 Hello everyone, welcome to this podcast, episode 49, this Cube Pod, I'm John Furrier with Dave Vellante. Live from Barcelona, we're doing the Cube Pod from our studio here in Spain. Dave, we're doing a special podcast while we're here because we're flying back on our Friday recording day. We're doing the podcast here in Barcelona. Love it. It's hard charging, but it's not dry air like Vegas, but the voice is still from all the talking we're doing. We're continuing to talk with some of the interviews here at the Cube, great events for the Cube. So special pod, so let's get it going. We've got a lot of great Silicon angles to talk about because this show was about networking and Silicon. We had Broadcom's president, Charlie Quaz. We had Rami Rami, who started his career in Silicon at Juniper, and they were blown away. At one point, Rahim the CEO says to you, are you, do you know Silicon? He said, he asked me, are you a Silicon guy? I was like, no, I'm not, but I love Silicon. I've been following it forever. We've been covering Silicon chips since we started 14 years ago. I mean, Silicon angle, you can't get any more Silicon than in the name, Silicon angle. My first exposure to Silicon was when I, after I graduated from college, did a little programming, wasn't very good at it, started at IDC. Back then, IBM was everything. IBM was by far the leading semiconductor manufacturer in the world. And so we would go up, we would tour their facilities. I remember going to Essex Junction, Vermont, where they had this, you know, semiconductor facility and getting educated by these, you know, IBM research guys. So I mean, we go way back. When Silicon angle started, we've been covering chips. Silicon, from a data center perspective with Intel, AMD, Intel was dominant at that time. But then when ARM hit the scene, we chronicled the chip business from data center to cloud and we broke the Annapurna story. So before Amazon even bought Annapurna, I held in my hand the first chip. The guy from Israel was over at a reinvent, I think 2013. I think the point is we've always understood the importance of Silicon, right? It was interesting to hear Charlie Kawas talk about CMOS and what a game changer was. And I remember, I remember when IBM mainframes were running ECL, emitter couple logic and they transferred into CMOS, which was much low power. And they had to pay a performance penalty, but they knew they were on a new curve and they dominated that business. And of course, we know what happened with the PC revolution and Intel. And of course, it was really well documented in chip war. You got to get that book. The ARM stuff, we saw that early. I bring this up because someone made a comment to me the other night at the Cisco event and they said, do you guys cover Silicon chips? Well, we don't. And it was like, well, yeah. I mean, of course, we cover the industry. I thought you guys knew enterprise. The competitive dynamic, but how can you not cover Silicon if you're in enterprise? And so, and a lot of that too was David Floyer really knows this stuff deeply. He's an analyst emeritus at theCUBE Research and he was on ARM early. He, I'll pull it up, I'll show you the post. I remember. In 2011, he said Intel's in trouble. Here's why, PC volumes have peaked, started to educate us on Wright's law, said ARM is going to dominate, it did. And we've done a lot of work on this. And I'll just give you a quick stat. So, and people talked, one of the guests yesterday talked about everybody talks about tops. It's really about, it's really about performance per watt. That was Charlie Kowals interview. Was it Charlie? No, he didn't talk about it. You brought up tops. No, somebody else brought up tops. I forget who it was. He said the benchmark wasn't tops. It was another KPI. And I said it was, yeah, it was performance per watt, which is true. That's really the key metric. But I said tops are still important. And here's why. We all know Moore's law. Doubling the transistor density every, let's call it 24 months. If you do the math, that equates to about a 40% per annum improvement in performance. Okay, it's down now because Moore's law is waning, but let's call it 40%. If you look at what Apple has done, and Apple really got this started with the iPhone, and you compare or you combine the performance of the CPU, the GPU, the neural processing unit, all the accelerators, and you look at their progression of the A-series chip, which is inside the iPhone. And you're seeing the same thing with the M-series. John, remember, Moore's law, 40% per annum. The combinatorial factors of the XPU and Apple's chips are increasing at over 100% per year performance. 