46:20 Webinar
5 AI Predictions Every Infrastructure Leader Needs to Know in 2026
Join us as we explore five trends shaping the future of AI infrastructure, from unlocking the value of your data to architecting for production-scale AI workloads. Walk away with a clear roadmap for positioning your infrastructure as the engine that powers your organization's AI ambitions.
This webinar first aired on January 22, 2026
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
Hello everyone, and welcome to our webinar, Five AI Predictions Every Infrastructure Leader Needs to Know in 2026. I'm Erin Stevens, senior product marketing manager for AI here at Pure Storage, and I really am so excited to be here with you today. I'm also thrilled to be joined by my friend Ruchika Khurana from NVIDIA.
So she is a product manager for NVIDIA's AI data platform, which if you haven't heard about, don't worry, we'll talk all about it today. And she oversees work helping data storage partners like Pure plug into NVIDIA's ecosystem in a way that better enables you, our shared customers, to really bring your AI projects to life in a, a way that is, you know, more efficient, easier, so that we
can help make you successful. She and I met at a conference last year. We had a wonderful conversation. I know you all are going to really enjoy, insights as well, so can't wait to share that with all of you. Thank you so much, Ruchika, for joining.
Hi, everyone. Awesome. So, so let's talk a little bit about the webinar today. You know, twenty twenty-six is shaping up to be a really pivotal year for AI for, for a few reasons, and, and we'll get into this. But, you know, we're kind of starting to get past the hype phase.
We've seen what is possible with AI. There's a lot of really cutting-edge stuff, and there's some, some organizations who've really taken AI to do some amazing things. Now we're moving from that phase of kind of early adoption and, and proof of concept into that, like, major execution phase.
And, as you know in IT, that's where things get really hard, right? Bringing it to life in a way that's, that's sustainable, that's scalable, that's cost-effective. So, like, we really wanna dr- dive into some of those trends because we think this is really the year that infrastructure leaders need to kind of be ready for the, for the
onslaught of demand there. But at the same time, that also puts you in the driver's seat, and we wanna, we wanna help prepare you to be really proactive, in that journey as well. So, so, over the next, forty-five minutes, we're going to be sharing, five AI predictions based on research from Gartner, IDC, also from our wonderful guest,
Ruchika from NVIDIA. We'll be hearing, hearing her predictions as well. And then at the end, we'll, we'll take some questions. So if you have questions throughout, the webinar, please do put those in the Q&A section, and we'll get to as many as on, on at the end, live, or we will,
answer those via the chat as well, just based on how much time we have. So what do you think, Ruchika? Are we ready to dive in? Let's do it. Let's do it. Okay. So our first prediction here is that this is the year that agentic AI
goes mainstream. So I have this great stat from Gartner, and, and Ruchika, I'll ask you to comment in just a moment. But according to Gartner, forty percent of enterprise applications will be integrated with task-specific AI agents by the end of twenty twenty-six, up from less than five percent today.
And so this was a, this was a stat that kind of came out last year, really predicting what this year's going to look like. That's a huge shift from five percent. That's really experimental, right? That's kind of leading edge to forty percent.
That's almost half of enterprise applications having AI agents in them or form. So really again, moving from that experimental to, to operational phase. So Ruchika, what does this shift to more widespread adoption of agentic AI mean for infrastructure teams?
That's a great question and a great So this is my perspective. I think we're nearly there with agents, and I imagine a world where there's everywhere, just like the internet, just like the Wi-Fi, always on, and you don't, don't really pay attention to them.
But if they get slow or start misbehaving, it'll be like the world is ending. That's what I think is about to happen. But-- And what that really means is the world's gonna change for instruct- infrastructure teams. They're no longer just supporting apps.
They're supporting always-on digital workers. Agents don't just answer one question. They're just They're not a chatbot. They plan, they call tools, they reason, they retrieve data, and they go over this over and over again till they actually finish the job.
So infra teams start caring more about a few newish things. For example, I think latency and consistency are gonna become super critical agents feel slow, they could lead to a world crisis. Data proximity is super critical too because agents are constantly pulling from enterprise knowledge.
And governance becomes even more critical because agents touch WEKA more than, more, more than what a chatbot ever did, right? So you can only imagine, like, if governance, falls apart, you can kind of imagine the kind of mayhem that might happen, right? So, it changes all the compute patterns.
