 Hello, welcome to this CUBE conversation. I'm John Furrier, host of theCUBE here in Palo Alto, California. We're bringing you a special conversation with a venture capitalist, good friend of theCUBE. We're trying to figure out which year he first was on theCUBE, but he's an OG, original gangster of theCUBE community, Alan Cohen, general partner in DC, VC, many careers in running early stage companies, wireless to networks, Nacira, Airspace. Really here to unpack the future of how AI will be impacting the technology industry mainly around kind of the conversation that grew out of the Juniper HPE acquisition that HPE bought Juniper for $42 billion. Really around the crown jewel of missed AI, which Alan has some DNA. He knows some of the folks, there was an advisor. Alan, great to see you. Thanks for coming on theCUBE and chatting with me. It was great to be here, John. I feel like a hobbit at, you know, I'm revisiting Brie after the, or after the World of Wars. Yeah. Well, it's been great. You and I have been friends for over a decade now. theCUBE's on its 13 year or we're going through another inflection point. Again, what's super exciting is again, I always say in theCUBE, I wish I was 25 again, because this next wave inflection point is real. The AI hype is matching what is coming in terms of substance. So it's pretty close to hype cycle meets, you know, meet on the bone. It's going to look like the web, I think, but we can debate that. But this next wave is coming fast and it's AI is the buzzword, but it's really a lot of things going on with AI. And specifically, everyone talks about consumer, chat, GPT, whatnot, but it's a lot of other areas are going to be disrupted. And they're new things. So some of them are, you can pattern match a little bit. Oh, it's like the web, it's embryonic, it's going to grow and get faster. But the bet's going to be on how the infrastructure and apps kind of start rolling together. Who does what? What is going to be automated away? How human can be part of that loop? I mean, there's a whole nother level of thinking and you're a thinker. So what's your take on the AI wave right now? And just generally zooming out before we get into the MIST and or AI ops or impact to technology? Well, by the way, John, I saw your show last week with Steve and Zia and Jake and you guys unpacking. MIST, I was actually, as you mentioned, I was a board advisor at MIST five years ago when they got started. And that was kind of the first glimpse of using machine learning algorithms to actually replace human beings for the management of IT networks. But it's funny to see that the technology industry is actually catching up to the real world. So for the last five years, I've been investing in companies that use AI, computer vision, machine learning to run businesses we don't think about in the tech industry, rockets, robots, microbes, physical security. And they have been incorporating the ability to consume massive data sets for decision making. And anywhere we have, as my partner Matt Ock always likes to say, we do security as a service as. We have a company for them that is a drone detection company evolved where I was the chairman and we took them public. Those are companies that use computer vision and machine learning to actually protect people. And they've been doing that for seven or eight years. So to me, what's happening now in the tech industry is an industry that had been built clearly on writing software but not necessarily using generative AI, massive data sets to actually instruct their actions. They were really configured and managed by people. So that to me is like the wave that we're starting to see. And your point is that the app is at the infrastructure. Well, the generative AI is going to be the decision maker and basically negotiate between the two of them for how things work and how they perform. So pretty exciting time. And in an industry that employs millions of people, this changed very little over the last couple of decades. Architectures have changed, chipsets have changed, software has changed, but the human beings interaction with those systems has not changed a lot. Right, I mean, in the networking industries, people still brag about CLI. It's pretty basic stuff. But, and they have to because of the heterogeneity, the distributed and dynamic nature of computing. And then obviously the heterogeneity, you're the, you know, the godfather of super cloud and many cloud and hybrid crowd and multi cloud. Well, those are all about all these different environments that are now kind of banging into each other and they can't just have effectively human middleware manage them by themselves. So pretty exciting. And then obviously compounded about the enormous amount of data that is making its way into organizations through AI. That's a great, great summary. I think let's unpack some of that out of the box there because you got, you mentioned humans in that AI middleware. I mean, middleware is a concept that goes back to the old early days of computer industry, you know, hardware, applications and then middleware to kind of run things. That's just started to have distributed computing, mini computers, lands and whatnot and, you know, networking, internet working and TCPIP, create all that standards. Now today, the idea of architecture shifting is interesting and it's interesting that IT hasn't changed. I mean, we're talking about platform engineering, some of the stuff from theCUBE. It's like, wow, you