 Good morning everyone, welcome back to theCUBE, the leader in live tech coverage, live day three of four days of CUBE coverage of SD23 from the mile high city, Denver, Colorado. Lisa Martin here with John Furrier. We love talking with Stealth Mode startups, we're going to be doing that next with a good friend of theCUBE, Renu Raman is here, the founder of Teresa. Welcome to theCUBE, thanks for joining us. Thanks for inviting me, John and Lisa. So just exiting Stealth Mode? Yes, still in Stealth Mode. Probably in Q1, we'll come out more. Maybe this, I will see. Can you give us a high level overview of Teresa? What problems are you solving without revealing obviously anything proprietary? Yeah, life. Teresa is kind of, as John was saying earlier, my life experience put together in a bottle. Happened to be at, just to go a little bit back to give the idea of what that is, I was at Sun and we changed the world a couple of times while I was at Sun Microsystems, I was heading all the microprocessors, chip programs. But the three things I learned from Sun was we went from risk to multi-core and threading, I led that effort. That was an architectural shift. So we are in that, number one. Number two, we, like people say, we changed the ED industry, and ED industry was changed because we needed the ED industry to do aggressively against our competitors at the time, Intel to do chip design. So we are on the same boat now today as we build systems for AI, but use AI for systems too. So that's the second thing. The third thing we also did, at that time, I was lucky to be part of that company, was change from building proprietary manufactured systems in Newark at Sun, to I was the first one to go away from Sun's own manufacturing to ODMs and MiTAC and others. That was changing the supply chain, call about white box and others, from OEM systems to white box, et cetera. So that team plays into what we'll do in terms of how do we build new systems and deliver it as a service for, and the other transition that happened was, since Sun was, Solaris was a Linux, that was closed, open. And Sun was the first open systems company, but it lost its way eventually and became closed, but open systems came and take over and caused the next architectural shift. So the same phenomena is going to occur as Kevin was saying earlier too in the prior call. So Renu, I wanted you to come on theCUBE here on this show specifically because, again, as you laid out in your career trajectory in quick summary, you've been involved in major inflection points, okay? Besides all the great work, I don't want to oversimplify the work you've done. I'm having lived in that era and know what you've done. Really a big set of accomplishments at a technical level, but the timing of what you've been working on is pretty incredible. So, and we're in the same moment now. We've seen this movie before. When you have these massive inflection point shifts, you said close, hilarious to open, ODM to kind of change that model on the white boxes. There's consequences downstream. They're all embryonic shifts. We're seeing the same one now. We saw the web on mobile, everyone says, oh, so PC this, but you're actually in the tech side. So we're in the same perfect storm of open, source booming, the cloud. Peak guys are peaking. They're maxed out. They're still doing more, but they're looking more like snow stove pipes. Almost like proprietary noses. I've been saying on theCUBE, young people probably don't know what that is, but that's proprietary network operating systems. So they're in a dilemma. Then you have the emergence of the AI wave, which you've been writing a lot about on your substack around this new formation of a market that's developing out of the inflection point. So with that as a setup, well, first of all, did I get it right? And explain what the shift is that you see and what's the market formation happening in that AI enabled early stage embryonic, small little steps that will explode into massive change? I think everybody understands that AI is a massive shift and AI lifts all boards. I'm going all the way back to you and some people are working on Z-Series mainframes and changing COBOL and providing insights via AI with the old COBOL mainframe core. So that's already said. But if you've seen the movie, everywhere everything happening all at once. So this is like we are in 2023, like everything is happening all at once and everybody's confused. At the same time, everybody sees the opportunity to do interesting things. So breaking it down, I think there's a systems open, close models, but there's also think of on the tech stack, we used to be bare metal OS and I was part of VMware as I said, and VMware changed from there to the VM as a unit of compute. But now we've got the model as a unit of compute. Model becomes the abstraction layer between what is happening underneath and what is the end applications are going to be developed. So it's going to change as Dave Patterson told me back in 2017 where Unix was attached to, at that time, risk. He said PyTorch is to the new underlying machine. Actually it's more the models is going to form the new form of compute layer. And everything below is going to change and everything above is going to be what the API or the programmer is going to be. So how do you try the programming model that you've got to simplify to the new model based world? To the underlying performance model, what is the new system? So there are a whole set of innovations going to happen. We can talk at length on that one. So semiconductors and cloud kind of come together which can stay in the cube but what you're saying if I get this right again trying to piece it together is what we're seeing with chat GPT is the consumerization of an interface. Okay, that simplifies something's going on. Yeah, it's streaming sentences, doing stuff with images. We all go wow, that's magic. Educates everybody. An interface. But behind it something's happening. You're going one step first thing. The language models themselves, the foundational models are now an abstraction layer like a middleware was. That's now going to, on top of the infrastructure which has to be refactored and invisible. Invisible and also you're not going to be writing, I mean what Shopify has now got a million lines of code generated by co-pilot. You're not writing code pedantically or procedurally. It's all declarative. As we go on, a friend of mine, Eric Meyer at Facebook says, we don't even need to teach data structures to students anymore. Everything is a model and a collection of models. So everything is going to be declarative. You and I, everybody as Jensen also says, the new paradigm is the English