100%. Okay, so more than double Moore's law. And that's why I say tops matter. Now, Charlie's bringing up an interesting point, is it's got to be energy efficient, and I get that. But the point is, it's a whole new game in town. That's why people say Moore's law is dead. Is Moore's law dead? Well, essentially it's evolving, it's being replaced by the combination of all these XPUs. And that's the fascinating dynamic in the semiconductor industry. The last thing I'll say about it is volume still matters. This is where it's an interesting debate about merchant silicon versus, you know, custom silicon. But ARM has the volume right now. TSM has the volume right now. And that's why it's so hard for Intel to catch up. So, speaking of Charlie Kwas, who was from Broadcom on yesterday. So here's the quiz, pop quiz for you. He mentioned OSP. You remember what that said? Open scale power. Open scalable power. Power is the new constraint. So, we've been talking about clustered systems a lot here. And I think that's the new category. We're going to do a lot of research on this. But if you look at the new data center, or call it the new data center, but because data centers are shifting away from the old school data centers, rack stacks, machine servers, because moving the clusters. I mean, the cloud's a collection of servers. So, I mean, Larry Ellison will say the cloud was, it's just servers, right? So it's true. The data center of the future is on-premise and edge. Smaller, faster, cheaper, more as well. And cloud. And cloud, it's a cloud operation. So, public cloud on-premise and edge is distributed computing, operating as a cloud operation. Meaning, microservices, APIs, and now AI and generate AI stacks. So, okay, so that's going to happen. So, by the way, Charlie totally loved that idea and he validated that. Everyone does because it's happening. But power is the new constraint. So, there's no motherboard anymore of the constraint. It's power. How much power can you generate in the data center? Power and cooling. And then on the device, what's the power consumption relative to generating that AI or application? So, what's going to happen is if the game is still the same, it's just shifting distributedly. So, we talked about, with Charlie, the metaphor you used, I brought in highly available versus high availability. That's a storage concept. Highly available and high availability sound the same, but they're different. Highly available means you have data right there, ready to go, go. High availability means you go get it. Oh, it's available. So, high availability means I can get data. So, high availability means I got an application that needs data. It's available. Meaning, it's on a server somewhere else. You got to go fetch it. So, across the network. Highly available means it's on the device itself, low latency. So, when you talk about tactical warfare, like military operations, any kind of drone activity, the tactical edge they call it, or edge in this case, with consumers, if you need data instantly, it's got to be on the device. Not a query to a server. And that's what Qualcomm was talking about yesterday, right? Yeah, and that nuance, that's an old concept. That's a computer science concept that's been around for years. So, what's happening is these concepts are being recasted in a new architecture. So, these clusters matter. GPUs are all the sexy conversation right now, but it's Nix and Switches. So, Broadcom and NVIDIA are the only two companies that actually make both. Now, GPUs and NVIDIA has GPUs, so they got a little advantage. To your point, the GPUs are going to be only one element. So, NVIDIA's got a little bit of a lead on Broadcom on that point, hence the valuation. So, you're going to see a whole another generate, well, again, what does it mean? All of this means that an entire new generation of computer science and IT is coming fast. That is going down now. So, it's the same wine, new bottle, as the expression goes. So, all the same concepts are rolling into an entirely new generation of IT. So, IBM's all over it, all over it, and all the companies are all over it, and the telcos are like, well, we got the pipes. But we're not IT capable. So, that's why 5G's not working right now. So, 5G is awesome, you have great blanket coverage, but guess what? Making that enterprise ready? Not going to happen. I got to pick up on a couple of things you said there. It was a lot of knowledge you just dropped. So, I want to talk about something that I think I was asleep at the wheel on that we talked about on the last Qpod, where I've been saying, I feel like NVIDIA's got a substantial moat, and we debated, the industry is debating about, and I think there's a consensus, there's going to be all these alternatives that come out. AMD, Intel, things like LPUs, very specialized processors, and I have said, yeah, but none of them are going to have a monopoly, like NVIDIA. Well, where I was asleep