 Hi everybody, welcome to VMware Explorer 2023. We're here live at the green set with Rob Streche, Cube Analyst Dave Vellante here. This is our 13th year doing VMworld slash VMware Explorer. Really the same event, they just changed the name. Double sets, it's actually quite an impressive show. Last night, Rob, we were walking the show floor. A lot of customers here. The ecosystems may be a little bit smaller, but there were some big booths. You had some of the bigger partners that you would expect. Dell, HPE, IBM, Lenovo, NetApp, et cetera. Veeam had a big booth, Rubrik had a big booth, and several others. And it was packed. I mean, the customers are here in force, maybe trying to figure out, okay, what's next for VMware? Maybe they're up for renegotiating their ELAs. Maybe they want to hear what the future with Broadcom looks like, or maybe there's just more of the same, but it was actually more vibrant than I had expected. Yeah, I think it was a good showing early on for them, pre keynote and pre understanding what's going on. I think getting in there with some of those actual customers and talking to them about what they're thinking about and how they see this playing out to your point, what their expectations. I think there is a lot of learning going on. I think there's also people looking at what their job is today and what their job may be tomorrow. And I think there's a little trepidation around that is, hey, being a VMware admin, it's not as sexy as being a cloud architect anymore. And how do you make that transition? And I think they're here to hear the multi-cloud story that's really front and center. Rob, you and I in ZEUS wrote a breaking analysis this week. It's published on siliconangle.com and of course on our YouTube channels and LinkedIn and everywhere. But we said our expectations were, we said we're not going to hear much on the Broadcom acquisition. We heard more than we expected. Haktan was here, he actually was on video and then he stood up and waved in the front row. He basically said, hey, we believe in this multi-cloud era and a lot of opportunity and we're going to invest. I think he said $2 billion in R&D and et cetera. So very strong message and indicator that obviously things are going well. We had spoken to some folks privately yesterday who said, yeah, there's a couple little things we have to settle in the Asia pack countries. We're going through that. The FTC as we know, and we started out our pieces continues to drag its feet, but there's no reason why they won't approve this thing. The European Competition Committee, the UK rather Competition Committee approved it, I believe yesterday, gave their final approval. They had already given preliminary approval. So that's no surprise. So that's done and we did an analysis in the financials and we can talk about that. But the thing I want to talk to you about here is VMware's AI play, we're going to talk about that. They kicked off the event with all the CEOs from all around Silicon Valley. Who'd they have on there? They had Adam Solipsky, they had Silicon Valley extended. Satya, Satya was on there or not? No, Satya and Satya, TK, Thomas Kurian. Who else did we see on there? The head of R&D, group CTO I think at Hugging Face. At Hugging Face you had, and of course you had Jensen on stage, so they sort of had some good AI action. But what we said was we're going to hear more about cross-cloud services, Tanzu and Aria, which they're basically bringing together and developer extensions. Then we said incremental enhancements to the core. vSphere, vSan, and SX. And we're going to talk about that quite a bit. We said, I kind of had a question mark around it. I took it out, but around Project Monterey, our intent was, haven't heard much about that lately and we still haven't heard much, and I don't think we will hear much. And then the ecosystem showcase, which we can also talk about. But let's start with the core. The core incremental enhancements, vSphere, vSphere Plus, vSan Max, NSX Plus, let's start with vSan because that's where VMware started. Yeah, and I think with vSan, I think it is more than an incremental, I guess you could say announcement that they are doing this disaggregated storage play where they're allowing you to grow storage separately from CPU and memory. I think this is a direct shot at some of the other hyper-converged infrastructure players, HCI players that are out there. I think it's also a good way to be able to move between being, you feel like you're stuck on a single box and be able to move your VMs very easily using storage vMotion. That being said, I think, funny enough, I was at HPE, feels like decades ago and it was almost a decade ago that we did HCI using store virtual underneath the hood and did disaggregated where you could have a separate box for storage and you could have vSphere on top of it. So I think a big piece of this is how really innovative is that, and I