language is a programming language. So that is going to happen over the next 10 years. Okay, so a news tax here, what's going to be the change points? What do you see as the battlegrounds or the opportunities? I mean, I've heard people say, we got to control the data control plane. I want to have a semantic layer. Certainly data management looks like it's going to go upside down in terms of like how that script will flip. What are the areas that are going to change radically or areas that people will fight for or develop to? What are the hotspots? What's the areas? So I would like to look at the world as, it's not a zero sum game. The pie is going to grow bigger. And as the pie grows bigger, the new players emerge or new categories emerge and new things emerge. We're going from $200 or $250 billion of infrastructure spend to $600, $700 billion of infrastructure spend. So what are the new categories going to emerge? Clearly there is GPU and Jensen, crew on the bottom of the stack. Clearly there's chat GPT, which is replacing or adding to the web browser to the window system that we used to have, the UI part. The other thing that's going to happen is, what are the consumption surfaces to your point of data? You have, lots of people could not go to the public cloud because of regulated reasons or the amount of data that exists or sovereignty that exists. And do I go and build my own data center? No, I've already drank the Kool-Aid of a cloud consumption model. If I drank the Kool-Aid, what is the best way? So I don't like the word tier two cloud. I think Kevin was there. That's my word, I use that word. It sounds subordinate, I get that. Yes, the Cruzo cloud and Lambda labs, all of these. What are the emergent thesis? If you go back to the 90s, the same thing was, the white box emerged, a lot of people emerged, but they really are trying to drive the new categories. So there is going to be the private cloud. So are those categories specialty clouds or? I think they start out as specialty clouds, but they're going to create the new category. So the way I think of infrastructure is there's technology stack that is front and center, but my lesson learned in the last 20 years, if you look at the success of AWS and Google, it was operations first. So if you look at Outpost, it has been successful, not as successful going into outside of the public cloud. If you look at Anthos, not as successful, right? So what that lesson learned is the operating model is different when you go from a public cloud, elastic compute with million servers in one region to I have a collection of servers in 40 places and it's a different control plane, a different operating model. So not only the technology stack, the operating model and how do you serve all three change. So there are going to be new categories of players, whether it is Teresa or others, we will see. Talk a little bit about how you see the market evolving over the next few years and how do you envision Teresa as part of that evolution, maybe as a catalyst for it? So I think the biggest market, the consumer market is already in flight with barred open AI and a whole bunch of applications are going to emerge with that. The big transformation is going to be coming from the VMware, part of the VMware and I also did SAP is the enterprise market is going to shift. There's a little bit of, call it super cloud tier two or private cloud consumption surface, but what does the stack look like? And there's both a legacy component that's not going to change, plus the new compute model. So how do you blend the two? It's not just new workload, it's going to be a blend of the two on these OPEX model, which is closer to where the data is and deliver that with new, perhaps new business models considering the model as a service as well. So you're saying, I guess what you're saying is the market is the least question is unknown relative to what you can define and it'll be new so it'll probably be things people don't see. That's what happens in these deflection points. It's like some nascent player emerges like Google was considered a bad search engine at that time when they rose to the number one because algorithmic search was actually the best but then people didn't understand it. We had all the ideas at Sun doing clustering and the service, N1 was the one that Yusuf Khalidi did, he's at Azure right now. We did the whole public cloud thesis but we did not realize that the shopping store became the cloud. Do you see this as a black swan event on the horizon? It's more than a black, yes, it is a black swan event and there are some tea leaves are shown who's going to emerge. There's still more to be discovered. There's a lot more to be discovered. Yeah, what's interesting about what you're saying is that if the new stuff's going to come out, who's going to capture that opportunity? And what I want to ask you, because you brought up the enterprise and the old slower moving inflection points because we're in a high velocity. Again, the pace of play is higher now than it was in other runs because of what we're living in. Usually the consumer sets the standard first and the enterprise lags behind. That was always the old playbook. So now it's not that way. It's not the case. The enterprise actually has use case AI examples. Consumer check, what does the enterprise look like to you right now? Where are they in the IQ level of understanding where they are? Is it like in boiling water? Do they understand their predicament? Do they understand the paths? So I'll give you two. I think I've blocked about this. That is SAP and Oracle. I have a relationship with them from the last 20, 30 years. But they're also data breaks and snowflake. So juxtapose is to an imagine where they go. These two incumbents are in the enterprise. They're bringing the new tools, new capabilities, and they're going to be competing with those two. Who's going to win among the four? We will see over time. Parquet is going to change the game on the data side. There's another dimension also is geopolitics and geo. The last build out entirely happened in the US as a dominant and then in Europe. I come from India and I've just visited India too. I'm surprised like the wireless infrastructure leapfrog. All the new build out is going to be green field data centers. When you have new capital and new stack, don't assume, let's assume the new geo is also going to come up because of geopolitics to go build out. So that's another dimension that's going to play out. And the new stack is going to look like what to you. If you look at the future, obviously got abstractions, coding like co-pilots of the world, we'll maybe create glue layers on the fly, manage