at the wheel was Broadcom. Broadcom essentially has monopoly-like business. It's got very, very high gross margins, almost 70% gross margins, and of course, it's got VMware-like gross margins. It's getting better. So, I kind of missed that, just not really thinking about it, and I missed it because I was thinking narrowly about NVIDIA's GPU competitors. Okay, so, I think I would put Broadcom in the mix. They are the number two AI company by valuation. They're an AI company, right? Because AI is everything now, they're like perfectly situated for AI, even though Charlie Cowes would tell you, we weren't like trying to go after the AI curve. That wasn't our strategy. We were going after 10-year durable businesses, so that's one thing I wanted to mention. The other point is the low-latency edge, real-time data in the moment. From a silicon standpoint, Tesla is the best example. This is a company that said, we're not going to use Mobileye, Intel's chipset, right? They're silicon. We're going to develop our own custom silicon using ARM. Why did they do that? Because they wanted to get rid of LiDAR. Now, whether or not getting rid of LiDAR is the right thing, is the safest thing, I don't know. But I do know that they saved roughly $2,000 a vehicle by using low-cost cameras and developing their own custom chipsets with highly customizable NPU code that they developed themselves so that they could interpret what was going on and use low-cost cameras. And the last thing was interesting, I want to bring up. And who am I to argue with Charlie Cowes and Pat Gelsinger? What about chiplets? But Charlie admitted, he affirmed, when I said, chiplets are not anything new. I mean, Pat's marketing it like there's this new radical thing called chiplets. Chiplets have been around for 40 years. And Charlie said, yeah, that's true. Now, where? A lot of hand-waving there with Pat Gelsinger. I mean, are you saying it's smoking mirrors? No, it's not smoking mirrors. Definitely not. It's just good marketing. Pat's like, brilliant. And so, but he's making it sound like chiplets are this radical new thing. What is new is the silicon diversity that you're getting with the different types of XPUs. But I still contend that what NVIDIA has done, what Apple has done with taking all these XPUs and giving them access to a very large SRAM in that synchronous is different than having a bunch of asynchronous connections even though that's Broadcom's business, communicating through all those chiplets and all those different components. But they're still relative to what's happening inside the NVIDIA system. They're relatively slow and they are asynchronous connections. So, which is fine, there's nothing wrong with that. It's a huge market for that. It's a huge market for that. My point is, I think the NVIDIA approach of what people call monolithic, they make that a pejorative, I think that has a very long life because we're going to continue to be training big, giant data models. I don't think that's a pejorative. I think it's a feature. It's not a bug. It's not a feature. But my point is, we're going to need, and I think you'd agree, we're not like hitting up against the wall where we don't need anymore, we actually are hitting against the wall. It's not like we don't need more GPU power. The more GPU power you create, they're going to consume. That's kind of Jensen's law. I call it Jensen's law. The more you buy, the more you save. Jensen's law is buy more GPUs because we have software to run them. Buy more, save more. What Broadcom is getting at, and what we're talking about when we're at the HPC plus AI equation of supercomputing is that the supercomputing market is now mainstream because of clusters. We are seeing clustered systems. Monolithic is not a bug, it's a feature. So if you got a workload, you run it on a cluster that's going to service that workload. So collection of clusters is a distributed computing paradigm. And so even Charlie Coas and Broadcom, the president said the word superclusters yesterday. I asked him, why does AI need all these chips? And it's not just GPUs, it's combination of GPUs and chips around it. So GPUs will be one element in the clustered system management. So we are going through a systems revolution again. No question. So the shift is data center to cloud. Andy Jassy said data center is moving to the cloud. He was kind of right, but not 100%. Meaning cloud one, now cloud and on premise data center is going to invert into cloud operations which has to look like Amazon, which is clusters. So Rack and Stack the old way is over. I think the idea of super cloud will be a series of super chips, super clusters. And so all that's going that way. If you look at the GPU clouds, they're all clustered systems. So manage service, hey, you know who wins? The