got to be really engineering smart to do that. Not an IT person potentially is more operating. You got an operator running some vSphere, some VMware and you got alpha engineers working on hard core, re-architecting like large scale infrastructure. So clouds, for instance, like Amazon Web Services and Azure. And then you go, okay, well, on-premise has great for machine learning and AI and you got Edge. I mean, that's a data problem that's kind of mind boggling. If you're using an old data warehouse, who runs those? And then it's like, okay, then, okay, how do I mesh? One, how do I figure out what I want to do? So I think people are crashing and burning right now saying, you know, kind of banging their head going, what do I do? Well, everything gets compounded. And I'll give you a great example. The size of the datasets that you see in the kind of the large language model, the chat GPT architecture. So you'd say, hey, this is really good. It can write a marketing paragraph or sales paragraph or three pages of documentation for me. Of course, I have to burn down a small forest in the Amazon to run the computing capability to do that. So we're still in this kind of building phase of architectures, which large language models will run in on the monolithic platforms like chat or Google or whatever Amazon and Microsoft are building, which will be open source. Who's going to be the red hats of AI because one people don't want to pay the tax, not everybody can, and they are very sophisticated and a lot of the open source models are really, really, really good. And fast too. And fast and don't necessarily have to consume running the whole model just to run a query. And your point about burning down the rain forest, whatever you said, that's implying obviously the resource required to run a basic query on chat GPT. Or drain Lake Tahoe to make the transition. And the point there for the folks watching and listening is there's been documented that a to run a chat query on chat GPT just requires massive amounts of power and cooling. And yeah, enormous amount of water. So I think you'll start to see kind of distribution architecturally because there's not one size fits all. Some things can be done on an open source model that's fairly efficient to run. And I'll run that in my organization actually will be running on a, but I'm not missing it. The guys in Cupertino tell you it's going to run on your Mac or your iPhone. So that's going to be a lot smaller model and that's going to have to plug in all the time. Now that battery is not going to be around even though the power on the chips are pretty good. So, you know, there's, so kind of everything is up in the air, right? We have the underlying from the physical chip levels, right? It's been a really great run for Nvidia, but obviously there's going to be competition. The centralized versus distributed edge, you know, hybrid, right? That's all up in the air. What the apps can do. But the most exciting part, I think of all of this to your earlier question is I don't actually have to be that same kind of expert. While people talk about a lot of things you can do with, you know, I can create funny videos or I can have a picture of Gandalf playing at a Bruce Springsteen concert and on mid-journey, I can do all those fun things. But what I can also do is I can actually transition code. I can say, here's a code base. Rewrite it in Python or pick a more, you know, a new language, I can translate. It solves the cobalt problem. We would have didn't have it for the year 2000. There was a lot of good jobs, right? And a lot of us, I mean, I worked at Cisco. We were up till two in the morning making sure all the networks didn't go down and then we started drinking, right? But yeah, but we sold an enormous amount of equipment to even just deal with that. So the organization's consumption of AI because you can do things in natural language is going to accelerate. So there's gonna be a lot more people using it than the people who provide the underlying plumbing and capabilities underneath it are going to have to shift to do that. And so you see, and we were talking before we came on camera to get this out there now, this is a good point to talk about it. That AI will replace those jobs or elevate those jobs in IT that are mundane and are quite frankly grunt tasks or rock fetches as Dave Vellante calls them to make the machines run that, to shift the humans either into a new role or just automate away completely their role in running middleware and configuring and running networks. I don't know if it's a net or gross job killer but there are hundreds of thousands of open jobs in IT. You cannot run a business in any industry anymore unless your computing infrastructure is supporting it. You can't introduce a new product, you can't take an order, you can't support somebody, you can't reconfigure a manufacturing line and if your IT is slower, you don't have the people to fill those jobs, those are all delays which create opportunities for your competitors or lost revenue for you as a business. So the ability to operate faster could be one of the, and more efficiently and honestly, with fewer errors could be one of the great benefits. So for the, when you- An incentive for executives to fund and or figure out quickly or give the green light to engineers or people to figure out how to deploy AI to move on that opportunity. Right. So you know that by background, you may remember or you've probably purposely forget