Nutanix’s Lee Caswell on AI and Hybrid Multicloud

hi I'm James Maguire and on today's e beaks we're talking about a number of topics that all relate back to artificial intelligence and generative AI we're looking at the major shift in the tech sector toward decentralized Computing including the big generational migration to multicloud to discuss that I'm joined by Lee Caswell senior vice president of product and solutions marketing and nanic Lee thrilled to have you with us today I'm delighted to be here Jam great and then that is by the way that is that is your personal personal library in the background isn't it is I'm a reader a lot of books I'm a reader and there's even a guitar in the background so there you go we're gonna we're gonna have to get to the guitar at some point Lee I'm sorry I didn't warn you about that we're gonna to play a little guitar today terrific so before we get to my insightful questions what what is the brief brief elevator pitch for what newx does obviously it's a well-known company but for those who don't know it as well what does newx do yeah nanic provides virtualized infrastructure and so that allows you to run any application even the highest performant applications and we'll talk about AI in this uh you know in this venue uh and you can run them anywhere and this becomes important as customers are realizing they have multiple clouds for example today as well as on Prem and Edge infrastructure and so the ability to run that all together with one operating model high value going forward all right that makes perfect sense uh so you've written this sentence which I think is quite notable you You' you've said that quote AI will drive a decentralization of it infrastructure as AI models become more distributed so why why will AI models become so distributed you know what we're finding is there's actually relatively few companies that can afford to build what we call large language models right those are naturally suited for the public Cloud right you know 180 billion parameters for example right and hordes of data scientists right not every company can afford that but what we're finding right is that those large language models are now being made available to you know Enterprises across the world and now those Enterprises want to go and do two things they want to basically customize them introduce their own domain expertise right so it's specific to their industry so that's a tuning model right they also want to make sure their data is protected and that their data right is now an core asset and so that model all of a sudden looks like hey I'm going to go and customize that data probably in a data center at some you know some place and then finally I want to go and run these what we call inferencing so there's a training model and then your inferencing is actually running of the model and so we're probably going to run that now closer to where the data is being ingested so could be a retail store for example right where you're using generative AI to manage theft and Shrink could be a manufacturing facility where you'd be looking at looking at how you would do let's say maintenance procedures so we see this as a natural distributed data problem and a real value for a common Cloud platform across those different endpoints well it seems like one of the one of the problems or one of the challenges there is going to be is there going to be um uh the the processing power to Crunch the data there in that distributed world is is that going to be an issue it's it's a really hot topic right now I'll tell you right right so certainly you know there's gpus are being used like largely for this and you know they're incredibly powerful but also relatively expensive and they use a lot of power too right so so customers are really looking at what do the models start to look like right large language models in the public Cloud now I've got smaller models or compact models they're called right and so we a really interesting uh debate right now around hey where do I need to have gpus could I use CPUs for example right and are CPUs going to have enough processing power are there different ways this is a really interesting computer science you know discussion right at the moment what we're finding right is that that AI is coming to every industry and we're going to solve this as a way you know basically on the economics of how to make it you know feasible for every industry you know you mentioned large language models and I would love to take a short detour because I think it's such a Hot Topic these days I I think the companies are struggling with you know we know we need a large language model are we going to buy one off the shelf are we going to rent one are we going to make our own it's also expensive we don't really get it uh but it seems like that the large language model becomes the the there's proprietary Advantage if if you got the Bucks to make a good one you can really get ahead in the world of AI and any thoughts about that world that there just a challenge of today's large language models well I think the issue right is that you know competitive Advantage isn't created by doing the same thing that everybody else does right right that's R rarely the case right and so just like you go and take your Erp system or CRM system or right what people have used in the past um you basically customize for your industry right so if you're a Federal customer right and a dark site you're going to be thinking about this differently than a healthc care provider right now and you've got hippo requirements for example around where your data and how it's being used right or in a retail environment how you start thinking about how do I make sure that the point of sales system I have today is also able and capable of running generative AI right so one of the big solves here is to think about how do I find a platform and ideally a cloud platform right that allows me to have consistent operation across these but allows