hosting providers, Equinix, anyone with power, anyone who has a data center is going to be in business. So I had even someone off the record said to me, we're going to close down our data center. Turned out good call, not to. They're in huge demand. So I think the scarce resource is going to be data center, facilities, power and cooling and GPUs. And to your point, you know. And that's going to define the economics. Yeah, and to your point. Well, okay, so let's talk about economics, but I want to just say to your point, Jensen will often say that his systems cost, they weigh 70 pounds, right? And they have hundreds of thousands of components in there. And what's interesting, well, from an economic standpoint, this is where I think it's a bit of a wildcard, particularly what Charlie was saying about power per watt as the key metric. And I don't know, I mean, what does that mean for Nvidia? Can it get its energy economics to the point where it can actually continue to grow? That's, I don't know. That's something we got to look into. I'm sure we're going to hear a lot about that at GTC in a couple of weeks. And Nvidia is just amazing, right? They started in gaming. They got the tailwind from crypto. And then when crypto tanked, you know, everything went over to AI and building LLMs and then chat GPT. Now crypto's back. I mean, basically Bitcoin is flirting with its all-time high. I think it hit 62,000 today or yesterday. I think it's all-time high is 64,000, right? So that's a big story too, John. I mean, I don't know. Well, I mean, all this converging, the nexus of AI and cloud is really about distributed computing. And why I'm so hardcore on this whole clustered systems approaches, we're doing a boatload of research on cube research area. The cube research team is going to build on our arm work that you guys have done and we've done the old private cloud cloud computing work to this new idea that cloud operations is going full distributed computing. Means it's not going to be like cloud versus data center anymore. It's going to be the cloud is the operating model. Period, full stop. Hybrid cloud is the standard. So I got Amazon or Azure. I'm going to have an on-premise data center-like experience, but it's not the old school data center. It's the AI system. It's the clustered systems. And then you got the Edge, which will be device specific and or designed engineered systems that's connected to the cloud or clusters that are serving it. So again, everything's going to work in tandem in harmony. So it is basically mainframe. So if you get this idea like, okay, that's why the telcos are struggling because if they don't lean into being part of the system, they're going to be just the pipes. And so they'll have device access. So, you know, the dumb pipes, are they going to be the dumb pipes? Connectivity, that's what they do really well. And you know, it's a good business for them, although they're certainly chirping now in a big way. It's only one with dumb pipes, by the way. Give me an API, I'll run traffic for you. And it's a lucrative business and it's essentially a utility. The problem is that you're for the European telcos, especially there's a $200 billion CapEx shortfall, they're saying. And we're certainly hearing at this show that people are pumping the brakes on 5G, i.e. the CapEx players, the telcos, are slowing down their investments. You know, I don't know. So, Amazon- Let's do a thought exercise. Let's say you and I are the board of directors of the CEOs of the telcos. Okay, we're running the biggest telco on the planet. What do we do? Let's just, let's riff this. Okay, so what's our biggest problem is we're spending hundreds of billions of dollars on CapEx. Yep, we have massive end-user penetration. But the problem is the cost per bit keeps coming down. The data's growing like crazy and we're spending tons of money and we're not monetizing enough because those over-the-top vendors are making all the dough. So, what would we do? Well, what do we have? What's our asset? We got the plumbing, the network. We got the users. We got a ton of data. We got controlling the device. We got a ton of data. We got device access. We got capital. Right, we got a lot of intellectual property. What I would say is we've got to appeal to developers. Developers are going to win the next wave. I would say, I would say, I would say not the top priority. We got to control the developer community by saying, we have the end-users. Give them new experiences that make them go wow. But that's how you get to the developers. Make them go wow. Yeah, exactly. They're very much related. It's Yin and Yang of innovation. We got the users. And the data. And the data. And we're going to now make that available to you so you can develop apps. We're going to partner with you. We're