for good reason your own sanity by training I'm an economist and I always look at- I didn't forget I was gonna bring that up. I always look at how people spend. So if you look at IT spending, for the most part for the last decade, I mean there's been some spikes on and off, it's been in the low single digits. And so if you think about it, technology spending has been at the rate of in the last couple of years under the rate of inflation. So when you grow slower than the rate of inflation, you're actually shrinking. And that to me tells me one thing, the people in charge of organizations don't believe because if not they would be spending faster than inflation. So the myth that IT accelerates and drives your business, it's gone from that to this necessary tax I have to pay to keep the lights on and things running. For the first time in a decade and why I'm actually very excited about enterprise AI is that it can actually affect how products are built, not the back office, not the billing, not the marketing, not the paper flow, not the accounting, but the actual physical processes. I'll give you a great example. Last week was the GPM conference. One of our companies that we're investing. It's GPM. Oh wow, you've got another industry. JPM Morgan Chase, the largest health tech conference in the country. It's in San Francisco every week. One of our companies, Recursion Pharmaceutical, announced their large language model, which is an engine for small molecule, therapeutic and drug development. They have a model that they can plug inputs into, which allow them to come up with cures of cancer and therapeutics faster than they did when they had lab scientists and analysts. Yeah, that's great leverage. It's great leverage. They have data, right? They have an index. They have a huge proprietary massive data mode on small molecule biology. But now the technology can actually drive what they're building, which is actually vaccines and therapies. Not doing the back office and making sure Mary checked in and put in the hours and the PTO and all the things you think about. So we're finally getting to this kind of terminator phase where the technology, because of AI, is actually being infused into the products and services people are building and that they are primary. So we call that tech bio instead of biotech. I have an investment in a surgical navigation company that just had FDA approval that's operating on patients called Proprio. It is the Google Maps for surgery. It guides the surgeon in real time. It's like Waze. It's a little slow. It is actually Waze. Oh, we got to do a triple bypass while we're in there. Yeah, well, we're not quite on time. We're in the spine today, but we're working. So this is a fundamental shift, which actually makes technology more important. Well, this is why I want you here, because you're in your career. You've always been kind of at the early stages of companies and or Waze, wireless networking, networking, wireless. Cyber. Cyber. I mean, everything, you know. That's the DC with Nasera. Usually scar tissue comes with that too. And they're learnings. We are in another emergency. I'm only 25 years old. Look at the lines of my eyes. We're in emerging market now with Waze. So looking back and taking all that expertise, what is the critical success? Why are people getting excited? Because what you're talking about is a redefinition of workflows. So now that you have AI now part of the application process, you still need developers to code the apps. I agree. I'm going to back you up one second. Workflow is the IT combinational differences. This is a, I guess you could call it workflow. Or workflow or process, whatever. Process, yes. Something that's happening for me, the developer or app person. Or the biologist. Yeah, the user. Working at recursion now uses AI. I mean, yeah, they have a laptop they take notes on. But then the language model says, I'll give you an example. There's another company in our portfolio called Unlearn. They build digital twins. What that means is why instead of experimenting on Primate, they can actually run the lab experiment in a digital twin. They can physically simulate those things going on. To tie this back to Myst, just to prove I do have some memory for you, John. Myst was the first company that built predictive algorithms for wireless and edge networks. And they would see patterns. And they know at certain times of the day, they knew what congestions are. They knew when people moved around. And they were able to then reconfigure the network ahead of time based on that. And then obviously in real time. So the human behavior that it saw moving, packets into the network was actually being studied and they create their own version of a digital twin. This is why it was such a valuable piece of that because when you get to running, thinking about- I know AWS used them at re-invent. Of large scale events also are interesting too. And by the way, they also used my company Evolve for physical security. So you don't need as many security guards. 