and that's to your point is what we ultimately have to see. >> Yeah. >> You know or this bubble might burst but no no way for a while. >> The Yeah. I mean I think it's more of a monopolistic conversation. In fact, the Wall Street Journal had an article yesterday kind of hinting that people's concerns around monopolistic behavior and they canceled the revenue sharing deal and they they could just drop that like a bad habit because they don't need it, right? They have all the cash. But I think what you said when Jensen said yes, will that continue that reinforces that the buildout phase is now moving into execution and control and revenue and the fact that Vera Rubin's in full production as is in rock, you're going to start to see people getting their hands on this. So while people were waiting for the supply chain scarcity to kind of open up you they were really working on this AI native. So I think why I brought up the Salesforce numbers because it ties into the the the trend of buildout to execution and again you've been doing a lot of research on the systems of execution uh and that's building on systems of intelligence. So okay Nvidia crush expectations and remains you know somewhat capacity extreme but they are shipping units to the big players that matter. I think Salesforce earnings, yeah, they have a relationship with Anthropic. They were an investor. Claude Force, clever marketing. Good job, Bengh. They beat earnings. That bumps it up. But what I think that shows, Dave, is that the the simplistic, you know, rationale of that SAS apocalypse narrative is is dead, right? infrastructure demand remains extraordinary while the incumbent enterprise software companies are beginning to demonstrate that AI can expand rather than simply destroy their business. So to me the angle on the Cupid pod here is that the AI boom isn't ending software per se. I always say bad software always dies, but it's forcing software to become AI native while simultaneously creating an entirely new infrastructure stack. So whoever can take advantage of that and it's what we said if you remember our analysis on the SAS apocalypse was exactly the same thing they got the moes they got the data but they have to go AI native and that is based upon the new infrastructure stack we talked about this this week when we launched uh the exclusive story of Cisco's relationship with Nvidia conventional networking meets AI networking is what I called it that's basically what's happening that's going to open up the enterprise now you can connect campuses other things are happening so you're starting to see the transition ition of you know the core infrastructure buildout neo clouds now called AI clouds coreweave doesn't use that word neo cloud anymore whe they use AI cloud whatever you're going to call it it's the new cloud so that's done now we're moving into the enterprise so I think the salesforce numbers point to the opportunity in the enterprise because once this new infrastructure is in place and you can connect conventional compute conventional networking conventional storage distributed computing architectures to the new AI stack with you variety of approaches. You got Nvidia, you got Google, they all do the same thing. They pump out tokens, right? Tokens drives everything. That's the mathematics. Mathematics is getting commoditized. The models are getting commoditized. But the data is not getting commoditized. The data is becoming more valuable. So, whoever can move that new data that's now converted into math into into the math engines drives the stack. So the AI stacks are fully different. They're mathbased. They're AI based. That's done. Jensen's data points to it. Nvidia's numbers highlights it. The demand for it is off the charts. That means the entire industry will move this way. Now they'll keep existing stuff, but it will be AI native. Dave, that's my take. Because when you go to the edge, which we put out a post, that's going to change it, but it's still the same game as distributed computing, disagregated infrastructure. So I think that the Salesforce numbers not just show that they're a good company but the AI native SAS play has to be there and if not they're dead. So if you're a SAS company or you're using SAS companies products it's a simple simple binary analysis are they AI native or are they not? If they're not they'll probably be dead. So to me this is going to be a freight train and again Nvidia is laying the tracks down. They're providing the picks and shovels, laying the tracks, building the engines, the locomotives, the the stations to jump on, and the ultimately the destination. So, Nvidia, I think, is never going to die anytime soon in terms of their growth. And again, Jensen, you know, over the top. Yeah, of course we're going to do 70% growth. That's insane. I mean, Andy Jass used to tell us all the time, well, it's growth on the bigger number. Yeah, that's Nvidia. >> I mean, hello. >> What's the What's the question? Who are we wrong? Come on. What's >> What's up? >> It's funny. I mean, people still sort of nervous about, you know, Nvidia. I mean, of course, it's the market. I want to unpack some of the things you were just talking about. So, Ben off came on with um with Dario from Anthropic, and this is something that we've been debating with George Gilbert and David Floyer, you know, rest soul. Um Floyer's argument was always and I kind of was very intrigued by this was always that the um the the frontier model vendors absolutely have to get into the system of intelligence. They have to grab a big piece of the software stack. And George has always said they don't to your point they don't have the data. They may have some data but they don't have the enterprise data and they don't have the underlying application logic and it's not in their DNA. And Floyer's point was always, look, it's too complicated. AI is too difficult. The the the Frontier models are going to make it simpler. And what they're going to do is they're either going to acquire or partner or both with companies like Salesforce. And that's exactly what we saw the other day. Now, the