going to build an ecosystem. I would do the following. If I was in charge, I'd take over the telco immediately and saying, get the data, all the data. Operational data and user data. Get that foundational model built with AI immediately and offer that up to developers. And saying, here's an API. I'll charge by the drink on transactions. Because remember, when the cellular model went from voice to data, they made so much bank. Remember those days? Data's the money machine. So I would say that new business model's data. But the new data is operational efficiency and then end user experiences by using the data that they got. Roaming data, location data, every interaction data. And then saying, okay, developers, use my API to give end user experiences and I'll give you distribution on the application side and I'll get paid on every single transaction. Like Stripe does, like everyone else does. Okay, so the way I would do that is I would have an all hands meeting, okay? And I would hire Steve Ballmer. Come in and I'd see Steve. And dance around. I'd say, get all sweaty, okay? Hey, get the armpits going. And then come in and start ranting about developers, developers, developers. Because we got to solve this problem. That's how we're not going to get bogarted again by the over the top vendors, the way we did 10, 15 years ago. And then, so that's one. I would definitely get that Amazon vibe going if I was a telco. And then I would do is, I would nail the enterprise. I would say, look it, we got to get the business to business market down. I would get like highway nine networks which is Alwin's new startup, people like that and saying, look it, here's SLA's that we'll do for the enterprise. And just knock that down ASAP. More professional services as a service, make that work. So you got to nail the enterprise. And the users, by the way, work at the enterprise. So the users are the top priority. The data for the users, the operational data is the top priority. And then go to the enterprise and say, hey, you know, people have mobile phones on the network that you're on and run that way. So that. All right, you know what else I'd do? So somebody said this, yes, I forget which guest it was, is the telcos really, they don't really sell to the enterprise, the large enterprises, that the cloud guys own that. Telcos, they'll sell to the small business. So what I would do is, okay, so how do you create a cloud like asset that you can utilize to sell to the enterprise customers? So I would spend $50 billion. I mean, they're spending CapEx like crazy anyway. I would spend $50 billion and buy Cloudflare. I would pick up Cloudflare. Why would I do that? Because Cloudflare has all the edge capabilities. They have security, they have a cloud-like model where they're selling compute, storage, right? They got serverless. They got, and people look at Cloudflare, they go, well, that's not really fully functional. It's like lightweight Amazon. But I would acquire them, invest like crazy, open up my developer network and take out Cloudflare. 50 billion would get it done. Cloudflare is a great company, definitely good call. Well, Dave, this is, we're at MWC, we're in Barcelona. Obviously, Telcos, the world here. Great, a lot of news going on around the internet. Obviously, AI is the hot thing. Apple's getting out of the car business. Micro-Street. They didn't know that. Frank Salubin stepping down from Snowflake. That's big news. That's huge news. And Stock dropped 20%, by the way. 20%? Oh, that's a buy. Sridhar Ramaswamy is going to be the new CEO. The key to me is if, is Mike Scarpelli retiring too? If Scarpelli's not retiring, I would. 20% premium. Not that we give advice, don't take our advice on buying Stocks. Buy the hell out of it. Mike Scarpelli runs the operations of that company. Slutman's always had the board. And he had the vision. And he had the ethos and the go direct. I mean, you know. He won the race. Amped it up. Yeah, yeah. He's a sailor. He's reached his continent. But Scarpelli runs the day to day. That's why that company is such a strong. This same thing happened when Slutman left ServiceNow. Jim Cramer was actually, I thought he was going to cry. Oh no, Frank, don't go. This is terrible. Oh, ServiceNow is screwed. Stock dropped. They bring in John Donahoe for a two year stint. He was always a short timer. He wanted to be CEO of Nike. But then they bring McDermidin. Look at ServiceNow today. I mean, first of all, in my experience, Slutman doesn't leave companies in bad shape. Okay, because he's always been in it. I mean, he came back. He was in it for the sport. The last thing he wants to do is leave a legacy of, oh, I flipped the keys to a company that wasn't in a good position. Look at Data Domain. He with Frank Petrone, they created an auction between