 Okay, welcome back everyone to the live here in the Palo Alto Cube Studios. I'm John Furrier with Dave Vellante. This has been a great day. Vast presents build beyond with the final wrap up analyst panel here. We got Dave Vellante, Rob Streche, filling it from Merv Adrian who could not make it. He had a personal challenge. He's going to be taking care of some stuff at home. And Sanjeev Mohan with Sanjeev Cube alumni. Thanks for coming on. Appreciate it. Thank you very much. It's such a pleasure to be here. So you've been watching and absorbing all the data. Dave and I have been talking for months and months about data developer. Rob, we brought it out at Linux Foundation. The data developer new concept and the storage company becoming a platform. Global platform Sanjeev, what's your analysis? This is the most unique thing. In my career, I've seen a lot of data companies start expanding beyond their core capabilities into the stack, if you may. This is the first time I've seen a storage company come into the picture and start offering all of these new capabilities and these new, actually the capabilities by themselves are not new. The way they are offering them are very new. For example, I have heard more no's today than I have in my past. For example, we do tiering of data. There's no tiering in this case. We cash data for performance. There's no caching. We move data between operational and let it go, no ETL, no data movement, no data copies. We partition data to make it geographically available, but no partitioning. That's like six things that we take for granted are not happening in this new vast platform. So that to me is something pretty unique. There are multi-billion dollar industries that have been formed on all those nodes. Exactly, yeah. This is all of all those problems. This grip has been flipped. We heard the CEO run and say earlier, Dave, we're flipping things upside down. This is interesting, disruption. Well, so you've got Snowflake coming at it from data management, and I guess the cloud guys too. You've got Databricks coming at it from a data science perspective. Now you've got Vast coming in from the storage. I used to think this is what EMC was going to do. They bought Green Plum, right? Remember that? Pat bought Green Plum and you had HP bought Vertica and other infrastructure company. But they didn't know what to do with it. They really weren't getting into sort of data platforms other than storing data, Rob, right? Yeah, and I think that a lot of other companies are attempting to move this way. And I think like you were talking about in the previous things, I came up with kind of a quadrant of where are the storage vendors moving up and how close are they getting to being data platform versus the data platform, data lake, data warehouse, mesh folks moving down towards the storage. And I think that this one is the furthest along and I think it comes back to the transparency that they provide to the end user. We say furthest along, you mean vision-wise. Vision-wise, yeah, and I think that from a vision perspective, execution is still to come. I think we'll see, but the fact that it could be transparent all the way up to the data developer is huge and I think that way you don't have to worry about and I think we were talking about this earlier about the size of the files, being sub-parquet file level and having that performance or having one of those knows where, hey, the metadata can be actually stretched across. It's huge, I think that's huge. Well, the green plum, you mentioned green plum in vertical, those were columnar databases, so great performance but we heard Jeff Denmark say the transactional piece is what they bring to the table. Little complimentary there, so that's one I'd like to get you guys' reaction to. The second one is, there's a lot of knows but they're still doing NAS. So we heard Jeff also say you can do NAS everywhere now. So does the architectural shift where there's no, all these things that you talked about make NAS viable now at scale? I think it does because, but the way it makes it viable is when we did network attached storage, there was a lot of intelligence that would go into it, like how do we distribute the data? What VAST has done is they've abstracted all that. So with this whole concept of a global namespace, they may be doing NAS behind the scenes but I as an end user, I don't care. So I get NAS without having to pay the price. So NAS is not disrupted, it's just abstracted. Yes, correct. So that's interesting Dave, because we hear people say, oh, we're going to disrupt the NAS market by taking it away. Yeah, why disrupt? Like disruption is not the cool thing in real life. Would you want disruption in your family? We love disruption, we're the cute thrive on disruption. Absolutely, we love disruption. That's why we have a job security because of disruption. Chaos equals cash, someone wrote a post about that. But in the same way, a lot of the disruption in markets comes from being non-disruptive. Yes, if you could do it in a more BC red, why not? I mean, there's a backup example between like data domain and Avamar, right? Data domain, they just basically look like a tape interface and there's a narrow example. But Avamar, you had to change things and who won? Yeah, and I think a good point and they had one of their customers on with the Allen Institute. And I think if you look at what that BioLife Sciences use case is really interesting where they talked about the gene sequencers, they spit out terabytes of data per day. And along with that, at the end comes a file of metadata that kind of tells you, hey, all the files are there, these are the files and here's all the metadata and all the descriptions of the files. That goes into a project and then you go and do your, the data science goes on that project. I think what's really interesting is that they talk about that being, that's a file, they're all files, it's a file use case, getting closer to the gene, the sequencers and being able to know and trigger off of that. So bringing