think it's good. I think it's a little late to the market. I think also it doesn't go far enough for the multi-cloud vision that they have. So hopefully we're going to see some of that $2 billion that was talked about going into, hey, let's make this a true platform that can compete with an S3 or an EBS or an FSX type platform. So Rob, I remember when HPE, and it was HPE at the time, initiated that disaggregation of storage, it was at a big data show, where the term big data was in vogue. This is how long ago it was. And the guys came to me and said, we know you've been talking about this problem of separating compute and storage. We're working on it. We have this skunkworks and they wanted to pick my brain about it. This was like ancient history, to your point. And I looked at that announcement and I was like, wow, it's about time. I'm pretty sure I was on the cube in 2015 talking about it. Maybe in Barcelona, at VMworld Barcelona or something like that. I'll have to go back into the archives and look. But I think, again, it's definitely something that's needed. I think their customers will get great value out of it. I think it's a smart move on their part, but it's incremental. And I don't think it's earth-shattering. It's still not file for everybody, although you could theoretically use it. Object is through a partnership with somebody like MinIO. And really, the block is only for vSphere. And again, you start to look at it and how they've closed off their ecosystem and how the API set in the storage is really closed off. And it has been, since the EMC days, it started to close, they started to close the gaps, signing IO paths. I think the one thing, and we heard about this and we just mentioned it before on our other keynote analysis is they are good at IO. They can get high IO. And I think that does play well for AI and what their private AI strategy is, but there's a gap between the IO and the actual application. And there's that data layer. So, speaking of IO, when VMware in the early part of the last decade, so kind of 2010, 2011, even before that, 20, 2008, 2009, 10, when they were really on the upswing and IO was the big problem, right? And it was really complicated if something went wrong and you had to bring in guys with lab coats to figure out what went wrong with your VMware environment. And so VMware initiated and it did a really, actually really good job working with the storage companies. I think originally it was the VMware API for storage and then the Vasa API came out and they pretty much released the SDK to all the majors anyway. Majors being, obviously EMC got it because they owned them, but HP, NetApp, CertiDell, FreePar, they got it, they got the SDK and they all worked really hard and they became really good at solving those IO problems. It goes back to what Paul Moritz said back in, I want to say 2009, when he said we're going to build a software mainframe. The mainframe was really good at IO, it was really good at minimizing the impacts of virtualization on performance and that's essentially what VMware has done. Now, having said all this, I would argue that during the EMC ownership years, when you talk to the original folks that created VSAN, they felt that they were handcuffed a little bit, that they couldn't just go off and attack the market opportunity because their owner wanted to preserve some of the revenue for themselves. So they sort of neutered VSAN and said, all right, keep it low end, make sure it doesn't scale too much, don't encroach on our block business and now I feel like the handcuffs are off. Yeah, and I think for the last couple of years, that's why this is a first salvo I think for them into competing potentially in software defined arrays and being able to be on clouds with VSAN because right now the only way you can get VSAN really is on premise for that most part. Even they talked about there's ways to get some of the storage aspects of VSAN in cloud but at the same time, it's under the hood of the different services. It's not there front and center. VSAN Max will not be in the cloud during this first release and I think that's telling to potentially where the roadmap will take them. Yeah, and then the other two big core announcements, NSX Plus, which essentially is a networking super cloud where they're abstracting the underlying primitives and eventually they'll go to all the public clouds and they're going to take care of all that complexity. So there's an allure there but NSX adoption needs to increase as we know and they've got a lot of work to do there and of course VSphere Plus is, I mean, that's the core. Yeah, and I think the NSX Plus is a place where I'm very positive on that direction. I see it as one of the hardest places for these cloud architects to play. The networking is still really tough to go and do and I think, especially if you talk about multi-cloud or cross-cloud, even if you're say in an equinex using