For example, you'll likely see WEKA concurrency inference and a lot more focus on optimizations, things like caching, disaggregated prefill and decode to keep things super, super responsive. And all of this requires infra teams to kind of plan for accelerated compute and networking to manage all of this inference all of the networking traffic that goes
with it. So in-- I believe that overall, agentic AI turns infrastructure into more of an AI operations platform with a goal just like keeping the apps, keeping the lights on-Just kind of tracking uptime, but really delivering fast, trusted, and really cost-efficient intelligence, and at scale. Yeah. Yeah. That's, that's such great advice on, on
the things to, to think about and listen to, Ruchika. When you were talking about, IT and infrastructure lead- leaders having to support these always-on digital workers, I mean, that just really That caught me, especially because, you know, if you think about IT and, and IT is always inheriting these, these projects that someone else has built, right?
Well, now there's these, like, autonomous workers, right? These agents that are actually off, you know, making decisions, like you said. They're, they're taking a- actions on their own, and IT is going to be the one being held responsible to make sure that they're performant, to make sure that they're cost effective, to make sure that, they are governed appropriately, right?
So, you know, it's, it's, it's, it's just the, the same kind of thing repeating that IT is used to, but at, at kind of a, a more massive scale with, with, you know, arguably more risk here, too. So, yeah. So thinking about that data proximity, the governance, and then also the, the latency and
consistency and, and those types of things. That's, that's great advice. All right. That leads us to our second prediction. So our second, second prediction here is that data is going to become the make or break for AI.
And this really is kind of getting to the heart of the fact that your data strategy today is really what is going to determine your AI success in the future. So we have this, this stat from, from IBC that predicts that by 2027, companies that do not prioritize high quality AI-ready data will struggle scaling GenAI and agentic solutions, resulting in a 15% productivity loss.
So, you know, obviously that's really compelling, right? You know, that you, you have to really start with the data if you wanna see success with, with AI. So Ruchika, what does a good data AI look like in practice? Yeah. It's a great question, and I hate to break it
to all the leaders, but whoever's been off data strategy all this time, they're have to kind of put things into turbo mode going forward. But here's my opinion. A good data strategy for AI is not about, let's now move this data into this new place that can magically fix everything, and I hope this is this new magic thing, is sustainable
and, will work for the next year at least. It's more like building a solid foundation where your data becomes AI-ready, secure, and easy to evolve. So first things first, let's talk about governance, because AI without governance is basically just automated chaos. Like Erin, you spoke about agents and auton- an, autonomous agents.
You can imagine the kind of chaos that could break loose, right? So in real terms, you really want access control, and audit trails, and not just on the raw documents, but on everything that AI creates from them. So for example, if you're extracting text chunks, embeddings, you're creating embeddings, you're vectorizing, you're creating indexes, all of it, you want the governance to travel
with the things that the AI creates from the documents. For example, if somebody shouldn't be retrieve a document, they shouldn't be retrieve anything that the AI created from those documents. Second, you plan for messy data. Everyone knows that data is never clean.
It's always messy. It's not just about text or tables, it's all of the above. It's text, tables, charts, diagrams, scanned documents, video, audio, and there's gonna be more things that come next year, in the years to come. So your ingestion pipeline should be modular so you can add new modalities
and just keep letting it evolve with time without constantly rebuilding the whole thing. Thirdly, design for change. Language models will change. Embedding models will change. Retrieval, retrieval things will change.
You want a data foundation where you can support the business users that are on your, on your founda- on your platform, but you also have room to evolve. Very importantly, freshness. Freshness matters a lot. AI doesn't just fail when it's wrong.
It fails when it's confidently using outdated information. We all have been a victim of that by now with some of the models that came out, right? So you, you need a platform that is constantly re-indexing, refreshing, doing change detection, the basic keep it current. It has the basic keep it current muscle to be part of the strategy, right?