question is, will they be able to, as Alex Karp says, steal their alpha. I believe that that they absolutely intend to do that with a Trojan horse strategy. And so, you know, but you know, of course, folks like Beni off and and and Ellison are no fools, but I really do believe that this is the first step toward simplification of AI. So, and Benov talked about we're deterministic, they're probabilistic. You got to bring those two worlds together. And then the question becomes it's it's unlikely it's probably very low probability that >> Salesforce is going to start inventing stochastic software >> like frontier model vendors. It's h it's it's there's a higher probability that the frontier model vendors will acquire the they're clearly doing that with FDEEs and they're going to learn more and more and more about that underlying process >> knowledge and process logic and that's something to to watch I think closely and I think as to use an Andy Jasse term in the fullness of time that the frontier models are going to grab more and more of that AI software stack and if they don't they're in big trouble. I mean, I love I love the um Alpha comment. Again, Alex Karp, I mean, if he sat here on the cube here in the studio, I'd love to go toe-to-toe with him and challenge him, but he's throwing haymakers. He just doesn't know what he's talking about. I mean, he's like throwing haymakers from his perspective. His assumption is everyone's stupid now in this market. What Salesforce has shown, again, this is going to play out. There are dumb people out there and dumb companies, but if you're a smart management team, you're not going to just give up your territory. So he's right under the assumption that if no one gets on board, they're going to just lose their turf because that's what taking territory is all about. But in a perfect world, in a competitive strategy world, I'm not going to let someone come and take my territory if I'm a big incumbent. So he's right on one hand, but his assumption is flawed. He's just assuming that, yeah, you're going to lose your alpha. Now, I think if he's if he means it as a prescriptive warning, like an open memo type blog post, yeah, he's kind of you can see his perspective, but that's clearly not going to happen. And if you're like an existing player, it's like, okay, what's it take to not to lose? Okay, in the modern era, and we just I think we just laid out that alpha. So, if someone wants to give up their alpha, to use his that term, then they're idiots, right? So, like, you know, this is where you got to assume that people are going to be smart, you know, Jeff Bezos had that phrase, you know, your margins, my opportunity from that perspective. I think Karp comes from that perspective. And I think he's not wrong if he says if you're not going to get on the right side of history here, you will lose your alpha. But I don't think that's a mandate as in that's going to happen as a matter of fact. And I think you'll see the people who have that control plane. They have that build out. They're leveraging the stack in the way that AI native does it. They'll just be better. Their product will evolve and you'll see things happen. Where the control points are in their business, they'll have to lock them down. And if they don't, someone else will. That's just the way the nature of the beast is. It's Darwinism at its best, Dave, right now. I mean, this is what's going on in the market. So, you know, he's got the haymakers on TV, but he's kind of right, but at the same time, not realistic. No one's going to lie down and say, "Hey, take my alpha. Go take it." >> You remember the other thing he said is every every conversation I'm having with enterprises, they're they're they're leery of the frontier models. I mean, he threw that FUD into the market and I was like listening to him like really >> suppose Anthropic is going to do a hundred billion this year. That's like 10x 8 10x what you're going to do. So obviously somebody's buying this stuff. >> I mean if a company gives their alpha to the frontier models that's on them but the frontier models aren't maveli in the sense of like gatesy and like we're going to take over their business maybe if they want to let them do it. But at some point the frontier models have to make a decision. What what line do they push where their existing growth can be maximized. Now if their only play is to take everyone's IP and put it into their cloud then that's like that's kind of just a disruptive mode. The question is is it a disruptive market or is it an accelerated market? And this is a debate we had with Nvidia and Nvidia is very much in the accelerated computing narrative and their version is simply no one has to lose to win. Like so this is the growth of the market. So the disruptive mindset that Alice Karp has is no he's throwing bombs at everybody. He's throwing moltoff cocktails at everyone. Okay, whatever. But that's not the market. I don't think it's a disruptive. Now some places maybe, but for the most sake, everybody wins. And like I said, if you're a SAS player or an existing market and you move to the AI native with the infrastructure that's developing, you can win big if you got those control points, if you got the data and you build your new modes, protect your alpha or change your alpha equation. And Nvidia is doing it out in plain sight. They just bought hugging face for 12.9 billion. >> Khan must be pissed. >> They got potential perplexity investment. They got Cisco rack scale collaboration, new robotics hardware and inference accelerator aimed at agents. I mean, you talk about Nvidia controlling the most important points around their accelerated computing models. There's no there's no clear example of how to win than adopting the Nvidia strategy. >> They just don't they just don't dominate. They're just owning the influence and all the control points around it. You want a you want a cheap way to invest in all these AI startups, just buy Nvidia. It's underpriced and it's investing in in everyone. But you know your point about maybe a new type of alpha I think is really