14:34 Webinar
AI Infrastructure and What Comes Next
Hear how Everpure is approaching AI infrastructure, readiness, and the operational demands of scale.
This webinar first aired on 17 June 2026
Click to View Transcript
Thank you all for coming down to Las Vegas. Let me tell you what I've learned about the people in this room. Over the last year and a half or two years, you've all been asked to bring AI to life in your organization by your board, by your business leaders, by your CEO. And as Charlie said, the only way to get this right is by all of us
shifting our focus to data. Because the bottleneck that's stalling AI is not compute, it's not models, it's not tooling, it's data. Some of the latest research drives this point home. Earlier this year, we partnered with IDC and surveyed over thirteen hundred IT leaders, but
here's the number that stopped me cold. Eighty-six percent say storage is holding AI back. Eighty-six percent in a world where every board is asking about AI and billions of dollars are being invested. Why? Well, over sixty percent say their data
infrastructure needs improvement or a refresh. They say their data platform isn't connected enough or rich enough in context. Read these stats together, and the conclusion is unavoidable. Data is at the heart of AI success. But there's three things you need to make data work for AI.
First, you need AI-ready data, data that's refined and ready to be served up to AI models. Number two, you need AI-ready infrastructure. This is what Chad just showed you, a unified and intelligent data platform that not only stores but understands, governs, and protects your data.
And finally, you need ecosystem integration with partners like NVIDIA. Here at Everpure, our platform delivers all three. If you already run Everpure for your tier one workloads, your databases, your file systems, your object stores, the same platform covers every phase of your AI journey, from connecting and preparing data to training and inference.
No new systems, no new teams, no new silos. Much of the focus over the last few years has been on training and inference, and the Everpure platform is ready for both. Today, we help you train models which require massive throughput because slow data can hamper innovation. We built FlashBlade EXA to deliver over ten
terabytes per second of read performance. It's been benchmarked feeding more than ten thousand GPUs without a single idle cycle, achieving top results in both MLPerf and SpecStorage. But we don't stop there. We also help AI infer because users won't wait for a slow answer, and agents don't take coffee breaks.
Inference requires low latency retrieval and high concurrency. FlashBlade S delivers two hundred and twenty gigabytes per second and four hundred and fifty million IOPS. FlashBlade EXA handles four point six billion metadata operations per second, and KV Cache accelerates inference response by twenty X.
The output of all of this agents and agentic build-out is happening on containers, so it gets even better because Portworx is the engine for containers, making endpoints elastic, scaling automatically under load with sub-minute failover. Many platforms can handle training and inference, but they skip a critical step, getting your data ready for AI.
And here's how Everpure is different. We help you connect and prepare your data. Other vendors want you to copy all your data into their system. Some optimize for training alone. Some are retrofitting backup infrastructure for AI.
Every one of them solves a single slice, creates a new silo, and asks you to move your data to do it, which means AI is always working on a copy. And guess what? A copy is always behind. Instead, as Chad highlighted, Everpure data intelligence brings AI to your data where it already lives on primary systems in real time.
No copies, no silos, no stale answers. And our new product that we're unveiling today, Everpure Data Stream, helps you prepare that data, automating pipelines from ingestion to inference to deliver results faster. Classify, curate, index, vectorize, feed your AI factory. Today, this step can take skilled teams of data engineers months of work.
We believe it can be automated in minutes. More on that in just a few. Finally, your data platform must be optimized to feed your NVIDIA AI factory, so you can continuously power innovation and create intelligence with AI. Over the last few years, we've achieved NVIDIA storage certification across all performance
levels with validated reference architectures, including Enterprise, SuperPOD, and NVIDIA Cloud Partner. Together, the Everpure data platform is the foundation that turns your data into a competitive advantage, faster delivery, real-time decisions, and AI agents you can trust. Now's the really fun part where we get to show you something new.
To help me unveil, I'm thrilled to welcome Kevin Deierling, SVP of Networking NVIDIA, to the stage. Kevin, thank you for joining me and some of my closest friends here today. Great to be here. Let's just start Let, let's get down to brass tacks. When NVIDIA designed the AI data platform,
what problem were you trying to solve? Yeah. So enterprises have decades of knowledge sitting in documents, files, and databases, but their AI can't reason over it. Data was never connected across all these systems of record. Every application understood its own slice, but nothing understood all of the data,
the whole. So the AI data platform is our reference architecture to fix that, and it brings NVIDIA's accelerated compute and AI software stack directly to the data where it already lives. So agents can query enterprise knowledge in real time. And that's what we're here to unveil, our new service based on NVIDIA's
design, Everpure DataStream. That's right. NVIDIA and Everpure have spent the last year as extreme co-design partners across the entire stack building this solution together, and we couldn't be more excited to show it to you. I hear from customers all the time that preparing data can take a skilled team of data
engineers months of work. That's just not sustainable. We believe that it should be automated and do this in minutes. So NVIDIA and Everpure worked together to solve this challenge. Yeah, it's exciting. Our new product, Everpure DataStream, delivers
just that, allowing you to automate pipelines from ingestion to inference and deliver results faster. To walk us through the live demo, it's my pleasure to introduce Per Botts, Vice President of AI Infrastructure. All right. Kevin, you ready to show everyone what we've been up to? Let's do it.