3,000 people an hour can walk through and they don't have to empty their bags and pockets. So things that used people. I always wondered why that worked. Because I have a bag and I didn't even get pulled over. Yes, because you are a handsome 161 pound bag of sailing with a little bit of hard tissue. It reconstructs your body and it knows whether you have an iPhone or a gun in your pocket. And that's what the algorithms and it learns all the time about the changes. So it allows people to go fast and it sees it into your bag as well. It uses an array of sensors and magnetics and but also- So more data the better it gets. Yes. And that is the answer to everything. The more data, the better it gets. And what you wind up is that companies as opposed to having just technology modes, they're all going to have data modes. So you're going to compete on your data set as much as you're going to compete on your technology. We need Jerry Chen here, because he's the mode king. The mode king. The castles in the cloud. So again, this data is the new modes. I love that. That is definitely happening. Now here's the little haymaker I'm going to throw out there. Data traditionally has been in construct and distributing as a database driven and Oracle and all the success there. Data warehouse has been out there. Snowflakes got the data club up there actually. Got to lock in. Data bricks is out there. So now, okay, this flips the script. So if I'm going to have like good carbon emission computing, not burning forests down to do queries, you're going to have to have iPhones running inference. You're going to have data that needs to be exposed everywhere. That flips the script on data management and data availability. Is it highly available or high availability? So all these things becoming into play. So what's your vision on how you see that? Cause you guys do a lot of big moonshot investments at DCVC where you're a firm. And so you've seen machine learning for years. So it's like, what's the script on the data layer? How does someone create a data model where an app developer can write a code so that if I'm at the edge of the network, it knows a lot of things from multiple databases instantly. It's a great question. The answer comes through is how does the model work? Cause the model is the intelligence and how does the model onboard it? So you may have seen one of our investments was a company called Mosaic ML. It was acquired by Databricks for $1.3 billion, right? Great founder, former Intel. Yeah, fantastic, right? And we've backed him before. I mean, it's a prior company. But what's- Was he sold to Intel? She sold Intel, he's got her out. He's great. He's an alumni by the way too. Naveen is terrific, really brilliant. But what Naveen solves for Databricks is how do I make it easier to onboard the model? So you're talking about how do I use the data? Part of that is how do I onboard the model? And do I have a structure to bring that data into the Databricks engine for creating those models? So what I would say is that we're still in the early early, despite all the hype, we're still in the early phases of building out the architecture for all of this data, making it useful and then making it useful for a model. And what is a model? A model is a version of how the human brain works, how it processes things, how it gets the synthesis and then ultimately brings you answers. When you do due diligence on a company, obviously, some people are very light AI, I mean, we call our cube AI, cube AI and essentially transcripts, it's got some... You know that I've got 70 something clips, right? You knew right away, right? Well, that was the data lake. I just thought it was a query to the data lake, but the AI actually has... Made that very easy for you to query, right? Well, we have now perplexity kind of experience where you can type in a query, what is, how do you get the results with multimodal sources? But it's not real AI yet because we're just ingesting and we're just doing embeddings. It's more rag, it's, you know, retrieval, but it's essentially a wrapper, right? So we're using AI, open AI, Anthropa, we're expecting all the different models. That's not yet true AI, although it's a good app. So I was, and Gensli was against AI, I thought they were kind of trivial, but it's almost like a web page. You're in the web days. If you had a web page, you're in the web. You're not actually a web company, but you have a web page. But the thing that's a little different here is, so you know that I spent two years on the board of LLM and AI, which is Yashua with Benjio's company. So Yashua with Benjio, Eve Lacoon, they were students at Gert Jeffrey Hinton's, they're inventors of what we call AI. They created neural networking. Yeah, I wasn't going to the whiteboard correcting the alagos, trust me, I was learning. And but the thing about, you know, when you see a neural network, think about the term neural network, it is a model the way your brain models information, but things also have to happen. You know, neural networks and AI has blind spots. It has lack of judgment, right? But there's an element of reasoning in good AI where there's some meta reasoning or some reasoning, not just retrieval. It's not just, it does, it does reason. That's the whole, why they call it neural. But they, but just like your brain is you do retrieval, but you also have other executive functions in your prefrontal cortex that allows you to say, I'm not going to get angry and jump over the table, right? You know, like somebody did to me. Yeah, and I can use some help on that. And those parts are still being built. And so the, what you're seeing right now is the, you know, the beginning. Will data sets merge and have fusion? If I have a great cube linguistic model around B2B jargon, you know, I have all the jargon of the cube interviews. So one of my motes is I have good word combinations, but I'm not going to be the