the triggers, bringing those functions up to the sequencers, huge for these companies for speed of analysis and speed of drug discovery. I think that's really where this is a platform and not just storage. And I think the story, and I'm all kidding aside, I do like disruption from an innovation standpoint, because things do get disrupted, abstraction is disruption, non-disruptively means operationalizing it. They got all the standards, they checked the options. But I think that jumped out at me is this idea of the data engine one, the relationship with NVIDIA and the fact that NVIDIA is an investor and a major player in the platform and the fact that their customers are experiencing value at scale. So the question is, does this open the kimono for a new kind of approach that's going to open up the business side and saying, hey, let's rethink how we handle our data. Because AI is driving a lot of change too, Sanjeev. What's your analysis? Yeah, so my analysis is exactly to your point. We see there's a massive category confusion in the market. What category am I in? And categories keep shifting all the time. The moment you think a product company has achieved a category, everybody jumps in and then gets overcrowded and then you're rushing out, trying to find a new category. The advantage lies when you can have a full stack. So we all grew up on OSI 7 layer stack. So think of an OSI 7 layer stack. At the bottom, you've got the physical layer. That's all the chips, whether it's NVIDIA or it's TPUs or Inferentia 2 or whatever it may be. So you've got the physical layer. You've got the data layer. So you've got to own the data layer and VAS today showed their whole database and data store. Then you've got the metadata layer. And to your point, that metadata layer and data layer should coexist. It shouldn't be that your metadata is sitting in a SharePoint file somewhere, it's out of date. So if you go up the stack, then you've got the AI layer which is where a lot of activities aren't today. But if the AI layer does not have hooks into these other layers, then it's a standalone kind of a thing. And then you move up to application layer and then an ecosystem and you've got the seven layer stack. And if a company can operate in this stack, that's the road to success. And it actually goes back to what you were saying about our previous discussions when you start to look at platform engineering, right? Which we talked about at Length at KubeCon and Linux Foundation in Vancouver and all of these different places. This is what people want out of platform engineering is abstraction. I don't want to know it's a NAS. I want to be able to build my application against the data and go and do that. And I think this is the case. And Cloud did it, right? And AWS does NAS, but you never find out. Well, these guys are the first ones to use data developer term we coined on theCUBE. Two, data engineering is taking the same track as platform engineering. The security industry shift left was born out of the demand of taking that separate team, bring them into operations to feed the developers in their pipeline the ability to make decisions with guardrails. And that's coming with data. And these guys seeing that and their customers on the keynote were telling us that. And I think those advanced customers to me are a signal to what the mainstream enterprise might grow up into. For example, and this is where I want to get you guys reaction to. People would store data for a couple of reasons. Compliance, store it, we might need it later to prove some GDPR thing or whatever compliance or legal reasons. Got to protect ourselves, store the data. Now people are storing data for innovation because they see value in the data. And we heard that from Pixar. So now you've got customers that are storing data for all three purposes. Like AI legal, explainable AI. Did we use the right licenses? So we're going to a whole nother level of data intelligence Sanjeev where it's like the old way might be completely irrelevant. I mean, this is billions of dollars of industry being recast in just, if you believe what I just said to be true, that's disruption. Yeah. Well, and I think to the top of your OSI stack is critical for companies like Vast who want to build a platform. You've got to have an ecosystem. And I presume they're not going to do all the security and the governance and the lineage. I mean, they'll do some of that. And obviously AI, they're going to partner for AI. So they've, they had to announce this, right? And show the world. And now the execution piece comes from in part anyway building out that ecosystem. And that is the hallmark of a successful platform. Yeah. And I think in one of the slides that was in the keynote as well where they talked about Presto, Trino, Spark. So I think it's not about doing it all. It's about being part of the ecosystem. And, you know, with Trino, they had Starburst listed up there. So when you start to look at some of the ecosystem that they look at being complimentary and being, you know, building upon that platform, I could see them saying, hey, listen, you go build your Spark on us. It's going to be 10 times faster and 3,000 times cheaper than doing it, you know, on Amazon using Databricks. Well, where does their ecosystem come from, Dave? Does it come from new players, new entrants out of the organic growth of the platform? Or they come in from industry, like database companies, you've got storage companies out there. I think it comes from software, you know, attracting software companies. Yeah, 100%. Come into my platform, right into your capabilities into my platform. And, you know, Vassar has to figure out how to make it attractive enough for them to do that. Part of it is product. Yeah, so ecosystem becomes really important here. Because, you know, we talked about the seven layer and you were mentioning that in each layer, we could open it up and it's a whole world right out there. So if it's physical, then there's all the security, the networking, all that. If you look at the