me to customize at the same time all right makes perfect sense um I want to want to get into the larger question about the virtualized Enterprise it infrastructure but let's make sure we drill down to new tenic how how is new tenic addressing the AI and hybrid multicloud needs of its clients what what is the new tenic Advantage well we announced something called G GPT in a box right and it's not literally a box right but the idea was that it's a prescriptive way to say hey you can get started today what we found right is that only 10% of the customers we've surveyed right were actually going to build their own large language models so they're going to take large language models and they want to get started hey how do I get started customizing right so one of the things you can do here easily right is you can take standard servers I talked a little bit about you know the power of using standard server Hardware now you can take servers with gpus very you know fully certified by nanic and you can build a GPU enabled cluster and we've got services that allow you know us to help you now think about how to go and build the application now a lot of the early applications James are basically chat Bots for example right to simplify customer service for example or a lot of video generative AI for things like I imagine managing shrink for example and so customers are looking at those and saying hey nanic becomes a trusted advisor to help you get an AI application going and you can start small right that's a really interesting way to go and give you an option to go and get started today well that's pretty amazing let me make sure I fully understand so it's you're calling a GPT in a box yeah it's in a sense it's it's sort of like it sounds like it's a low code artificial intelligence development system it really is right so you know the way this uh you know the way nanic works our Cloud platform software Aggregates standard servers together servers you can add gpus to those are all qualified today from all of our major Hardware partners and then our software then takes AI as the next application turns out like the way our software has been used in the past virtual desktops was a big landing spot then people brought in databases and Performing databases then they brought analytics like Splunk for example and now the latest application is AI and so by adding gpus and helping size and configure what your application would look like we're helping customers get started saying AI is the next application you'd run on the nanic cloud platform so you you use the example of a chat bot which makes perfect sense what I'm thinking of this hypothetical customer sitting in front of the thex GPT in a box what else might they possibly build with that using AI well I gave the example of a like maintenance monitoring you know ai's been around a long time right you know I recall doing like locomotive automated maintenance for example right you know a long time ago and you know but what's different right now is that AI as proven by GPT is accessible to everyone people see the power of it and how fast it is right right so now what you might want to start looking at is hey can I go look at for example in a manufacturing environment and start looking at the output data that's coming from a manufacturing line and start looking at when predictive or preventive maintenance would have to be required a lot of cost savings if I can do that intelligently right or or maybe I'm on a wind farm right or in a retail environment think about what point of sale data is coming in that makes me want to think about hey a different endcap display and how can I go and present that so there's just a tremendous amount of opportunity for applying the speed of AI into environments of all sizes yeah you talked about nenic uh enabling this virtualized it infrastructure in in a cloud environment seems like a huge need for that how does that technology relate to something like AI Ops which is the artificial intelligence trying to the entire it infrastructure is there overlap between those two worlds or what's the combination or or difference yeah a little bit so you know I think of AI Ops right which is operational basically being automated and predictive um you know improvements right um pretty interesting to think about you know I'm talking Futures here right now right you can start thinking about now that you have a hybrid multicloud which now you start thinking hey I've got multiple clouds I've got my data center I've got my my Edge start thinking how do I optimally locate applications and data over time and you know you might think of I think of there are four key areas where you might you might think of it for cost a lot of a lot of focus on cost these days right around the cloud was you know very agile but uh pretty expensive right so I might relocate for cost how about performance it turns out like modern users are extremely sensitive to Performance in fact the fastest you do something becomes the standard by which every other interaction is measured right right and then you've got data sovereignty of course right and uh one last one I'll introduce is sustainability you might start thinking about how to go and locate applications and data based on the power and cost right of that power and also you know sustainable goals right that you have and so there's a really interesting opportunity of automating this in the background making recommendations and then being able to have the ability to basically derive these and put to you know place these optimally across the multi hybrid multicloud environment really interesting idea because I I sometimes hear like oh you know the myth of of application portability in theory one could move it between hyperscalers but you know you you really can't you know the the new boss is the same as the old boss and just you're stuck once you have your but you're saying that this this theoretical AI system is going to move this heavy duty application