comprehensive, large language model. I don't have enough data to do anything, but maybe I integrate with the model. Do you see models talking to each other and passing parameters and learning from each other? Yeah, I mean, all the time, you know, one of the things one of my companies did, they were looking at, they were doing AI work a couple of years ago and retail, they bought out 20 years of serious point of sale information. They told you what was sold, what time of day and where. Right? So you can, that was used in training a larger model. So there are data sets that you could acquire. Like I suppose, you know, you could buy the, you can buy the Andrew Ross Sorkin CNBC data model and you're giving perspective, he gives perspectives on the market, you give perspectives on the market. You could be merging those two. I'd definitely kick his ass on all the commentary for sure. Absolutely. He rolls on the enterprise. Like Andrew, you're stepping the side. So yeah, no, but I think, you know, there's enormous activity going on. You know there's enormous investment, there's enormous hype, but there is real fundamental change on how things are done and there are good data modes and then there are kind of some worthless data modes. How should entrepreneurs and business executives and or developers or people who are leaning into AI to get on the right side of this wave, to ride the wave and to create value? How should they be thinking about what they're building from a durability standpoint, sustainable standpoint, funding standpoint? What's your take? Because you know there's a lot of noise right now and I know a bunch of startups in this market that are kind of in that net of bad companies that are actually trying to fight their way to get a path. There's a lot of noise around companies that are walking dead in the kind of the last model seeing the numbers there, but there's still a lot of promise. So there's a lot of entrepreneurs on the street right now and business people scratching their heads saying, I got to get out of this. I want to get to the right side of the street here on the AI. I want to be on the sunny side, not the dark side. That's failing. What's your advice to them? What should they do to demonstrate how they should tell the story? Is there any tips? Well, I mean, you know, there's a couple of things that look, today's Monday, right? So I, before I came over, we were in our deal review. I serve four deals. They're all about AI. That's all we see these days. The first one is are you actually solving something really matters? Like I don't need the 13th business plan to tell me how to write a paragraph better. Like I'm going to be better than Grammarly. Here's why, right? I mean, because yeah, I mean, to be honest, it'll probably all wind up in Microsoft Word. Right? The spell check on Microsoft Word will be the AI writer. Clippy becomes a company. Yeah, I mean, you know, so that's going to be, you know, so is it an area that I have a proprietary advantage in? And then number two, is it real? I mean, it's not any different than anything else, right? Is it real and pressing that, you know, if I don't solve this, like, you know, the world is less better off? Will people absolutely rush to want to use it immediately? Is this a, I mean, there's the old adage of the aspirin and the vitamin. It's true in AI as well, right? And steroid too, the growth thing. And by the way, is it going to be efficient? Because like, if you look, I mean, chat, I love, I mean, if you look at open AI, absolutely amazing. The most amazing thing about open AI is they did $1.6 billion of revenue last year and you're sitting there with me. I'm paying $20 a month to, you know, to get a better search than Google or have somebody write a paragraph for me. So they have built a, and obviously there's, I want to use your experience game. Yeah. Big time. Except it doesn't work for everything. Yeah, exactly. It works if I have a paragraph. Once if I need a table and I need to see things side by side. So if I'm in IT and I'm saying, I'm evaluating 12 products, I don't want to read 14 pages about it. I want to see them beside by side. I want to see, you know, what are the features and capabilities they have? So there's still plenty of, you know, where to go on the UX side. But they did make it bone dry, simple. Well, I mean, this is why the web is a good analogy because the web was nascent early and it was like primitive, but you didn't need to solve the connectivity problem because that was getting better as more people came online, but the experience got better. And you can argue that the AI is good enough. Well, cool. I'm writing some summaries. It does help, help me. And this is my next question, which is, you know, in all these major inflection points, there's an element of a pattern. If you can reduce the steps it takes to do something, okay, and it's easy to understand and use, it works. And so in this case, ChatGPT, it helps me get some content out there. Also goes through email lists. It cleans up spreadsheets. You can tell- Oh, it does all kinds of really marvelous things and apparently- A lot of grunt work, a lot of heavy lifting as they say. So yeah, there is always- Small utility. So in that case, Chat and some AI is done what technology has done since IBM launched the 360 in the 1960s. So what is that? 60 years ago? It's getting to be a long time. It removes crummy, low-end manual tasks through automation. So technology has always used automation