interesting because >> you know alpha yes proprietary data proprietary process etc. But maybe there's a new type of alpha which is speed new operating model. And if these startups and new companies and emerging companies can balance the frontier model excellence with, you know, using open models where it's appropriate, they could build an operating model that gets them to market really, really fast that begins to dominate their sector. Yeah. And then it may be a winner take takesmost market. And that may be the new mo the new op the new alpha is is that sort of winner takemost software-like marginal economics for all companies. >> Yeah. And and I totally agree that that has to be looked at and and and and watched very closely. And on the heels of the Nvidia earnings, I wrote a post. We were comment at the beginning of the podcast. >> Yeah. Let's talk about that. Great post. >> And the important part of that post was not to be get a hot take on why Nvidia is winning. We already knew that a year ago. It's just now everyone's catching up. But if you they they're winning the AI factory game where intelligence is created. And if you look at intelligence as the new monetization opportunity, whether it's measured in gigawatts or or tokens, whatever, AI is manufacturing intelligence. These factories do. But how do you distribute it? So the important development is from this is that the AI factory um is escaping the hypers scale data center. Okay, meaning okay what's next after the big hypers scale data centers? Well, it's edge and smaller AI factories. So the Cisco Nvidia deal this week is evidence that rackscale AI architecture is moving towards the enterprises. You got sovereign clouds and the neoclouds as service opportunities. That's the next TAM expansion. And the enterprises can plug in to coreweave NScale very easily. They can go to Amazon. That's distributed computing. And in the Cisco deal, if you squint through that relationship, that's conventional networking, which is distributed computing, connecting campuses, going to the edge, telco. You got um the AI uh market in the enterprise which was looking a lot more like Amazon's outposts. There will be data centers in hospitals, right? There will be racks, but they won't be the monster racks. They'll be 30 kilowatts. >> So, so how do you get that done? That's all networking. So, that's the network is the computer. Again, back to Scott McNeely, right? Comment, the network is becoming the supercomput again. >> Said I should have called it cloud. >> So, so but it's interesting. That's but if you think about it, you could plug if you're an enterprise, you could plug into a service like a corewave or an nscale and get what looks like pure endtoend IT like environment. Amazon's already had it with VP um their their technology. So okay, if you're now have disagregated infrastructure, call it scale across scale across the network, you got to connect it. So Cisco's conventional networking, you got optical and photonic fabrics emerging. You're starting to see the next mountain that Nvidia will climb. Now Nvidia doesn't like to talk about it because they don't have product yet. They can barely serve the current market. So that to me is a really interesting distinction. We kind of brought it up at MWC last year. This year at MWC, I'm certainly at supercomputing and OCP, Dave. It's going to be top of mind of everybody. >> Well, I I I thought your post was quite visionary. What I liked about it is, you know, everybody is going to talk about the quarter, you know, like I was just given all this, you know, spewing all the stats and the data center and half is is NeoClouds, etc. What I liked about your post is it was forwardlooking. You're basically, you know, looking at okay, the market is for inference is growing and that's going to drive a shift from centralized to distributed was your kind of core thesis. And then you laid out, which I thought was brilliant, this this density mismatch. And and I didn't I didn't have the numbers, but you said that modern AI racks need like what 140 kilowatts of power, I think you said, but the Telos are capped at like 30 to 50 kilowatts per rack. Y >> um and so the the idea and I guess you talked to Nvidia and Cisco about this is to disagregate that that that compute and use high-speed networking with optical and and high-speed fabrics and that to your point was a TAM expansion market creation and then you started talking about robotics which was that is the the next huge wave. So this is why I think Nvidia is going to be like I think they be a 10 trillion plus company. I mean >> I mean they their one of their big announcements in the earnings was the robotics piece and then the weekend before the earnings unit out of China showed that impressive robotics demo robots playing tennis marching like Attack of the Clones on Star Wars. It's like I was like what world am I living in? They're crushing it. >> They're fast faster than Hussein Bolt. But of course, you know, we can make planes that go fast, too. But but still, that was pretty impressive that the the Chinese, you know, uh humanoid robot games. >> Yeah. Well, the the point about that post though was also to to point out that, you know, the partnership between Cisco Nvidia is a normal partnership, but the the networking is key. And all the all the discussion over the AI factories the past three years past two years specifically has been all about this AI cycle this first cycle obsessing over footprint power land and shell GPUs cost of GPUs and I think last podcast you talk about the bottlenecks right the next bottleneck is going to be networking memory power inference economics and distributed execution so by post was trying to tease out that okay if you're going to put a factory node in call it a node on the new factory network It's not going to be as big. So where do you put it? If you're like a telco, you got a building, you got networking and power and facility. So they have land power and shell. It's just not that big and it's not a lot of power. But it's also they have like a bunch of other towers or central