data pipelines, Kubernetes services. What's the areas that needs to get stabilized in terms of core tiered pillars in the stack? Yeah, this will go back to a lot of Jerry Chen's thesis of systems of record, systems of intelligence, and systems of engagement. I think that framework really applies invariant with time. The details matter, the details change, we can go out at length. So the systems of record, which used to be databases, now models could be systems of record. I think that's where it's going to be, and where it's going to shape up for the open AI world and then the enterprise world, that's going to shape up. Systems of intelligence is where the new innovations and actions happen all the time, which is the app engine. What is the co-pilots agents and aggregation and make that all work together? And how do you make the two layers work very well from a performance model? The third thing is that the systems of engagement with this is a chat, but the browser doesn't go away, other things don't go away, we add on to it. So these three things, how do you re-architect and put the systems together? Can you share a little bit about how you plan to differentiate what the services that Teresa is going to deliver in a no doubt competitive market? And I got to ask you, what's in the name? Is this related to, it turns a wave? The name is actually first, when I was trying to figure out the names, I used to have a thing called Terza Italian in Italian three, but the lawyer said, pick a name that is nobody understands and abstract nothing because your company, you start today, this is not the company, you will be tomorrow. So I was searching for some tiles for some work in my kitchen for my wife and then found Mexican tiles named Teresa. So that's how it was Terza, Teresa. So that's how the name evolved too. We're a little too early to say where we're going to differentiate and service. We are still in stealth mode, we are figuring out all the pieces. But I think the general thesis I would say is the emergence of the new categories of consumption surface, the geopolitics, and the new stacks along where the systems of intelligence we have to drive. What does the new App Engine look like? That is tight to make the performance model and the programming model tight. Full system thinking, it's really important to come full systems thinking in this case. No, that's why I like the systems thinking because there's consequences when you change things. But are we just living in the same old formula hardware, middleware, and app? I mean, isn't that the kind of construct that you got the cloud? You know, I mean. Very, very good point. I think the question now is, I'm not smart enough. I probably have to bring Eric Meyer and crew. What does the model world look like? So Andre Carpathi has already put out, if you've listened to his, see how this tweets. So I think we are in a brave new world as some of the people discover is what does the model world look like and how does it change that three layer stack? Well, we're excited about the power law that we put out. Also, we're excited about seeing how the super cloud narrative evolves. That's now two years old, Dave, and I put that out, you've been following that. What do you see this super cloud emerging? If you can call it super cloud as this new thing, just as a word, what does the super cloud vision look like to you? So I was in the multi-cloud world in the VMware site and then you came with super cloud and I was a little bit confused. But now I think I realize where your vision has been going. I think NVIDIA is already going in the direction creating, if you look at DJX cloud, it's probably the first real instance of having their components running on the public cloud. They have a DJX cloud and a partner. So you can now take the thesis forward. There's going to be an open, closed version, new systems in the king, and that's going to be the evolution and that may tie into your super cloud thesis. Fun time, Lisa. Fun time, fun time. And then by the way, I also have to say, you said 25. In the year 25 is when you'll be 25. Yes, we all have to be 25. AI is the fountain of AI. I like that. You're still 23. All right, yes, yes. All right, we're still in stealth mode. Still in stealth mode. What can you share with us in terms of as we're watching this space? So, so... I'm breaking on the cube. Give us a little bit of taste. What can you share? What can we expect in the near term? I think there is new systems to be done. What is happening is we have got one type of CPU, or one type of computer, one type of memory. We've gone to two types of computer, two types of memory, at least today. But really the components of a system is changing from just silicon, and nothing else. If you go look at a rack today, you have water and cooling, which was not there. You have cables going from copper to optics, which was, so you have, as my friend Shaheen would say, it is light, sand, and you have, obviously you have silicon in it, and then software, right? So, the whole system is going to change to new material set, new components set. So this is a systems part, and how do we deliver in the super cloud, better name for tier two? Yeah, tier two is bad, super cause better name, and I think we've seen these trends, agile startup, design thinking, I've been saying on the cube, you know this, we have a systems thinking paradigm coming where, whether you're coming out of college, whether you're going to learn data structure, have it done by AI, you got to understand that a system has actions and reactions, and this is a systems construct, and systems thinking is the new trend, it's a thing, and it has to be because everything will be connected, because the personalization on the consumer side has to be broad, and then precision, and real time. In order to do that, you need an operating system. And you need an operating system, absolutely. That would be the layer to focus on. We, as I said, we have Unix, we have VM, what is a new operating system that is memory-centric? That's probably the last line. All right, there's a tease right there. I have a sense that's going to be part of the startup. That is a tease, buddy. Thank you so much for joining John and me on theCUBE today, sharing your background, your history, the catalyst to launch, Teresa, what you're working on in a non-preparatory way. We are going to be definitely keeping our eyes on this space, and congratulations on what you've accomplished so far. Thank you, thanks for the time, and good to see you. Likewise, thank you. For our guests and for John Furrier, I'm Lisa Martin, and you're watching theCUBE live from SC23. We'll be back after a short break.