to basically speed up things and make things happen faster. Sometimes it works really well. Sometimes it's actually worse. You know, we've all been through this where you've written 13 drafts of something and if you had written it down with a pencil and a paper and done it right the first time. So you're automating and I'm spell-checking and I'm cutting and pasting and I'm shipping and all these people but I'm not actually thinking. The second thing that I think has to happen is it actually has to mirror your perspective or your needs, whether you're a consumer or you're a business. I always love like all of the review sites and I always call this the tragedy of tripadvisor.com or Yelp. And you say, hey, I'm gonna go to this cool restaurant down the block from the Cube studio. Looks really good. I love Moroccan food. And I read the first 10 reviews and like five say it's great and five say it's lousy, wisdom of the crowd, data moat coming together. But none of them say whether I will actually care. And you know, like, I mean, the reason why people tune into the Cube is they see guidance from you and the people you bring in that help them in decision making. So just aggregating data and presenting it efficiently is a piece of that. But then there's perspective and we're not yet at the perspective side. If something is very technical and very, you know, very scientific, having a clean data set, accurately presented, sourced massively is a fantastic need. But it doesn't tell you whether you should do it or not. Well, this is where your view is. I like this angle of increasing the productivity to get to value, right? And so the time to value can be accelerated on that next step. That's what's coming. That's what's coming. So can it actually reflect the question you're trying to answer, the decision you're trying to make or the data that you're trying to say is good or bad? Because AI doesn't know if something's good or bad. It doesn't have context. We're a long way, I think, from real context in these areas. But if I'm running a network and you're saying this is gonna overload the network and this is not, I don't need a lot more context. It's like not doing it's a good thing, reconfigure it. So MIST is like, okay, change the channel assignments, change the transmit power, be able to bring on this large crowd at this time, distribute the traffic. That's what it did. Because the context was just make it perform. But when the issue is, should I take that medication or not? Should I make this investment decision or not? Yeah, different story. That's where the intelligence needs to be. And that is the part, we're not yet at the quite a determinator phase of AI. We're the terminators. It is better to kill all the people to keep the planet safe. Yeah, you know, that's the meme, the guy's job is to pull the plug, right? Yes, I think so. The new platform engineering job. Al, great to have you on. Well, to wrap up, great chat, very podcast style. Love the conversation. You're awesome. As you look at the HPE buying Juniper, it speaks to the consolidation. You got a perspective of management, make that acquisition work. And we've got the competitors like Cisco out there, looking at this and maybe it's opportunity while they figure this out. But it's also there's a customer perspective. So as this industry is evolving and with this inflection point emerging around how people will do their jobs, how they will consume technology and whatnot, what are people thinking? I mean, how do you see that? I mean, what does HP do right now? How's that going to play out? How do you see that playing out? And how do you see customers reacting to this kind of big mega acquisition? Because it's consolidation. It's a lot like Broadcom. Well, look, there was a news, I think today about synopsis making a $35 billion acquisition in the tool space, like for building steppers and underlying, you know, goes into building all of this tech. So I think when you shift through errs of technology, there's always consolidation, right? Because what it does is it says, okay, we're kind of at the point that things are starting to become commoditized. So critical mass, scale, cost, reliability, the things that you expect from commodities. And I think a lot of the traditional client server to cloud industry is kind of in that phase, right? But now there's this new phase. So I think, you know, look, I think traditionally us folks who've worked in tech have seen this brand identification. I'm a pure person. No, I'm a Hitachi, right? You know, like in the storage, right? I'm an Amazon, I'm a Microsoft. I'm an Amazon, I'm a Microsoft. People are a bunch of brands. And then, you know, there were people I missed, but I'm my HP. So, you know, we'll be watching very carefully. It's an interesting culture. I mean, in networking, you can't please everybody. I mean... Well, no, I mean... It's kind of loyal, loyal base. So does Cisco. When we sold airspace to Cisco, I remember doing the first technical advisory board. So our biggest wireless customers come in and I give them the new roadmap of the two companies and they go, well, we love these airspace features. When are we gonna get this on our legacy Cisco equipment? I go, how about never? Is never good for you? And they were furious. I go, what do you mean? I said, well, look, if I try to write features for two different platforms to do the same thing, you're gonna get zero innovation. So you have to make a choice. Now we'll help you