offices. Enterprise have multiple buildings. So the disagregated power and disagregated footprint can be managed with intelligence if it's connected to a factory big factory. So I think the I think the rack scale factory systems are going to be this big wave now. Then the smaller rack scale AI factories that are tied into the footprint requirements are going to come into production. I think once that happens phase one is over and the system becomes uh the unit of the computing in phase two the totality of the connected factories. So if you connect all the factories, you have a phase two uh model that says what is the total gigawatts per per revenue per gigawatts. That's why I think Nvidia is confident because they know and they're working on it. We saw them at MWC. We've been in the briefings. They're not saying it directly, but if you connect the dots, you know, OAN, the demo they showed in San Jose at GTC, they had a factory powering Metro Network blanketing San Jose. the RF was connected to a factory that shows the infrastructure and then the Salesforce is just pointing to the fact that the next app in the in these regions are going to be powered by that infrastructure. So again, the stacks will be different, footprints will be different, but it'll be connect. It'll be a connected system that's going to be very interesting to watch over the next 12 to 24 months. Who owns those control planes? How do agents sit on them? That's all going to be a big kind of in the weeds and networking is key and Nvidia is working on it. You talk to anyone inside the ropes, we do, they're all like networking, networking, how do I get this fast from this here? You know, >> it is the new bottleneck, right? I mean, it used to be used to be spinning disc. Flash changed that. And then, of course, STX is is >> and Blue and Bluefield what the Bluefield initiates all about storage. If you look at the NVL72 rack, it's a monster rack. >> Almost half the rack is switches. >> That's not compute. So, they have to put this the nicks in there interface cards to kind of manage the connections from the GPUs to make sure they get the low latency. And then the KV cache is working, all that dynamo work. So again all that system in the one factory you multiply that complexity by like you know hundreds and hundreds of factory nodes. Okay you can put software to work there too. So CUDA becomes important the evolution of the software is going to give you hedge a hedge against the the supply chain churn. So as they're waiting for the next chip to come CUDA fills the fills the void. And that's the same thing we're seeing with companies like Alterara who does FPGAAS. Um so a company Resolite which does photonix fabrics they're knocking down massive innovation there. So again I think the the AI anxiety is going to be lightened up a bit when the economics hit and then you know the politics of the data center. Don't get me started on that but you know the economic when the economics start coming in to your point two pods ago >> then the idea of what we're in is reality. But I mean talking about the Cisco um great job by the way there. I mean to me the the the significance of that Cisco Nvidia relationship is you've got all this sort of general purpose enterprise IT infrastructure out there and you've got this new thing called accelerated computing and what they're doing is they're creating a bridge. It's almost reminds me of the the sort of Nvidia and Intel you know investment and relationship. um it's sort of a bridge to accelerated computing and and CUDA and and so you've got you know trillions of of value that's installed with general purpose computing and I think it's going to be absorbed by accelerated computing and this is just another sort of I think proof point and that relationship is critical and Cisco becomes you know overnight like exceedingly relevant again. >> Yeah. and and the Cisco case here the networking and if you move up the stack on top of the factories you got the AI native worlds we're always referencing companies like fireworks AI but when you look at some of the um execution environments where AI is rendering itself agents are top of mind I love the physical robotics physical AI because that's going to be a tell sign but if you look at like the cloud and the enterprise and the desktop and the device and the physical side of that world not the robotics you're looking at trends we reported this week AWS acquired duck labs around their duct DB and their embedded analytics. Perplexi launched an ondevice agent clean as a desktop workspace. Anthropics advanced both memory machine generated agent standards. T Thompson Reuters launched a proprietary model for legal work. Google announced something for legal work. Obviously Harvey's the winning company today. All of that points to the fact that if you look at the the dis the fragmentation of the environment that's interesting because you have different approaches. Everyone's trying to figure out where the puck will be on the execution side. What do the AI native apps look like? Are they are they asset light? A new term we're hearing a lot about. I don't want the capex. I'm just going to call a large capex cloud. So what is that execution environment look like? Is it fragmenting? Is that a good thing? Is it a feature? Is it a bug? So, this is going to be an interesting question, Dave, because at the end of the day, as we get more horsepower, we'll see the models emerge, the successes will emerge. So, I I'm I don't know how I feel about this. All I know is that the guy's changed. It's the prompt is get me third quarter sales projections versus writing the logic in a SQL query, which is the logic you're typing out of your hands from your brain, versus saying what you want and having the logic happen in the AI. That is a completely fundamental turn in the b in the data business. >> That weird example is just now happening at scale. That is a that is that is a user change. And that's why these headless systems are becoming popular. Meaning >> right >> I don't need software >> to go to the user because whatever they