So to sum it up, a good AI data strategy is a system that keeps knowledge continuously current, governed, secure, flexible, so AI can actually scale in the enterprise. Yeah. Does that make sense? Absolutely. And, and that part about, you know, scaling into the enter- into the enterprise, it really strikes me that this is sort of the year where, you know, we've seen so much of what's
possible with AI in more of the, kind of the public consumer space, right? But we're now starting to bring AI so much more, especially with agents, we're bringing that so much more behind the walls within our own enterprises, and that's where-Things like data freshness, things like data governance, they become so critical, and you also get so much more complexity from all the different types of data sources you're bringing.
Like, you know, I just I think about, you know, setting OpenAI, setting ChatGPT or, or Gemini or, or Claude AI loose on the internet, right? That's a lot of information and data, but they don't have to worry about governance, right? They don't have to worry about those types of things. Now, as we're bringing these large language models behind, you know, the doors of our
enterprises and wanting to then supplement that with data from our enterprises and build agents who have that, that Kontxtual what's happening in your organization and com-complete tasks there, you just have so much more complexity with the data, right? And so much more comp-- you know, you just have to think about governance, you to think about getting the data where it needs to be.
And again, these are all things that IT has had to deal with, right? You know, whether it was for data analytics 10 years ago, you know, or, so or predictive analytics, right? And it's now just even more complex, right? And so, yeah, yeah, that's, that's really good perspective on, on some of the things to, to, keep in mind and plan for.
And, and also what you said, if, if you're, you know, a little behind on your data strategy, yeah, this is the time, right? You know, make sure that, that you're thinking about these things, so yeah. Awesome. Okay. So for our third, prediction, you know, the AI,
data platform is going to become the foundation of the AI factory. For this, we're going to do something a little bit different, and we're actually going to cut to our studio in Santa Clara, to hear from Casey Lai, our VP of AI here at Pure Storage, and also Jacob Lieberman, who Ruchika works with, who leads the AI data platform initiative at NVIDIA.
So let's take a look. Hi, everyone. My name is Casey Lai, Vice President of AI at Pure Storage. Really excited to have in the studio with me today Jacob Lieberman from NVIDIA. Jacob- Hi. Very nice to see you again.
Thank you. Yes. Would you mind telling us a little bit about yourself and what you do at NVIDIA? Sure. Yeah. So my name's Jacob Lieberman, and I'm a director of enterprise product at NVIDIA, a new initiative we have called the AI Data Platform. Very cool. Very exciting.
Well, Jacob, the first thing I wanna talk about is I think everyone when they think AI, they think the coolest co-pilots, right, the biggest, baddest models, and they often forget about data. And so love to kind of talk to you about that. Why do you think data matters for AI? What's your perspective? Well, this is one of my
favorite subjects, Casey. So yeah, there's been this massive rush of enthusiasm around Gen AI and all of its capabilities, but somewhere in the midst of all that, we lost sight of the fact that data is still king. So whether you're training a model, fine-tuning a model, or retrieving addi-additional context through RAG to inform
your LLM generations, you need secure access to high-quality data. Basically, you don't want the garbage in, garbage out problem. Right. Crazy hallucinations, right? Yes, that's still true. All right. Right. Well then, and in addition to that, I
think the performance also matters, right? Because you need to make sure the data gets to the GPUs quickly enough, so that way you don't have idle GPUs. Nobody likes that. Right. So without GPUs, it's really not possible to prepare data for AI/ML at
scale and to keep up with the velocity of the data, the rate that the data changes, and the rate that the data grows. So that's number one. Number two, you know, building these pipelines to make data AI ready are complex, and they have many stages. And there's many handoffs between the different personas and users of the data,
at any moment during any one of those handoffs, somebody could drop the ball. Yeah, I think, I think I see this as one of the biggest challenges that's, getting in the way from an AI inference perspective. You know, up to now, most of the workloads has been training, and so people have been really, really focused on that. But now it's going to shift, where most of the
time, effort, and money is actually all gonna be focused on inference. And so the reason why this is interesting is 'cause what you just said, the minute you get to inference, right, you can only get good inference and good consumption if the data's actually AI-ready. Well, there's many challenges. I mean, first of all, data,
enterprise data is unstructured data. Ninety percent of the data that an enterprise acquires is unstructured data in nature, and there's many modalities, video, audio, text, PDFs with graphics, images, presentations, spreadsheets. Combine those things, it becomes quite challenging to really extract
insight from the data. Right. Well, if you can't even get the data to be AI-ready, you're not getting any insights, right? Right. I think that's, that's definitely key. And so I think that's why, like, we're very excited in terms of what
we're doing at Pure Storage. We announced at GTC last week the introduction of a new product called Pure Storage Data Stream, where we are specifically focused on this challenge. So first part that Data Stream's gonna do is address that specific area, so that way you get one workflow, one product that's going to automate the entire process-
to actually generate data sets for AI/ML in minutes. Second thing we're gonna do is we're gonna make sure it's super easy to consume the output, put some governance around it.... Right? What they should use, what they should not use, who can see, who cannot see.