migrate. There are some things. If I'm a customer, I'll go to the cloud and I'll run Juniper switches in the cloud. Well, I will, as long as it's cost-effective. So the other thing is that right now cloud and on-prem... I'd be worried if I was a customer. I'd be like, okay, I wanna make... Cause Juniper and HP, I mean, there's overlap there, but also misses a crown jewel, right? That's gonna be a big piece. Do they lose people? Do they keep people? What's the migration look like? What a customer? People after integration into a large company? Never happens. Never happens. We'll be watching. No, yeah. So I think, you know... That's a big concern. It is, but for the ecosystem, there's a lot of really brilliant people who come out that folks like me get to invest in, in the future. I mean, why don't we just say the quiet part out loud? And so, you know, I think they'll make HP better and some HP customers will benefit. People that, you know, look, there were people that were missed customers, you know, that said, oh, why did I wanna be part of Juniper? Juniper really did an amazing job letting Mr. Run Mist became a driving force inside of Juniper. That may, who knows, that may be what happens inside of HP. Like Juniper may be the company that gives HP at second, when third or fourth, or 18 wins. Yeah, the RUPA has some gaps, a lot of complaints on the RUPA side, so... Okay, closing out. I was in that war too in the beginning. Final question as we wrap up. Outlook for AI, as you sit through the pitches and all your experiences, what does it tell you that what's gonna happen? How does this next 12, 24 months look like as people who are out there trying to figure out, okay, AI for good, AI for guardrails, and what is AI? How should people be thinking and framing in their mental model how to look at the opportunities with AI? And what are some of the cautions, if any, that people need to rent? Now, my opinion is, let chaos reign in the face. I look at your question the way people looked at SAS, the way they looked at Bandy Off 18, 20 years ago, and the way they looked at Andy Jassy and Amazon. And you remember like, hey, this SAS thing, we'll keep an eye on it. It's perfectly okay for salesmen, people don't buy package software anywhere, right? You buy it at SAS. And everyone, when Amazon Web Services was just doing S3, everyone says, oh, no one's gonna put mission-critical workloads in that thing, right? So I think people, this is an inevitable move. And it is going to affect everything. It is so powerful. And this is not a hype statement. This is going to consume a great deal of technology that we rely on as both individuals and as businesses. And there is timing that's appropriate. It's like, if you bought the model, if you bought the roadster from Elon Musk when it came out, it was probably that it broke down on Sand Hill Road, I used to see them. I was an early buyer of the Model S. It was pretty good. And then Model 3, just flawless, right? So people make their decisions on when to adopt based on when they're ready to adopt. But the adoption is going to be in everything. And, you know, did you ever have a stick shift car? Yeah, I loved it. Okay, when's the last time you drove a stick shift car? I can't remember. That is the point. That's not available. I love to buy one. 75% of the cars in Europe, you can buy one, but there aren't as many. But there was a point in time that it just doesn't make any sense. It's coming to tech. The tech industry, again, on the human side, I think it's going to be massive transformation. Again, this wave is coming. It's exciting. I personally think the entrepreneurial equation is going to be significant. I think this chapter is going to close very quickly of the overfunded zero interest rate period. And it's going to move quickly to value creation, open standards are going to swing back with open social dimensions for you those. We're watching those on theCUBE. And of course, the ecosystem, the new wind system of the crowds are going to be part of this. Now the data is going to be available. And again, The talent that entrepreneurs that we see now in AI, kind of like Gen AI, there was, I don't know, it was last one, Gen Z. The next gen is Gen AI. I think we're going back to the alphabet in front of the alphabet. That includes us old guys who have been through multiple cycles. There's a lot of older entrepreneurs now coming out. I'm seeing, I mean, comment on that. It's not just the young in the 20s. I just funded two people in their fifties that are unbelievably talented. And it's an AI play. And so what you see now is because of the disruption. It's just some, a lot of the older executives have systems thinking. That's right. That's a big part of this, isn't it? Yeah, but not everyone has the ability to completely decide on how to blow things up. That's, that is the young person. Revolution is here, Allen. Great to have you on. Great chat as always. Again, podcast style. Great, great masterclass there. Great insight. Thanks for coming on. Pleasure to have you. I'll come here. He's a general partner at DCVC. He's sitting on the other side of the table writing big fat checks for deep tech AI. And, you know, really moonshot projects that are now part of the mainstream. Once was once a moonshot is now a real deal with AI again. The world will be changing very rapidly. The scripts are flipping. Entrepreneurs in charge, developers in charge. If you have a data moat, that's the playbook. Thanks for watching.