want, I'll just respond to whatever they want and provide intelligence and I'll figure out the business logic based upon the data. So that's going to change the front ends completely >> and the client surface is you know completely changing to your point and is also a closed loop with the back end what we call the system of intelligence or the ontology or you know the context you know semantic layer whatever you want to call it we call it the system of intelligence but that closed loop is critical and that's why I love what data bricks is doing because they've got genie in the front end and they're building the system of intelligence at the back end and they've got the agent execution environment. They've got the ontology and it's just they nailed the the strategy. The one other SAS apocalypse company that we have to call out here kind of got lost in all this um Nvidia news and and and Salesforce was CrowdStrike. CrowdStrike absolutely crushed its quarter. I mean, it blew away all expectations. They're almost a $6 billion ending ARR company now, growing 25%. that they they did 333 million of net new ARR and they're what the the fundamental premise here is and they said this George Kurr said it was coming in in their Q1 he said the mythos you know event is going to be such a tailwind and it's exactly what happened the postthos fear >> has turned into cash that chaos has turned into cash for companies like Crowdstrike and and presumably PaloAlto as well. But CrowdStrike upped its guidance, new records. It it it increased its its uh outlook for for fiscal 27 by substantial 630 basis points, their CFO said. So I'm writing that up as part of my breaking analysis today. But the cool thing that I want to share with you in the audience is I'm I'm introducing this week a new firm called Qualitate. So Qualitate is an AI native. It's a it's a data intelligence primary data platform and the premise is that public data you know it's declining in value first party data proprietary data is is escalating in value. So they've automated this expert interview uh across a network and for doing market research for investors, you know, primarily and corporate strategy people. And the company's built this incredibly intelligent AI moderator and they're capturing these expert insights at massive scale. Any industry, of course, tech is part of that, but but robotics, semiconductors, healthcare, you know, you name it. And interestingly, John, the firm was started by Saga Kadakia, who was the one of the early employees at ETR. He was the head of data science there and one of one of the colleagues that I originally collaborated with. >> Yeah. >> And so, you know, they're they're they're bringing a lot of nonobvious proprietary intelligence to the to the clients. And I I was able to they were allow me to cherrypick some of the data um this week for the breaking analysis. And then um they've got a survey platform that I'm gonna gonna test out and I got a demo of it the other day. It's unbelievable. You just you prompt what who you want to survey, what you want to survey, and they've got, you know, tens of thousands of instances and you can just pull up you want Cisco customers, you want Oracle customers, you want Crowdstrike customers, you want Crowdstrike customers that also have Microsoft and you can just query them in record time at a substantially lower cost. >> This is the difference between modern >> going to take off. This is the difference between modern and old because you mentioned uh and I mentioned that SQL query. So for the people out there that don't know SQL, it's like you query the database. You have to figure out the logic be in the query while you're writing a longass query to query a big database. It's got tons of rows and columns and the work done was the by the by the data scientists to actually create the query. That's the business logic. Okay, you have to basically figure out the business logic first. Okay, then your objective, then the business logic that maps to the schema of the database and you go and it runs and it cranks an answer. You get a report. The new way is just give me third quarter sales or hey give me what what's crowd strike with uh embedded into a Microsoft account. I'm making this up. So it just the logic is built into the AI. So the modern era goes the way the latter which is let your outcome your that you want be the query. That's a prompt. It does its prefill and decode. All the mathematics kick in with the AI engines. You feed it more math, it gets more math and and the survey work is going to be key. So then it's going to be faster decision-making and that's going to make companies perform better and and and what you'll see is lagards and winners. You mentioned CrowdStrike. Okay. I was just talking with someone this morning about this. Their market cap is now 222 billion as of close of today. Okay. The the the curve goes like this. you know, even with their little dip in their um event in in 2024 went from 97 plummeted to 76. We talked to George Curtz and it's just peaks if you look at the technical analysis. Compare by contrast Zcaler. It hasn't moved. Those are two peers, right? So Zcaler is actually down from 2021 one. They've lost value. They're peers. That's funny. >> They're peers. >> So, okay. So, so you know, if you're a company out there, you do not want to have a curve like that, okay? You need to look like Crowd Strike. Let's look at Peloto. You know, Peloto, Peloto looking really damn good. Better than Crowdstrike, a little up to the right, but you know, 300 billion, >> but they popped heavily since 2026, March of this year. They're they're huge almost. They're up from 2021, but remember they made that platform shift. >> Yeah. Platformization. Remember too that CrowdStrike still hasn't recovered fully from that incident a couple years ago in 20 what is 2024 now? They still haven't fully recovered. Now, at the time, customers were pissed. They were saying, "We're going to leave. We're going to leave." Crowd Strike played it. >> They didn't, but Crowd Crowdstrike played it brilliantly. They they had this thing called