Those type of things should be in there. And then third is we're entering an age where we have agents, right? Agents are part of our digital workforce, so you have to now think about how agents are going to consume. I think these are very, very important capabilities in Data Stream, right, to
accelerate and simplify the process of making data AI ready, right? And so you can think about Pure Storage as really taking an active role to, to be there for the, for the customer for every step of their AI/ML journey. And what I love about it is that it's all centered around the data, which is really the core competency of Pure Storage is protecting that data. Yeah.
And then it builds on top of that, but the data's always at the core. Jacob, it was a pleasure having you in the studio. Yeah, thank you so much. Thank you to Deluxe Water for, for doing this. Thank you. And we really thank you for doing it. Had a blast. Yeah, me too.
Awesome. Thank you Casey and Jacob. Some, some great insights there. Ruchika, I love the way that Jacob put that at the, at the, kind of towards the beginning that data is still king, right? And you live in this world every day building the AI data platform.
What are you seeing with the customers that you're working with who are trying to get their data AI ready? Yeah, absolutely, Erin. So totally and honestly, getting data AI ready is like meal prep. We both know, like everyone loves the idea of a really good meal- Mm-hmm but nobody wants to
do the prep. But doing the prep is the reason why everything is better for the next week or the next few days, right? So here's what I'm seeing with a lot of customers. A lot of people start thinking that, hey, it's just upload documents and chat, and then maybe I can hook up an agent to it.
And then they realize AI ready really means building an actual pipeline. It includes things like OCR, parsing, building out tables, layouts, generating metadata, chunking, embedding. Have I said enough words? But it's Well, you get the gist.
It's, it's probably a lot more words and phrases that I use that they have to do. I didn't even mention the operationalization of it. But doing all of it in a reliable, manner, not just a bunch of scripts, is a lot of work. And then the enterprise reality hits.
It has to work across millions of files. It worked great for a POC where I had a little script that curated stuff, but it now, now it has to work for millions of files and, jobs have to finish, and these curation have to finish in my lifetime, right? So, that's when, that's when like rubber meets, meets the road really.
And that becomes, that means like ingestion, retry logic, monitoring, cost control, all that, plus keeping the knowledge base fresh as data changes. But the piece that really surprises a lot of teams is that after they've done all of that, they now need measurement. They have to define evaluation sets, run regression testing, benchmark extraction,
extract, quality tracking, tracking, quality, tracking end-to-end generation, answer generation quality, impact to the agents, track performance benchmarking. How does this work under load? How does it, how does this work when there's a gazillion agents hitting this backend system? Tracking latency and throughput, and making sure it's predictable and consistent.
That should give you an idea how much work that is, and it's hard. So I think, I think enterprises are realizing that making data AI ready one-time project, it's basically a whole engineering practice. And they need to do this r- irrespective of what the agents are doing. It's, it's a foundational piece of their, of, of their agent building practice, if you will.
Mm-hmm, yeah, absolutely. You know, I've heard you use the term pipeline a lot, and, you know, I think five years ago it was, and 10 years ago it was all about data lakes, right? And before that, you know, I used to think of just databases. They're kind of this, this stagnant- landing, you know, landing place, right?
You know, you like it goes in, maybe it comes out, but those are very distinct from one another and, and there's a huge time, timeline there. And then a data lake is kind of, you know, kind of I, I always picture this sort of black pool where all the data's going, but then how do you get it out, right? And what I, what I'm struck by is now we sort of need to take these giant lakes of data that
we've created- Yeah and now we have to do everything that you just mentioned to build a pipeline of that data into the AI models, and it does. It has to be fed back at times. It has to, you know, it has to have the right data at the right time to keep things moving fast.