VMware there and using VMware in your, say VMware on cloud, the VMC in AWS, it's still hard to make the VPCs and all of that work together even with the physical links being there from equinex. Yeah, that multi-cloud VPC. I mean, basically the core problem they're trying to solve and they talked about this in their little breakouts yesterday. You need a common cloud operating model, that's super cloud and DevOps wants to move fast. So those are both true. And so you've got to evolve NSX to NSX Plus where it's got these kind of multi-cloud standardization. And I think it's moving it beyond just being a product to being a platform and a platform as a service and I think they look at what the clouds have done really well is simplifying it down to a couple clicks in the UI in their console and going and doing that. Now, how does that play if I'm going between one console and another console? Are you going to have consoles of consoles and how do you do all of this? Are they going to automate all of that through the Azure APIs and the Google APIs and the AWS APIs? They kind of have to, they have to. If they don't then what's the real value? It's just the manager of managers. They got to go beyond kind of policy management. They got to really, really create that true multi-cloud abstraction layer. And then there's a data lake which is providing intelligence which is very important to have the metrics. Let's talk a little bit about the AI play. Jensen's on stage, they're betting a lot of their AI on NVIDIA. Like a lot of companies, you know, okay, that's a good bet, smart to be there. But it's a safe bet. But it's a safe bet and you have to be there. It's almost like table stakes. It's not going to bring a ton of differentiation, but you got to have it. And then their database is Postgres, and I guess Green Plum, but Postgres is where they're starting. It's very functional and great Oracle alternative and relational capabilities and fairly mature, very mature actually. And then they've got Huggingface for their repo and that's where ostensibly they're going to get access to customers will get access to multiple LLMs. I would say this, and we talked about this again in breaking analysis. The cloud vendors, and we've got, you know, anecdotal evidence from some of the developers that we talked to, you in particular did this research where they're saying, hey, I'm actually quite impressed with what I'm getting out of AWS. I got LLM optionality, Lovin' Bedrock. I'm playing with all these different LLMs. They're fenced off, the LLM vendors are fenced off from getting access to my customer's data. So that's a plus, plus I got encryption at rest, encryption in motion. I got all the cloud security and I'm moving fast. It's making me productive. I'm putting out product. We're actually going beyond MVP. So very, very high marks. What you hear from the on-prem crowd, HPE will have LLMs as a service in fourth quarter. Dell, we got Project Helix and we got some reference architectures. You know, you got that same thing here. We didn't hear LLMs as a service. But so the point I'm trying to make here, Rob, is that the cloud has the advantage of innovation and speed. The one advantage, ostensibly, that the on-prem guys have is, and VMware trotted out its general counsel today and said, we've got the fud of fear, fear uncertainty and doubt of legal and compliance risk. So you better do this stuff on-prem, so wait for us. But the market's not going to wait for them too long. No, I think it's a private versus in-the-cloud type of, and I think it's private in-cloud or private on-prem, and that's going to be some of the big choices that people have to make. I think that VMware did a good job of talking, even in one of the slides, had Dolly, Falcon, Lama. They talk about, I think it was Platypus or one of the Platypus two from Boston University, Go Terriers, but I think it's one of these things that when you start to look at how they're bringing it together, it's still a 60-page reference architecture. And I think it's going to have to be simpler to be a platform as a service like Bedrock is if they really want to achieve where they're going in this market, they got to make it simpler. It just has to, it can't just be, yes, we use Nvidia's workbench. You go in here and this is how you're going to keep it secure and here, by the way, we encrypt it all, all the way through and confidential computing comes in there and how they're doing all of their encryption. So they do have a good story there that they can build upon and good infrastructure. I think it becomes a, can they make it simple enough to achieve that, to get there fast enough? I mean, the big concern I have is what is their differentiation in AI? Obviously they got great infrastructure and they can bring that infrastructure, evolve VMware infrastructure so that it's capable of handling AI. I don't question that, but what the gap to me is the data