Falcon Flex. They had it before the incident, but then they deployed it. And their whole emphasis on their go to market was keep the customers happy. Don't let them churn. Use Falcon Flex. Allow them to use these new modules. On board them, make price concessions. Whatever it takes. That was their go to market motion. >> Yeah. And remember >> earlier this year, >> if you go back to our Cube Pod, by the way, we have that analysis and we predicted >> that they would not not only lose, they would maintain and take correct more territory. >> That's exactly right. And and that's what happened. And then earlier this year, the signal was to the sales team, okay, we're through that now. We got through the knot hole there. We want to use Falcon Flex to really drive new logos and new growth. And that's exactly what happened. And it timed was timing was perfect because they initiated this is at their SKO I believe this year earlier this year and then what happened was the methos event and it was just it was just this gift for companies like crowdstrike who were consolidating and same with PaloAlto. Now the one thing I will say is customers they're they're like I say they're in chaos and so they definitely migrating toward we saw this to crowd strike. I think George Curts said to Kramer the other day that still customers mo many customers still don't get it. Of course, he's going to say that he's touting book and I'm going to talk to him next week. >> But here's the thing. Customers should absolutely consolidate should simplify. But when you do that and you give more control and more data and more telemetry to a a consolidated platform, your blast the impact of your blast radius, you got to think about it, right? because your blast radius is now way more you know tied to that one platform and so the trust bar raises dramatically. So people want to make sure that they coming back to sovereignty amid governit you've got to make sure you've got financial sovereignty operational sovereignty your tech stack etc that you've got control in case something goes wrong. Yeah, I mean it's I I think it's going to be big thing, you know, the Mythos um event from Anthropic and there's a project called Glass Wing that came out and that's been kind of like they had they took it they did the right thing. They went out to all the security companies and all the major platform players that were relevant and said >> we'll give you this advance and every single company that I've interviewed as part of that program has said they found significant vulnerabilities. In some case the word was we thought we were Swiss cheese before but we had no idea how Swiss we were in there huge gaping holes they found them they fixed them now they get updates and that's a big deal but the thing about the mythos is Dave it's as big as chat the chat GPT moment so in the security business mythos is the chat GPT moment for security because we all know when we got chat GPT how good it was the reverse it's like oh my god this is so good oh my god it's so had for us. So it is a it has been a call to arms. So obviously one it highlights the state of the union of cyber security is not high as healthy as we everyone thought it was be maybe was false comfort but it exposed and so it puts crowd strike everyone in the sector that's why I brought up zcaler because I'm looking at all the performance of these companies and the ones that don't have a kickup are the ones that are either getting killed by mythos or aren't executing on what the current landscape is. It's kind of like fighting warfare with drones, with Patriot missiles, right? Like, you know, you don't do that. That's something that we're seeing in the defense tech. These new the new battlefield is different. The rules of engagement are different. And with with the models, it just changes the surface area is a long tale of a lot of vulnerabilities. All you need to do is crack one spot and you're in and now you got agents which are like drones like everywhere. So agents could go off the rails. They could be manipulated. So, Egentic and Methos together really line up as a call to arms for the cyber security industry because it is just a massive problem opportunity. Everyone wants to deploy models. They want to have the data platforms. They're going to have a lot of agents. And guess what? >> It's unsecure. >> Well, >> that's going to be a huge opportunity for CrowdStrike, PaloAlto, and all the different vendors. And by the way, Obsidian, a company we've been in, same thing. a lot of different new brands might pop up. So zero trust, endpoint protection, all good. I call that the patriot missile strategy. You know, like they're built for they even cyber resilience back up and recovery based upon old landscape. So I think it's a whole reality and you know Sanjay Pun was in today talking to me about this. He's the CEO of Cohesidity. You know cube alumni he passes hello to you but he said the same thing. He's like look we have to be very integrated into glass swing. We got to have resiliency not be a just recovery because these agents are going to be having all these little paper cuts that are vulnerabilities. How do you roll back from that? So, so the the whole methodology is changing. The mechanism have changed, but the game is even worse. >> 100%. AI detection and response is, you know, now the big thing. And you remember everything's changed. You're absolutely right. You remember patch Tuesday? >> Yeah. Patch Tuesday, you know, Microsoft and others would put out their patches on the second Tuesday of every month. Patch Tuesday used to mean, you know, hackers Wednesday. And so you had this 24hour window. The window there's no windows anymore. You have agents continuously doing updates, finding vulnerabilities, updating, patching. But the there is a new type of window now that occurs. And we kind of learned this from the crowd strike incident which is you want a window where you're doing planned rollouts. You're you're you're doing them in phases in case you have to roll back. So there's a window in which you're doing you know kind of your initial