But I love that visual of thinking about, thinking about data for AI as more like a pipeline running through an AI factory, for example, right? It's a really different construct for, for how you- Mm-hmm wanna think about delivering data. It's not about where you land it, it's about how you take what's there and deliver it quickly, right, to the models. Yeah.
Yeah. So, you know, just as a follow-up to that, why are storage and data management so critical to success with AI/ML in, in this, in this construct? Yeah. Absolutely. But I just wanna call out, Erin, I think you're being extremely kind by referring to a
data lake as a data lake. A lot of people I know actually refer to it as a data swamp. Um- I haven't heard that, actually. That's so great. Yeah. But, you're absolutely spot on.
I mean, storage and data management is absolutely critical because without it-You can't really succeed with AI, with, with AI, and, Casey actually referred to it well. He, he talked about it like he referred to it as garbage in, garbage out. Mm-hmm. But, here's, in my view, and not to make it look like, oh, something else is easier, or okay, they got the easy problem, I'm stuck
with the hard problem. But I seriously believe that a- w- when it comes to agents, the model part is maybe the easier part. And the hard part is actually trying to figure out and make sure that the agent is getting the right information, and the truth is told to the agent so the agent can kind work reliably, reason through properly, and, execute properly and reliably.
So, and, and I truly believe that agents are only as smart as a retrieval layer. They, don't just answer questions. They're constantly doing things like retrieve, retrieve the data first, tools, take actions, and again retrieve the data. So where I'm going with it is that the data layer, h- is absolutely critical
because it becomes this agent's, muscle memory, because the agent's retrieval depends on this data layer. So if the data is messy, stale, or scattered across a gazillion silos, the agent just turns into that coworker who's super confident but is still wrong. We all have know We all know at least one person, like, who has those attributes, super
confident but gets it wrong, but leadership loves thos- loves that person. Anyways, not to digress. That's why data management is so critical. K- keeping content organized, up-to-date, searchable, and consistent with change detection, refresh cycles, indexing, all of the associated metadata management, and all of
the lifecycle policies. It's all the non-blingy stuff that is absolutely essential to the agents, aspiring to do a good job. So, yep, storage isn't just about where data sits anymore. It's basically the supply chain for intelligence.
If it's slow or outdated, your agents just can't function, period. Mm-hmm. Storage is the supply chain for That is great. That is, that is so good. And I think, you know, especially for infrastructure leaders out there, it's really important that you help teach others in the business about how critical having the right
infrastructure to do these things is, because I, I couldn't agree more. The supply chain for intelligence is, such a good way to, to think about it. I also feel like we have discovered a resolution to go with this prediction, Ruchika. 2026 is the year that we're gonna clean up our data swamps. So maybe we can resolve all together on this webinar to, to do that.
Yeah, I, I just, I, I love that. I love that. All right. So prediction number four is that we are going to see a new wave of infrastructure modernization. And I think, you know, when we, when we think about this, you know,
infrastructure modernization as a term came up, yeah, again, probably about we started to think about cloud-native, think about moving to the cloud, think about, containers and microservices, all that, and, and we've been talking about it for 10 years, right? And I think, you know, we've realized it's just kind of a continual thing, right?
You're always modernizing your infrastructure to get on the latest technology. Mm-hmm. But I do think we're about to see this major gestalt shift again, and think, start to think about a fundamental change in how we think about architecting infrastructure to support AI versus, to support applications, right? And then this is something that we've heard a lot from you all at NVIDIA, and, you know, I
think we're all kind of following your lead here. But I have this quote, from, Jensen Huang, of course the CEO of NVIDIA, from his keynote at CES, which, is that, "Computing has been fundamentally reshaped a- as a result of accelerated computing as a result of artificial intelligence. What that means is some 10 trillion or so of the last decade of computing is now being
modernized to this new way of doing computing." And what I hear from that is, okay, everything that we have been modernizing and everything that we've been moving forward, it's time to, to modernize again, right? You know, it's this, it's this new wave. It's an enormous amount of infrastructure to really think about modernizing.