NetApp and EMC, Joe 2G1, because he had more cash. Slutman stayed on and helped clean up EMC's kind of dysfunctional data protection business. Remember Avamar versus Data Domain? Slutman stayed on. He didn't have to stay on. He could have said, hey, I'm out of here. He stayed on. He did the right thing. He did real work to make sure that before he left EMC, their data protection business was in good shape. It kicked ass after he left. Same thing with ServiceNow. He left that company in great shape. He made sure that they were expanding their TAM. They were reaching new markets. They were launching an app development platform. He did all of that work, clean that company up. And then look at Snowflake. When he came on Snowflake, he even told me this to one time. He said, these guys had product market fit years before they built their go-to market. He came in and he goes, you guys have such great product market fit. You just don't have go-to market fit. We're going to solve that problem. And of course, the other thing that Slutman did, he continued to invest in engineering talent. So that company is in very good shape. They're facing some challenges. There's no question. People don't like their pricing. They think it's too expensive, but that's a good problem to have if you can sell the value. So that's interesting. I didn't realize the stock dropped so much. I guess not pay attention, but I would say that's a buying opportunity. Let me ask you a question here at MWC, which is basically an enterprise show at this point. You got a lot of vendors here. Classic, enterprise, cloud, data. Snowflake's mentioned here, one of the sponsors. What's your angle on that? Because I think to me, my takeaway is, is that you always had American companies here, but more enterprise than ever before. I think that this show is converting quickly into being a combination of re-invent and CES, right? So you got those two worlds coming together. And the opportunity that the GSMA has here is to keep this show to be global, telecom based, cloud and developer and make the US show MWC about the enterprise. So to me, why would I want to go to the Vegas show if I'm going to come to here? And I think their opportunity is to kind of create a line extension for the enterprise. There's enough demand, Dave, for this show to be in Vegas as enterprise. Meaning IBM, HPE, VMware, Broadcom, Dell. Even the telcos want the enterprise business. 5G is enterprise. Look at 5G has not yet delivered. 5G outside of blanket coverage of great fixed wireless, outside of fixed wireless blanket coverage and connectivity. 5G has not found the killer app, okay? Consumers won't pay for it because they expect it now, like plumbing. And the only way to make money with 5G is the enterprise, business. And I think they got to make this an enterprise show at some level because it's just too lucrative. 5G will happen in the enterprise. It's just not happening now because of slicing is not ready for prime time and IT integration with telecom is terrible. So what happens when there's demand and inadequate capabilities? Entrepreneurs step up like Alwin and Highway 9. So entrepreneurs are going to step up. The big vendors are going to step up. If I'm Dell or HPE, I'm all over clustered systems. I'm all over this next generation data center. So John, I'm going to go back to a previous topic. Snowflake, I'm just looking at it. Snowflake's going to open up. They're by $75 billion market cap right now. They're going to lose about $18 billion of market cap at the open. 18 billion evaluations is going to. You think it's going to drop on an enterprise? Yo, it's clear. It's the pre-market. It's down. The stock closed yesterday at $2.30. Sloopman evidently announced after the close. It's going to open at $175. Maybe $180. I mean, we're talking about 15 to 20 billion of market cap vaporizing. I do think that's a buying opportunity for people because Shridhar Ramaswamy, the company they bought to do AI. They were a consumer AI company that Shridhar ran. Why can't I think of the name? It's probably because I'm hungover. But... Yeah, I mean, look at it. The stock's not dropping. I mean, I don't know what you're talking about. What are you talking about? It's going to open down huge. It's up 26 points this month. Yeah, but it's going to open up 54 points down at 9.30 East Coast time. Over the past year, it went from $189 to $2.30. Yeah, so maybe it's letting a little calling of the herd there, you know? It's going to get crushed when the bell rings this morning on the NYSE. They're NYSE. So, yeah, it's going to open up 25 points down this morning. That's 15 billion. That's a prediction. I'm looking at the data. It's going to be 15 to 20 billion dollar market cap hit. All right, and so they're going to have to do, you know, damage