test you kind of spring training and if you something goes wrong you've got to be able to roll back. So the whole concept of that window that patch window is changing. So that's that's new. And the other thing is you know the data from qualitate suggests that people generally when they think about you know mythos and how to respond they're coming at it from application security. So you think you know code but the reality is they then immediately want to know okay if I got vulnerabilities in my code where else am I vulnerable what other parts of my my my application estate and are could be compromised. So now posture management becomes a whole big thing that was the you know what whiz popularized right and then that leads to cloud security and identity and I mean it just all the crowdstrike's got all these modules that they are enabling >> and it's just this incredible flywheel and you're seeing you know I mean they've got multiple modules that could be you know IPOs in of themselves and of course they're all in one platform. It's interesting because PaloAlto has a different strategy was much more acquisitive, but they're a great product company. So, they do the engineering work and and and do the integration. Whereas CrowdStrike, you know, generally was sort of always a platform company. They do acquisitions. They're very acquisitive, but they they do these tuckins that go into the >> into the uh into the platform. They did a an acquisition called Pangia, which is I think identity. So, that's you know, agent identity is going to be really important. They did another one called Signal SGNL. Um I'm want to learn more about that next week at Falcon. >> Well, we got a big week coming up next week. Um got VMware Explore. You're going to be at Crowd Strike. I'm going to be in New York. Um a lot of action happening on the earnings front. Um PaloAlto, GitLab are reporting. Snowflake, Broadcom, HPE, NetApp, uh and then Dell, UiPath and Zcaler. So, you know, I'm I'm checking out these companies. I didn't mean to call out Zcala, but is notable in comparison to the crowd strike you mentioned, but you know, Dell, I mean, Dell, I don't know if you saw the news, but JP Morgan picked up a ton of stock as as a equity holder. Now, Dell's kicking ass. Um, NetApp has some storage action going on with Nvidia. So does HPE. All of these companies um have a sovereign angle, right? And I think you're gonna see a lot more cyber security. I think the convergence of cyber um cyber security AI infrastructure and cloudnative sovereignty like where's the domain of my network what data stays in where's the revenue I think sovereignty is going to be a whole new word that'll be redefined in the next 12 to 24 months because everyone is talking about how to use their infrastructure and intelligence for sovereignty so you know we'll watch these reports Dave it to see how Palo Alto is certainly during the week of crowd strike you'll get the data there get the scoop and you know for sure Broadcom will be very >> that's interesting I mean Nvidia blowout you know you got to figure that means that's a good news for Dell right yeah because >> Dell is essentially a distribution channel for Nvidia AI factories you know obviously they add value to them but I mean they are the number one you know the you know engine computing company that >> they nailed they nailed the rack scale. I mean what what where what Dell deserves credit is they understood the timing of the software stack which we know we were close to Dell and they looked for our advice and also we watched them build out. They knew there was no real stack yet but they were first on the packaging of rack scale systems you know um liquid cooled turnkey they ship essentially a rack scale system that takes turns on in six hours. In fact, the word stand up is not even being used anymore. The old old language was stand up some infrastructure. You know why they say stand up? Because it takes weeks and months to do that. Now you now you ship a rack from Dell and it ships on the loading dock as the rack. It's a big ass rack. You come in and you turn it on basically. Now it's not like turning on the lights. It's six hours of turn on but they're not standing up anything. It's stood up. So the words from coreweave, the words from Nvidia, the words in the industries lexicon's changing the rack and stack language is over. It's turn on, not stand up. So very interesting nuance. I thought I'd share that with you because I I see I've seen that coming out of all the interviews. No, no, we don't stand up, we turn on. And like okay, I'm corre thanks I stand corrected. Thank you very much. But there it's interesting that you think about it that way. the scale of of turning on infrastructure. >> That's why I wrote the small rack at the edges. You drop a box in, plug it into the distributed AI factory network, then you have an intelligence network. So, it's like a really I think this is not being reported at all. I think we're first to even call it out, but it makes total sense. I don't think we're wrong on this one. >> Well, you've been talking, as have I, you've talked even more the neoclouds, and like you said, Corweave uses the term AI cloud. All the other Neoclouds don't mind Neo cloud, but the point is called >> Neolabs now that that's get the new word out, >> but they they are exceptional at infrastructure, right? I mean, you got >> people like Hen Goldberg and they make a strong case that look, >> you know, we're starting with a, you know, green field >> and the hyperscalers got to bolt this stuff on and we're going to move faster and we're focused and you've seen this kind of disruption before. Of course, I don't not predicting the hyperscalers go away, but I think I think many of the the neo clouds are going to do quite well. Not all of them. >> Well, coreweave core has their first event. I interviewed Gene English as a preview. She's the CMO cube alumni. She was at Juniper before that and you a variety of other companies. Um, industry legend, veteran. She knows the business. But coreweave is interesting. She said they have their first