perhaps between hyperscalers here's what's available today right always look at the standards that become available that make portability possible like tcpip right became a really big thing for you know being able to move like you know from a network standpoint there are two standards now to watch carefully one is kubernetes you'll hear about kubernetes you know over and people like what is it well it's really a way to orchestrate these new what we call containers that new applications are based on containers because you could develop them more quickly you can place them right across different environments more quickly so kubernetes this is now a standard for portability for applications right right so now I can move applications more and so this is really interesting Enterprises have have standardized on kubernetes different development environments right but kubernetes is a standard for portability of applications similarly there's a standard for portability of data and it's called S3 and so S3 compatible data can now be floating so we're going to basically free ride right you know look for how you can use these new standards to make applications and their Associated data more portable right across the hybrid Cloud nanic can help with that all righty well let's look to the future I mean that's that's there there's there's massive amounts of money out there so um you know Lee tell us where where is it all going and and where can we put all our money I mean the future of AI and hybrid multicloud that probably is about a 4-Hour question uh if you just address it in just a short amount of time what are some key Milestones you think you might see in in the years ahead well think everyone is wrestling with predictability and how to go and forecast the future right it's an unpredictable world right now right you know economic uh changes all over right uh you've got uh Global geopolitical events um and and you started thinking about skill shortage as well right and so what we start to look at is customers are looking for how do they apply their scarce skills in new creative areas and stop doing things that were actually less Val and so I what I see us doing forward is everyone acknowledges right this is going to be a multicloud world you're going to have you know resources across different uh different clouds the on-prem data center is not going away right and the edge is growing right where 50% of data is forecasted to be at the edge so where does it go I think what happens right is you start thinking how can I get this portability platform in place and tap into the power of server right so a server based architecture gives you the flexibility to have the degrees of freedom you're likely to exercise over time without locking you into deciding today well I'm thinking I mean I think about the future um there's this future image where we almost leave servers behind they're they're a commod commoditized component of it all but we're we're going to stop thinking about servers in the future are we not I I think so right and you know one of the advantages right is it used to be that you'd buy this big refrigerator sort of thing and plug it in and it was there for next five years right right but there was a great risk that you bought the wrong thing so people tended to buy a lot of it right and now with server based things it's like hey these are like chicklets right I can basally bring these in in small increments I can scale really seamlessly and easily new technology comes seamlessly integrated into these right so I don't have to worry that I bought the wrong thing you can always scale and basically build out of that that's a huge solve for customers who are worried that hey I I KN I know I need to invest but I want to make sure that I bought the right thing and I can scale over time makes sense you know in closing I like to share with you sometimes a doubt I have about the future of Enterprise it I think sometimes in my in my less optimistic moments I think we're almost going to collapse based on the the the total complexity of it all we're not going to be able to hear it's all going to be so complex the system is going to collapse under the weight of its own complexity I guess the answer I mean the idea of a virtualized infrastructure that can span clouds in the edge is is an optimistic view it it almost makes it all attainable uh but I still sometimes think H there's a the level of complexi is in a hocky stic moment I'm not sure that we humans are totally equipped to handle that all uh thoughts about that very broad idea well you know you can look at phones as an example right you know like phones and and you know all the new technologies that came into a phone right so now like I don't have to have a separate flashlight for example or a separate Compass right or a separate right I mean all of those things got in but there is by the way a learning curve right you can see this right as you know folks uh you know trying to make sure that they can actually you know take advantage of all the capabilities but what happens is once you get to a standard platform that you become comfortable with now all of a sudden it becomes easier to assimilate net new things into that platform particularly as it rides right across in our world right across multiple clouds and that's a big solve because otherwise what happens is you have to learn every cloud independently you've got to learn every on-prem technology and like you said you you basically run out of Cycles right so you know basically standardize on the parts that you don't want to spend time on and then you can focus on the new things well I I hope that Vision comes to pass that that is a positive Vision uh lee You' said it I've learned a lot uh thank you so much for sharing your expertise today and please come back and talk to us again sometime James always pleasure thank you so much

This transcript was generated automatically from the video's captions and may contain errors.