control. I mean, all-time high is $353, right? Look at the pre-market on Snowflake. It's going to open at 180, maybe, maybe lower. Yes! Yeah, it's at 230 right now. It's going to open at 179. Yeah, there you go. It's a stock closed at 230 yesterday. Okay, the lowest stock price that Snowflake had was 127. Yeah, it opened at 120. It went public at 120, and nobody got it at 120 unless you had an insider, you know? It dropped in May 2022. May 1st, 2022 was the low watermark after the IPO. Snowflake's been a very volatile stock. And so it didn't, but the IPO price was 120. Nobody got it at 120. If you bought it on IPO day, you probably paid 280. Let me ask you a question. Wait, wait, you probably paid 280, right? And then it's been down to as low as, what'd you say, 123? It's been up and down, up and down. But Shreeter Ramaswamy ran Nuvine. That was the spin-out out of Google. He jumped, popped out of Google, tried to do basically an AI search that this is not going to, we're never going to beat Google, turned it into enterprise. Snowflake buys them, he now becomes the CEO. So you got an AI CEO. That's a smart move. Yeah, I know. And by the way, he came from an acquisition. Yeah, Nuvine. He's not a Snow, Nuvine. No, Nuvine. Yeah, exactly. So let me ask you questions, because you're from Snowflake. Just to compare public-private valuations, Stripe just sold about a billion dollars worth of internal shareholders, employees, what not, for some liquidity, because they're not yet public. This happens, it's basically some liquidity event. Not public companies, so we don't know what the trades are, but the valuation's 65 billion for Stripe. These guys are valued at 75 billion at the close. What's Stripe's revenue? Don't know, actually don't know that number. Off top of my head. So public market values still- 14 billion. 14 billion? They're 14 billion. In revenue? They're revenue in 2022. Snowflake's got a couple billion in revenue. Two to three, let's call it three billion. And they're more valuable than Stripe, so. Well. Obviously the street thinks they've got a better business model. Let's get into the little, that's my analysis here. So 75 billion a little bit over, compared to Stripe's revenue, okay, high from a multiple standpoint. So is this a normalization, and then what does that mean for Databricks? What's Databricks' last valuation on the private side? I think Databricks is worth at least 40 to 50 billion, maybe even more now with the AI hype. I think Databricks, honestly, I think Databricks and Snowflake should be comparably valued. And the reason I say that is because a lot of Snowflake's revenue goes to AWS. Databricks doesn't resell AWS. They don't package AWS into their pricing. You got to go out and pay AWS directly. Snowflake, a portion of Snowflake's margin is AWS revenue, because they bundle it in, okay? Now, so the Snowflake gets its margin by negotiating great deals with AWS. So in that sense, it's maybe got a better business model. At the same time, I think customers feel like Databricks is less expensive. And I think part of that is because they're buying AWS services, compute and storage, directly from AWS, not bundled into the Databricks pricing so they feel like it's a lower cost. And whether it is or not, I haven't done the analysis on, but I would say that those two companies would be comparably valued from a valuation standpoint. Let's simplify it, revenue multiple. I think Snowflake and Databricks should have a comparable revenue multiple, but I haven't seen Databricks' income statement. So I'll reserve judgment till I see it. But based on what I know, if they can get an income statement that is reasonably profitable and it can show SaaS-like gross margins of high 70s, maybe even low 80s, that they will be valued comparable to Snowflake. And actually maybe even higher because they have a higher growth rate. So maybe John, their overall valuation will be lower because Snowflake has more revenue, but I could see Databricks revenue multiple being higher because they're a higher growth company. Hopefully that made sense. I'm a little incoherent this morning. Well hey, Ford Asians and Mobile World Congress were here live in Barcelona, love the city. I put on five pounds while we're here, it's so good food. You know, obviously theCUBE here has been a smashing success. We had the CEO of HPE, Antoni Unirion, Juniper Network CEO, Rami Rahimon. We had two technologists of AWS on. I saw Chris Niederman earlier. He runs all the industries now. So Amazon's web service is here in force. So Cloud, big presence. You got Clint Crozier's here. I didn't see him, but I know he's here. He's got the space and the satellite stuff going on. You had president of Broadcom, Charlie Kowals came on and you had some startups here. So we had a broad swath