corp uh conference, their user developer conference, whatever you want. It's not really a developer, but it's everything allin-one. They have practitioners there. This is their first event. Their numbers were off the charts, too, on their earnings a couple weeks ago. So, >> end of se end of September, right? >> Yeah. Yeah. September, I think 30th to 31st. Um, and >> we're going to be there. Yeah, >> we're going to be covering it live. But if you look at what they've done, they've built a full >> John Dave show, right? >> We'll see. Yeah, I think that's Yeah. Oh, yeah, definitely. Yeah, >> I'm coming. >> I think it was an IPO that day um that we might have to cover. So, >> maybe I'm slow. I was supposed to be at Cyos. I was supposed to be going to Cyos, but I'm going to go to the Cororeweave. >> I'm I'm scheduled to go to Coreweee. So, unless it's an early >> the big banking show in Miami. I was scheduled to go, but the Cororeweave thing is come up. So, >> well, I by the way, speaking of Crowdstrike, I just saw news come across the wire from CNBC. Crowd Strike CEO exposes dangerous cyber gaps that legacy tools can't handle. So, go get the scoop at Crowd Strike. We'll be we'll be uh getting all the VMware. I had a pre-briefing from all the execs. You're going to you're going to see AI factories thematically all throughout. And again, IT environments are going to like what virtualization did for cloud. You're going to start to see a similar AI native >> Yeah. >> infrastructure be moving very very fast. So, we're watching Nvidia closely, the Neoclouds, and everybody else. >> I got that preview, too. I I would just say this. Obviously, it's it's embargoed, but so I can't say too much, but you saw it. I'll just say this as we sort of reported when we had Charlie Kawas on people just they don't understand Broadcom. They think Broadcom is just this like PE that sucks dry all the all the profit, but they invest. They are an engineering company. Yeah. >> And you can see that in what's coming at at VMware Explorer. >> Yeah. And and by the way, they're working with everybody who's building their own chips. So yeah, you know, you got to know anthropics building their own chips. So, you know, I hear scuttlebutt on that. I haven't confirmed it, but no one's on the record. >> And yeah, let them, you know, you heard Jensen's response to that. Yep. Okay, >> good luck with that. >> That's good. They still buying from us. They have to buy from Nvidia. Nvidia is the highest performance, lowest cost, best product. You heard him on the call. We got the best training, the best inference, the best robotics, the best libraries, the best software. And it's true. >> Yeah, it's true. I mean, we're in a we're in a better mousetrap wins market, Dave. The better product wins in these inflection transitions. So, you know, to end on the Nvidia note, they're they're going to they're going to be very be interested to see what happens over the next few quarters now that Vera Rubin's shipping now the optionality on the different paro curves gives you now tokens at different price points and you got the open source models and the open weights. I think it's going to be a really great era for the neoclouds and the enterprise as they gear up. Yeah, I think the agent stuff will get rained in with guardrails to begin with, but it's going to be very interesting to see the impact to productivity and revenue for these companies. And I think Salesforce is a tell sign that you got to get your app right AI native with the new AI infrastructure. So to me, I think that's the story to watch and of course Nvidia's land grab dominance will be one we'll watch as well. But you know, the demand is so high. It's really not a monopoly when you have the the new car and everyone's on horse and buggy, right? That's what's happening. >> Exactly. >> Wait a minute. You're a monopoly. The auto the horses are pissed. >> I haven't seen, you know, Yes, I do believe they have a monopoly, but I haven't seen egregious monopolistic behavior. Hugging face is interesting. >> Yeah. But but Dave, it's the analysis is simple. They're the automobile manufacturer and the argument is like you're putting the horse and buggy out of business. That's terrible. Like, oh my god, you're evil. Like, okay, what? >> You know, there's still more more people coming in. You got AMD, you got Google, you got Cerebras, you have compute still booming. So, I think, you know, the say Nvidia's monopoly is ridiculous. Now, they're mopping up because they got a better product. But it literally is like comparing horse and buggy to the automobile. >> Yeah, that's true. But and and there is a lot of competition. And I think the competition is a good thing. I think that what people I I feel misunderstand is Nvidia is the new general purpose computing you know call it accelerated computing and the general purpose leader which is Nvidia is going to get most of the market just like Intel got most of the market you got companies like Cerebras and they'll compete at the the tip of the pyramid for sure and others but you know I mean Dmatrix we love we love SID and some really great innovations going on >> the scale flux is doing great scale flux >> D matrix you have a lot of the embedded systems again robotics will be a tell sign on the edge and you're already seeing a lot of innovation like I mentioned the unitry in China that's going to uh compel the American entrepreneurs and engineers to to level up I think you're going to be see a surge in robotics and the technology in robotics has to deal with low latency no zero hallucinations because that's safety issues you got graphs you got multiple databases tons of sens sensors, embedded systems, all run like an AI factory. So, it literally is an edge factory. I mean, a car or a robot. It's it's it's it will be the test. >> That's John's read John's post. It's right now it's on the front page of Silicon Angle. So, read it before it disappears. >> All right, Dave. Have a great weekend. See you next time. 136 is on the books or in the can or on the pod, whatever you want to call it. See you next time.