All right. So, so to, start off, I wanna show you what you can do with this new product we built, DataStream. It's powered by FlashBlade, and it's built on NVIDIA GPUs. And the whole goal is to make every one of your arrays AI-ready. And what I'm gonna do, I'm gonna show you an agent, an agent that we built on DataStream,
and it's used for traffic court. It's the place where you don't really want to have hallucinations in AI. No. We don't wanna get anybody in trouble at traffic court. All right, so let's get going and showing off what we built here. So DataStreams ingest and curates data.
It does that to create clean AI-ready data. And clean AI-ready data make it easier for an AI model to find the correct data without any hallucinations or duplications. And you know what? We're using NVIDIA NIMs to curate the data. Oh, that's great because NVIDIA Inference Microservices are containerized,
packaged to be able to make building applications easy. Exactly. Exactly. Now, once we have curated a DataStream, we index and vectorize it to make it ready for inference. And what this does, it enables the court, enables the court clerk to semantically search
the legal assist stream for similar cases that's been before the court. And here you can see how an AI found similar courts to cases has been before the court to determine if the court follows established guidelines. This is great. It really lets you find things instead of just search for them. We finally have find.
Exactly. Exactly. Semantically find. Now, sometimes you really need to have even more advanced AI. And, and so what we did, we added support for connecting DataStreams to other modules and agentic workflows. In fact, this legal assistant needed to do more than search and find similar cases.
It needed reason about the data. So Kevin, we're real excited that we have developed the capability to connect NVIDIA's absolutely most advanced reasoning models to DataStream. And this is what we're showing here. We're predicting how this case will progress through the court based on all other court
cases visible in that DataStream. It's a great way that you've shown an application that's leveraging all of the infrastructure and the models that we've built, and I think you can build applications like this across many different disciplines. Oh, absolutely. Anything that's information-driven, you can
build these types of applications on. Now, you do wanna see the actual DataStream product, and this is how the product looks like. This is the user interface we developed. And our goal with DataStream was to take regular enterprise file and object data and make it AI-ready with built-in governance.
And so here, let's pick on the legal stream, where we define the source, the classification, and the creation governance rules for everything that's been ingested to it. The court cases came to DataStream as files from a file share, and the ingest policy for DataStream allows you to pick which data types and what update frequency you should use. I love this 'cause you're able to take the data from many different places and enrich it
and make it AI data-ready. It's, it's so true. I mean, this, this could come from any array, and we suck it in into DataStream and make it ready for AI. Fantastic. Now, I saved the best for last. A DataStream doesn't just store your data, it lets you talk to it.
We ship several LLMs right inside, so you can chat with the data, and you can uncover insights that go so far beyond search. DataStream delivers the complete stack, data management, curation, classification, and LLMs running on your data securely and in your data center. Every one of the hard problems has been sol-solved easily with just a few clicks.
And we developed Legal Assist in this for the-- on top of DataStream literally in under an hour. It's that fast to develop. And finally, let me share one more piece of good news. Thomas Wagner that we illustrated the Legal Assist with, he doesn't reall- he don't, he never really got a ticket. We generated him with AI.
Oh, good. Thank goodness. All right. Now, we're about to wrap up our demo, and I really wanna thank you for our, for the partnership. It's been fantastic to work on solving this together with NVIDIA. Extreme co-design is something special.
Yeah. And I really like-- I saw some app calls in there. Do you have APIs that are defined so that people can build these applications easier, other applications on top of, Data Stream? You, you're absolutely right. I mean, the reason why apps are so fast to
develop, on Data Stream is 'cause we built really strong APIs, standardized APIs. We validate against open source packages, bring your own agent, use ours, build new, build your own or, or, or deploy open source. The choice is truly yours. It's, it's amazing platform to build stuff on.
And on a personal note, I mean, the engineers, we really wanna thank NVIDIA. This, this has been a great collaboration. Thank you so much for the partnership, Kevin. Yeah. It's been great working with Everpure. Thank you. Yeah.
Thank you. Now, we're gonna show you more details at the breakout session. Please come and join the Data Stream breakout session. Sean, over to you. Thank you, Kevin. Thank you, Par. Pretty good stuff, guys.
What do you think? And just to be crystal clear, Everpure Data Stream is available today. Folks, this isn't theory. More than fourteen thousand five hundred customers run on Everpure, across financial services, healthcare, manufacturing, and some of the largest AI clouds in the world.
These customers are already running AI factories at scale on Everpure. So as we wrap up, here's what I want you to remember. For the last thirty years, every system maintained its own truth, its own silo, its own context, and you all paid for that fragmentation every single day. That era ends now.
Applications used to define the enterprise. Data now defines the enterprise, and the organizations that can operationalize, govern, and understand their data in real time will define what comes next for AI. We at Everpure built this platform for this moment. The question every organization in this room needs to ask is this: Is our data platform
ready for what's coming next? At Everpure, we're confident we have built it, and we are just getting started. Thank you.
Shawn Rosemarin, VP, Customer Engineering, walks through how Everpure is approaching infrastructure, readiness, and operational considerations that organisations face as they scale AI.
Personalize for Me
My Updates
No updates at this time.