data or the metadata layer, you've got the data engineering world surrounded by the orchestration companies, observability, data catalog. There's a whole slew of a world out there in each of those layers as you move up. If you look at the AI layer, there's MLops piece and we know how many companies showed up in MLops and now there's a whole LLM engineering responsible AI and off you go into serving architecture, inferring architecture, so Vass plays across the stack and then they need these ecosystem partners to provide these ancillary services. Yeah, and I think an important part on top of that, layer on top, is the personas they're going to be talking to and how they go to market. And I think you brought it up and George talked about it earlier as well. It's going to, they're going to need that ecosystem to go out and have these conversations. They can't do it alone in there. So the question for you guys is this, do you think that these guys are a breakthrough in data management? Are they streamlining this enough with the elements that it's going to change the game in data management? I think they're going to make people think a little bit differently. I think, it was like when we were talking to Ren and I was asking them how much luck, how much good. I think they had the foresight to say like, look, let's bet on this architecture that is going to be appropriate for AI. He even said, we didn't know when or even if it was going to happen. And I think it is happening and it's happening in a way that it's clearly not a bolt-on. It's not like, oh, hey, we're AI too. There's not a lot of AI washing going on here. They're basically saying, look, we're the infrastructure for AI. You've been talking about this for years now. So I do think it's going to cause some people to think differently about how they approach this and they think you're right. It comes down to execution. Yeah, and I think it comes down to them partnering up in the ecosystem and not trying to do everything. Having focus and I think from that focus becomes success. But I think all of the other storage platform vendors have been talking about this. I think, but have never been able to execute on this. And I think that's been the key. You hear it with GPFS and GPFS everywhere and other stuff. And I think it's, that's great, but that's one little tiny piece of it. That's not a platform. That's a file system. Yeah, the data engine jumps out at me. The data space jumps out at me. The engine, you got some compute vibing. I love the functions and triggers. That gives you programmability, automation. Sets the table for kind of self-driving data, if you will, pun intended. Then you got the data space, which I find interesting because we've been talking about data clouds for years. Now you have data interactions. So as you have data that has to interact with each other, having this distributed global name space or data space allows for data to kind of sit anywhere and work with each other and still be small and manageable or big and manageable. But when you start to address that, how does that change the game? Can you see that? What do you see in there? You're in the game, right? I mean, clearly. You got Databricks and Snowflake and the Cloud guys. Right. And then you got all these other database companies that are solving problems, but they're not considered like the data platform. So they are vying for it. Vass now comes in and you look at the other storage companies, look at Pure. They're not in a position to do it. Dell is big enough and HPE's got their Esmeral thing, but that's not in the conversation the same way this is. It's more like, okay, we've got a captive business and we're going to sell into that captive business. Snowflake and Databricks have a bigger platform and presence in the market. Now you've got Vass coming in from the storage. I just don't see another storage company. I think it's going to be tough. Even thinking about doing this. Could it be Innovator's dilemma? I mean, they are too big. Like Dell cannot just overnight try to retrofit itself into what Vass is doing at the cost of their existence. Well, it's interesting. Dell is an investor in Vass. HPE is an OEM of the Vass file system. And it is an investor and partner. So these guys obviously looked at this and said, there's something here. And you wonder where that came from. Was it the investment arm? So who buys them? Who buys them? Which one buys them? I think they... They're in the public? Look, Renin, the team, Israeli company, he said, I got to move to New York or whatever, New Jersey. But this is going to be a U.S.-based company. And we've seen how many times have we seen some of this with Aguazio, right? They kind of tried to have some big ideas. But they got to find... In-depth of J-Frog, well, all the other ones. The vast majority of these companies get buffed. But I think that Vass, no pun intended, could be ripe for an IPO when the market comes back. I think they've got enough market momentum. They claim to be the fastest growing infrastructure company in history. You know, it does happen. If they capture a big part of this AI data market that where it starts to start setting up, as they start building now, you start to see architecture decisions around business and technical around, how do I store all this data in a way that's going to be the lowest cost, highest performance with most agility. And this is DevOps for data. I mean, we've been saying this on theCUBE. But go-to-market stuff. So Vass has that product. They could run the table on these use cases. So it's got to go-to-market, ecosystem, I mean, all the other things that matter, right? Yeah, if Vass were to go IPO tomorrow, it'd be on the basis of the storage success, which is phenomenal. But the data story is yet to be written, in my opinion. It's immature. Yeah. Which is a data story. Their data story. Yeah, it just came out today. Right, I mean, like the whole... It's a good story. The story is great. The vision is amazing. These guys are amazing thinkers. But we know how long