play. I'm going to bring my data to the cloud, I'm going to bring my AI to the cloud because AI is there and B, my data is there. I'm going to bring my AI to Snowflake, which is also in the cloud, by the way, because my data is there. I'm going to bring my AI to Databricks, whether it's in the cloud or on-prem, because my data is there. Now, VMware doesn't have a strong data play. Obviously there's a lot of data moving around the VMware infrastructure, but it's not in a cohesive way. There's no data platform play there and so I think they can build that, but it's got to be more than just, hey, we're doing Postgres. They've got to have tighter relationships with the data platform players and that means database, that means governance, that means privacy, that means data cleansing, data management, all that sort of ecosystem that exists that you see when you walk around Databricks and Snowflake shows and AWS and other, Google will see it next week, that's lacking here. Yeah, I think you're dead on and I think they don't even have the story about how you can get at it with SQL or how you work in these other tool sets and use those to get at the data in VSAN. Because I think it was, the play is very similar that they have VSAN, now VSAN Max, it's going to be disaggregated. What if you put these interfaces on top of it? SQL, for instance, and you allow people to do that. You have file already. Other file, Databricks uses a file system of their own that stretches over S3 or over file and actually runs really fast on file so faster than it does on objects. So if you start to look at how things could be built, that's a path forward for them that we're just not seeing that vision out of them this week which is a little surprising because you had even just a couple of weeks ago when we were with VAST, they were talking about this and same thing with IBM, talking about it the week before that. So you have Dell, IBM, VAST, HPE, with Esmerell all talking about how you bring the data to the storage and make the data lake be the storage and not separating those pieces out. I think the difference is, and it's true, virtually every technology company has been playing with AI or doing AI prior to the open AI announcement of ChatGPT. The difference is there's a certain sector of companies that were basically born to do AI and provide AI services to customers versus those that are embedding AI to make their infrastructure run better or maybe do some chat botting, whatever. But that's I think the difference. Look at VAST, but VAST basically has a platform that is a strong infrastructure for AI, at least ostensibly. IBM obviously has a long heritage in AI even though they sort of fumbled Watson, but they have a lot of experience there so they were ready to take it. To the point about Snowflake, they can kind of bolt on AI because they got the data play. Databricks has always been deep in AI. And they bought MomentSake ML so even give it another layer up. So you have these companies that are, have truly been thinking about how do we deliver AI services to customers? The cloud guys very clearly, really it's the three cloud guys and Databricks are the big ones. And then now open AI is shot to the lead as well just in terms of presence in the market and awareness. And then you had a lot of other companies, Anthropic and Huggingface and Cohere and on and on a C3 AI and DataRobot and all these specialists that are also in a pretty good position. Yeah, I mean all of them, yeah. Absolutely and so they've got good businesses but to our power law you're going to have the giants that and there's going to be a few of those very large language models and then the long tail is going to be very highly specific, maybe industry specific models and there are going to be a lot more of those but they're going to be a lot smaller. Yeah, I talked to three CIOs so far this week just in the last two days about this and what they were talking about is they're going to be very specific about how they use AI and how they build their models and their LLMs or specialized or segmented language models for doing specific things. So you're going to have to your point a long tail of very small models that don't need 10,000 GPUs to go and train. They need a couple thousand or not even a thousand or maybe tens of GPUs, not hundreds of thousands of GPUs. Yeah and Jensen up on stage today was giving away some GPUs that were signed by Jensen and Raghu. I do think because Raghu's a technical visionary, I think Jensen respects that. They got the Silicon Valley bromance going and so that was kind of cool. They've got five of them, they're giving away or raffling them away, presumably to partners or customers. But Rob, great analysis. Thanks for spending some time on theCUBE. More to come, wall to wall coverage this week from VMware Explorer. We're live, theCUBE's coverage 2023 VMware Explorer. We're right back right after this short break.