So Ruchika, how is this wave, in your opinion, different or similar to, say, of the cloud? That's, that's a really interesting question, Erin. So I think it's similar to the cloud in the sense that it's a massive platform shift, but what's different is we're actually dealing with two shifts at the same time, and this is,
this is how I'm thinking about it. The cloud era was mostly about where computing happens- Yes and a shift from the cloud with promise of additional elasticity and cost efficiencies. Yeah. But this wave is fundamentally about what computing even is, because AI is pushing us from tr- traditional CPU-style computing to
accelerated computing as the default. Like Jacob rightfully said in the video, you Some things are not even possible without accelerated computing. So it's not just migrating workloads, it's modernizing the fundamental stack, right? And that's the big shift.
It's not just one product you install. It's an end-to-end system, data pipelines feeding models, models powering agents, agents taking actions, and the infrastructure has to keep the whole loop running continuously at scale with predictable performance.Because I s- and remember I said when p- when you don't get the With these digital workers, when the performance isn't good, it's gonna start
seeming like world crisis. Yeah, yeah. Even the op- economics are gonna change. You're gonna still track the usual storage networking compute, but you're also gonna start tracking inference as a first-class line item, with costs broken down by, like, different kinds of tokens, input, output, context tokens, retrieval tokens, et cetera.
But basically, your infra bill is gonna start looking like a token ledger, if Mm-hmm. Yeah, yeah. Absolutely. And I it, it's, it's interesting because I think as you talk about this not just a shift in, in where your applications running, right? You know, it's, it's more fundamental than
that, right? It's a shift in, in how we com- how we do compute, right? And, and, how we think about architecturing, you know, the whole, the whole stack to, to better, optimize for that token performance. Again, it strikes me that in There was, there were these predictions in the early days of cloud that hardware was going to become a
commodity, right? Hardware was, you know, it Everything was gonna be on the cloud. You just think about compute and storage or, you know, the services that you were buying. And yeah, a lot of that has, has been the case. But at the same time, I think NVIDIA kind of made hardware cool again, and really brought a
lot of relevance. You know, we spend a lot of time now really, really talking through and thinking through, exactly how, how compute is done to optimize for, for AI/ML. And I find that really fascinating, honestly. And I do think that that's another opportunity for infrastructure leaders
that, that, hardware is not a commodity, right? It's a fundamental building block of your AI factory. And the success that you're going to have, with that factory and the, your ability to op- optimize, your tokens and your output, that is very much dependent on the hardware that, that you're using. And I think that is where, you know, we have
some, you know, we have some, some, shared principles in the way that we are, you know, designing the hardware that we provide to customers, right? From, from our inception at Pure Storage, we have been all about, rethinking the way that storage hardware is actually written, right? At the, at the NAND level.
And, so I think it's, it's really cool that we get to kind of, dive back into aspects that I, I find really, really fascinating. Awesome. Okay. Are we ready for our fifth prediction here? Shall we go to it? Let's do it.
All right. This one is all for you, our audience. Our final prediction here is that the infrastructure leader is really going to become the enabler of AI, and this is the year for that, right? So what, what I have heard a lot from customers and what I have heard a lot from, folks who are building AI is that a lot of what has been done so far has been proof of
concept or experimentation, at least within the enterprise, right? We've seen a ton of movement on the large language models being put out there by, you know, by the big leaders, but we are still very early on in bringing that into the enterprise. There's a lot of really promising work that's been done to prove the value that AI can bring
to the enterprise, right? With AI agents, with physical AI, with so much more. And but, but once you wanna take that from pilot or proof of concept into production, well, that's where everyone is about to turn to you, on the infrastructure side, right? And this is a moment where, you know, you can
either be kind of dragged along or you can kind of or, or have to come in and clean up maybe a mess later where someone tried to, to build something in a way that wasn't scalable, or wasn't secure and governed, or you can proactively go to other leaders within the business and understand what they are trying to do with AI/ML and then give advisement on how to bring that to life in a way that is scalable and secure and efficient, right?