of amazing guests come on, real chronicle analysts. I mean, look, I feel like I learned so much this week from our interviews and it confirms to me, again, back to my view of this systems revolution and what I mean is we are witnessing a historic moment in the computer industry where AI is driving massive disruptive enablement around how data centers are going to be built in the future, cloud operations is going to influence user experiences we've never seen before. I was talking to Samsung and some R&D folks in the hallways and some of the parties. What's coming on in the user experience side is all kinds of new interfaces to users, not just on the devices. Devices will be walls and screens, see-through screens, you got vision pro-like experiences, and again, vision pro from Apple. I mean, I personally don't like that product. I want to never wear the goggles, like the ski goggles until I can feel comfortable doing that, but it'll get smaller. It'll be Ray-Bans like what Facebook's doing. Let's end up having augmented realities coming big time. So device connectivity is absolutely going to be the thing and the back office that powers that has to be re-architected and it's not a complicated architecture, it's distributed computing. So mainframe theory, mini computer theory, network theory, computer science principles we're all going to come to bear around how do I design the system and what's the system software look like and what's the end user software look like? So it's going to be a renaissance of systems, it's going to be a revolution in architecture and the game is still the same. You can have servers connected to other servers, thousands of servers connected. The network in the cluster is now the new data center. It's a massively distributed system. It's a massive computer system that's going to go down. NVIDIA is just the tip of the iceberg. If you look at what they're doing with AWS, unbelievable architecture, they're using Infiniband which we know from our covering days of Mellanox. Infiniband is critical for those super monolithic machines but Ethernet will win the day everywhere else. So these guys are telling us we've got to wrap. I just want to say one thing, John, I learned a lot this week too and I'm so proud to be working with you and the rest of the team to serve our audience. You know, a lot of people, I would just, there's a lot of fun out there, okay? We have a real audience and the reason we do is because we do great editorial. We invest a lot in that editorial and I'm really proud that I can wake up every day and ask myself, what am I going to do today to serve my audience? And I tell you, we have great clients who support us. They want to come on theCUBE. They want theCUBE research and our digital capabilities and they enable us. Their funding enables us to serve our audience. That's why you and I formed this company, you know? Because we want to have fun too and create jobs and we've done that and the rest has yet to come. I got to say, when we combine research and real-time reporting with data and theCUBE and video is amazing. Now that we have 13, 14th year with platform as well and AI coming in, it proves that our bet was to reinvest and someone asked me the night to think, so you guys must be really making a lot of money. Like, no, we actually reinvest it and unlike others who have their hand out, we're reinvesting, okay? And that's the key and we love what we do. Love the business we're in and it's fun. We're self-funded. It's fun and we're self-funded, no outside capital, no VCs, no board pressure to get an exit. We reinvested the good content, great community. It's awesome, 18,000 CUBE alumni and growing. I mean, this is obviously an investment being here at MWC, right? It's not cheap to build something like this. But it's worth it because again, we can serve our audience, we can serve our community, our analyst community, our CUBE collective and it's fun. Well, episode 49 here in Barcelona. I'm John Furrier, Dave Vellante. Thanks for watching. A little bit shorter edition than our normal hour podcast but we're going to be on the road traveling back to the US. Dave, thanks so much and a great, great event here. You're watching the CUBE pod. Thanks for listening. We're at silkenangle.com, that's where all the action is. The cube.net, you'll find out the videos. You'll see where we're going to be next. And of course, check out thecuberesearch.com. That's the rebrand and Wikibon, our growing CUBE research team. Amazing team developing, there's more coming. I can't wait. We've got the CUBE AI division, we've got five divisions. We might even have six soon. Six. Silken angle, the CUBE, the CUBE cloud, CUBE 365 research and the CUBE AI. So we've got three groups growing, just awesome. Thanks for watching. See you next time.