it takes. Like, you know, we just discussed how long, how many companies have tried and tried and tried. You know, Oracle is still trying to perfect its solution 40 years into business. So you don't just get transactional consistency out of day one. Also, operationally, they've got to get a sales team. They've got to get SES. They've got to build their go-to-market. There's a lot of work to do. Again, the question is, is that, to me, the AI tailwind right now, how real does this happen? We were talking about the chaos. Hemipillar in production. Whereas the revenue, Microsoft didn't see their earnings pop from AI this past week. So is the AI money coming? And when? When does that hit? So it's like, you got Databricks said, okay, we're going to go buy Mosaic ML. You got Snowflakes saying, okay, we're going to essentially containerize the NVIDIA stack and rely on that. And of course, they made another acquisition as well. But look at all the companies. You got Dell said, we're going to do Project Helix. That's, we have to have an AI story, Project Helix. HPE at Discover. Well, we've got supercomputers. So we're going to turn HPC into an LLM business. Okay, we're going to take what we have and then we're going to promote it. With Vast, it was like, well, the market's coming to us. We don't have to make any kind of real changes here. We just have to present our architecture because it's suited. At least that's my take on this. Do you guys agree? Is it suited for this new AI era? Yeah, I think so. I think it's a new way of approaching it. I think the distribution and the chunking and the metadata management are the three magic pieces of it that are suited for this workload. I think to the other point, at what size? And it becomes a cost value conundrum. But also if you go into the repatriate act and stuff like that, where is the data coming back? And is this a better, more valuable way of doing it? And they don't care, right? Because it will run in the cloud, right across any cloud. Well, it is really... Repatriation question. I was just going to bring that up. Repatriates as we call them. We found that at AWS last week in New York when they announced the big gen AI summit is that there's a lot of right sizing going on in the cloud to save money. And we just did a big special on that. Cost of saving in the cloud. But they're taking that money to invest in AI. So the question is, is it the data that's going to be repatriated or the actual hardware? So now you have that repatriation equation. Because we've seen AI, people are deploying there because they want to over provision because they can't get enough of the stuff. So it's not like they're over provisioning NVIDIA right now because they can't get it. So if they have the gear, I can see that on premise. That's where most of their stuff is. So to me, the data and the platform is going to be everywhere. So right now it's on premise and edge. Yeah, edges. I mean AI inferencing of the edge is going to be enormous. And even if you're persisting a small percentage of that data at the edge, it's still a huge amount of data that's going to be persisted and certainly processed. So on-prem is not going anywhere. On-prem is going to stay important. I don't believe as much in repatriation. There might be some of it happening, I'm sure. But it's a tiny, miniscule percentage of the cloud migration. If I look at mainframe, the mainframe's going away. We've been saying it for 20 years. But actually the mainframe business is growing. Surprisingly, tiny amounts, but it's not going away. So I... Well, repatriation is semantics. But I'm going to pull back from the cloud. I don't see that happening. Net new AI capabilities on-premise is interesting. Edge, obviously, is not repatriation. It's never been patriated or whatever. Yeah, that's true. But I think the hybrid operations is the key. This is where I think this global namespace and the unification angle is probably the most important aspect of this story. Yeah, because AI is going to go where the data is. And data will have to be managed with commute and workloads. We didn't talk about workloads. So the next question for me is what's the workloads look like? Because when you move workloads to the data, that's a different discussion, right? Correct. And I think this is the most standout feature of what I've heard at the vast story is that as the data comes into the platform, the data engine is inferring the header or doing some inference and it's deciding what functions to execute. It could be compute. It could be data quality. It could be trained machine learning model. So basically data has become the center of gravity and everything is being driven by data. This is new. All this time, I as a data person have felt like a stepchild because 20 years later, it was all about infrastructure people ruling. Then it was all about application people stealing the limelight. So for the first time, data is stealing the limelight. Guys, we've got one minute left. BuildBeyond.ai is a website. Check it out. The vast presents buildbeyond.ai. Final minute, let's go around the horn here. Summarize, Dave, the data platform. What is vast presenting? What does build beyond mean? Go around the horn. Dave, start with you. A new way to think about infrastructure for AI. I think that's really what today was all about. Yeah, I think it's about data anywhere and bringing it to the right place, either compute to that, the data or storage to the, or data to the compute. And I think that's huge. I think it's making data a first class citizen and the driver of the story, AI story. I think data is the oxygen and the bloodstream that flows all through the organization globally. Has to be freely available in applications. It's going to be the rise of the data developer. There's going to be a collection point for data as code. And I think that's going to be a big thing. So go to buildbeyond.ai. Check it out. This is theCUBE's live performance here in Palo Alto for theCUBE team and for vast data presenting build beyond. Thanks for watching.