And you know, a- again, I, I think this, this moment, reminds me of the move to containers. It reminds me of, of, the introduction of cloud, right? These major moments of, of, technology transformation. And I think this is such a great moment where in IT and in infrastructure you get to go and
essentially say, "Hey, look, we're not a cost center. We're not a commodity. We are the enabler of AI in the enterprise," right? We are the ones who are going to make this actually work in practice, and that comes with a lot of complexity, so trust us and come to us, and we can help enable that success for
you, and we can do it in a way that is, that is scalable, is efficient, is secure, and that is going to lay the groundwork to continue to do this faster and better in the future," right? If, if you partner with the business. So, 2026, this is your moment. So you know, we really encourage you, to, to really, you know, spend some time, especially
as the year is starting out in these first few months of the year, spend some time educating yourself. You know, take, take a beat, take a pause. Now is the time. This is new for everyone, right? This is new for everyone in IT, and so you have that opportunity to go and learn and lead
with this. And those are really unique moments, right? So, and, and we'll be here. We'll be bringing NVIDIA back and, trying to, to give you as much, in terms of resources and information as, as we can and, you know, to, to help guide you through this.All right. So, we'll take some live Q&A here in a few minutes.
But before that, Ruchika, I wanted to kind of, switch to, to more of a, a fun, fun personal question, if that's okay. Any New Year's resolutions that you have made for 2026? Oh my gosh. Yes. I need some of the GPU acceleration goodness in my personal life.
AI's moving at warp speed, and I'm still running on chai without milk. So I wanna learn some white coding ASAP, code up a couple of agents to kind of automate some of the busy work, and I hope to spend a bit more time this year on advocating for sheltered animals. Oh my gosh, I love that.
I need some of that, too. So, you know, teach, teach me everything you learn. And I can steal it for my life. But I also love that, and I think, again, this is sort of the dream, right? Technology is here to enable us to, to automate the busy work so we can spend time on
what really matters, right? I just I love I know we talked about, talked about your, your passion for, and, and I love that, that you're going to be able to free up some of that time with AI to, to spend more time taking care of, care of animals. That's fantastic. Awesome. Okay.
So I do see some questions coming in, and we have time to, to go through those, couple of those. So, let's dive in. So our first question here, Ruchika, I'll direct this to you. Mm-hmm. We are seeing a lot of AI initiatives stall out in the pilot phase.
What is the biggest blocker that you see? Is it technology, is it data, or something else like organizational readiness? Or, you know, sort of help us out. Like, how do, how do we speed this up? How do, how do we remove these blockers?
Yeah. Quite honestly, I don't think it's the technology. In my mind, it's the data and organizational readiness. I see a whole bunch of pilots being super successful because they're small, managed man- manually with a curated data set, but they absolutely have no way to go to prod.
There's no path to go to prod because in order to do these pilots, teams, take, like, curated data sets from their fully governed sources, get, do a dazzling demo. But putting it into prod is so difficult because all of a sudden they're trying to put Band-Aids of, like, monitoring, et cetera, and it just doesn't work out. So I think it's the overall data and organizational readiness which is blocking a
lot of these pilots from Everpure getting, advanced to production-ready, like, agents or workloads. Yeah. Yeah, absolutely. Yeah, real world data is something different, right? And anyone who's ever worked with data knows that, whether it's building an application or
data analytics, all that. Mm-hmm. You know, you can do a beautiful demo with, with perfectly curated data, but then you get real world data in and it's a totally different thing. And, yeah. So, so again, you know, going back to, back to those principles of cleaning up
your, your data swamp and building a pipeline, right? You know, taking what is in, in your organization, what's in your enterprise, and actually building a pipeline to feed your AI models is, is really critical. All right. We've got another question here. For organizations that are just getting started with AI infrastructure, where should
they focus first? Upgrading storage, investing in GPUs, or cleaning up their data? Yeah. I think they absolutely need to start with the data. Data is pretty much core to the success, as we kind of spent a lot of time on.
Data hygiene and work on your Make sure your access controls are solid, because any little piece of data that lands up in the wrong hands or with the wrong agent you up to lawsuits. Don't do that. Get your data, data in order. Get your access controls in order.
Next part I would suggest is work on, data management. Make sure you have, the right pipelines and the workflows to ingest all of these multiple, multiple modality, data that might exist. And once you're ready with that, make sure you have the GPUs ready at hand the data AI-ready and abs- and scale.
Because part of making your Part of making your data AI-ready, means making sure that you have scalable and consistent, and performing pipelines. So, chances are that given the speed at which technology is moving, to avoid behind, you're probably have to gonna do all of those things together if you're starting from there. Mm-hmm.
Yeah. And, you know, I, I feel like this is a great moment to point out that there are, like, a ton of partners out there who, you know, shared partners with us and NVIDIA who- Mm-hmm have done this. You know, you, you don't have to reinvent this and go it alone, right? We have experts, here at Pure Storage.
Obviously, tons of experts at NVIDIA. Yep. And we have partner organizations that we work with that, you know, again, you don't have to start from scratch and do it yourself. In fact, we don't recommend that. So, you know, definitely an opportunity, to do that as well.
Okay, this is actually a nice segue into what I think will be our last question. So how do you balance the pressure to move fast on AI with the need to do it responsibly, especially around data governance and security? So, the, success I have seen with teams is in situations where I've seen teams actually, work with the guardrails in place from the get-go.
So start small, keep access locked down, but make sure that governance principles are being followed end to end.Uh, not just for the raw- for the raw docs, but all of the assets associated with the raw docs, like the charms, embeddings, indexes, et cetera. Basically, everything that the AI actually touches. But start small without skipping important steps, like governance.
Start small with that so that you can absolutely ship quickly. But, And don't do this thing where you say, "Okay, I'll add security later." Mm-hmm. So just don't have it as an afterthought. Have these, major principles in your pilots and in your POCs from the get-go. Yeah. That's my- And, and again Yeah.
Yes. 1,000%. And I think again, this is another, you know, moment where infrastructure leaders have, have the opportunity to, to lead and partner, you know. Because you know how to do security, you can bring in your security owners, right, and your security partners, into the beginning discussions, right?
These are things that you should be discussing before, before you ever anything, right, is, is how, how security and governance are, are part of, what you're doing. So fantastic. Okay. So, in closing, I have one final question for you, Ruchika. What is the one thing every infrastructure leader watching this today should take away?
I think, I think the primary thing is don't think of AI as just another app you're hosting. Mm-hmm. It's a brand-new operating model, and I think the leaders who win will treat this like building a platform. AI-ready data with accurate compute with guardrails so agents can reli- su- can
reliably, scale and be consistent is, is absolutely critical. And you're not alone. You have some amazing partners like, like Pure and NVIDIA to kind of help you in your journey. Fantastic. I love that. And I, I couldn't agree more, Ruchika.
And, you know, for, for everyone out there, you know, the decisions that are, are going to set you up for, for and, and be the defining, defining point in whether or not your organization leads or, or lags in the future, right? And not just with AI, but that But whether you are leading with AI is whether or not your organization is going to be leading or, or lagging.
So, be bold, be assertive this year. Find opportunities to go and evangelize to the business why IT and infrastructure leaders are a critical part of AI success and need to have a seat at the table, and, and, you know, are critical to, all the conversations about bringing AI to life within, within the enterprise. So thank you everyone so much for joining us.
Thank you for all the wonderful questions. Hopefully we've answered all of them either live or in the chat. And, I'll spend a, a few minutes afterwards, you know, looking at, at, at the last few and, and answering any we didn't get to. But if you are interested in continuing this conversation or learning
more, please visit pure.ai. Or if, if you'd like, you can also join us on the community. I'm active on there. We have a number of other AI experts from Pure Storage active on there. So we'd love to see you there.
Thank you so much again for joining us for the conversation today. We, we can't wait to see you for the next one. And thank you, Ruchika. Thank you so much for, for being with Thank you. Thanks, everyone.
Ruchika Kharwar
Product Manager (NVIDIA RAG Blueprint | AI Data Platform), NVIDIA
Erin Stevens
Senior AI Product Marketing Manager, Everpure
In 2026, AI moves from experimentation to execution—and infrastructure leaders are in the driver's seat. Organizations that build the right data foundations now will be the ones turning AI into a true competitive advantage. Join us as we explore five trends shaping the future of AI infrastructure, from unlocking the value of your data to architecting for production-scale AI workloads. Walk away with a clear roadmap for positioning your infrastructure as the engine that powers your organization's AI ambitions.
In this webinar, you'll learn:
- How to get your data ready for AI
- All about the NVIDIA AI Factory and NVIDIA AI Data Platform
- What it takes to scale AI from pilot to production
- Practical ways to bring AI into your organization
We Also Recommend...
Personalize for Me
My Updates
No updates at this time.