NVIDIA GTC 2026 | DD Dasgupta, Equinix

DD Dasgupta, Equinix

In this interview from the Nvidia GTC AI Conference and Expo, Rev Lebaredian, vice president of Omniverse and simulation technology at NVIDIA, joins Varun Parmar, general manager at Adobe, to talk with theCUBE's John Furrier about how 3D digital twins are bridging the gap between product design and marketing content production. The pair detail a new joint solution that converts CAD files into pixel-perfect digital twins, enabling enterprises to carry a single source of truth from engineering through campaign activation. Parmar highlights that brands expect content demand to grow 5x over the next two years, making identity preservation critical where generative AI alone risks hallucination. Lebaredian explains why simulation-based rendering of physical products has been NVIDIA's own practice for over a decade, producing keynote visuals long before chips leave the fabrication line. The conversation also explores the dramatic compression of campaign timelines, with asset production shrinking from roughly 30 days to minutes when digital twins replace traditional photography and set design. In the automotive industry, Lebaredian notes, marketing teams can gain a year's head start by working from finalized CAD data while vehicles are still in manufacturing. Parmar unpacks how Adobe's Firefly Creative Production platform layers brand governance, content analytics and agentic orchestration on top of the digital twin pipeline, freeing creative professionals from the 60% of their time currently spent on repetitive resizing and reformatting. He also underscores the risk of a "sea of sameness" as generative models commoditize basic content, arguing that agent-assisted workflows elevate human creatives to focus on the big ideas that differentiate a brand. From the Alliance of Open USD standard unifying cross-departmental pipelines to a conversational interface that lets any marketer generate on-brand assets in plain language, the discussion provides a practical blueprint for turning physical-world products into scalable, activation-ready content at machine speed.

Share this session

Clips
More from NVIDIA GTC 2026

DD Dasgupta

VP of Product Management, Cloud and Compute

Cisco

play_circle_outline Distributed Multi-Agent, Multi-Model AI: Enabling Specialist Models Across Cloud, Edge and Private Data Centers Without Vendor Lock-In

play_circle_outline Data gravity: move models to data; inferencing at the edge.

play_circle_outline Equinix Distributed AI Hub: Connecting Global AI Ecosystems with 280+ Data Centers and 10ms Reach to 90%

play_circle_outline Large ecosystem of partners: hyperscalers, neoclouds, security, observability providers.

play_circle_outline Vendor neutrality and architectural flexibility enabled by Equinix's neutral hubs.

Info
Transcript

DD Dasgupta, Equinix

DD Dasgupta

VP of Product Management, Cloud and Compute Cisco

In this Cube Conversation, DD Dasgupta, vice president of product marketing at Equinix, joins theCUBE Research's Bob Laliberte to discuss how the distributed nature of data is reshaping enterprise AI architecture as organizations shift from training to production inferencing at the edge. Dasgupta explains why moving models and intelligence to where data lives — rather than centralizing everything — is both more economical and more effective, especially as industry-specific workloads demand specialized models tailored to financial services, healthcare and reta... Read more

explore Keep Exploring

How is enterprise AI architecture evolving, and what challenges do organizations face when scaling distributed AI workloads across multiple clouds, private data centers, and edge environments? add

Why is distributed AI — moving models to where data is and performing inference at the edge — becoming increasingly important for organizations compared with centralized, cloud-based training? add

How does Equinix’s infrastructure and ecosystem support distributed AI workloads (especially inferencing), and what is meant by a “Distributed AI Hub”? add

Why are customers often stuck with a single cloud/vendor for AI and unable to use models from other clouds or their own homegrown models? add

How does Equinix’s neutrality and its Distributed AI Hub help customers avoid cloud data lock‑in and enable them to mix and match AI models and security solutions across clouds and on‑premises? add

bolt Powered by CUBE AI

DD Dasgupta, Equinix

search

Bob Laliberte

>> Hello and welcome to this discussion on distributed AI infrastructure and the Equinix Distributed AI Hub. I'm Bob Laliberte, principal analyst with theCUBE Research. As organizations move beyond AI experimentation into production and increasingly agentic AI infrastructures, the whole part of the infrastructure itself is really becoming a critical differentiator. AI workloads are no longer confined to just a single cloud or a data center, but instead they now span multiple clouds, multiple private data centers and edge environments. And there's also a rapidly-growing ecosystem of GPU providers and AI platforms. So, that level of distribution creates new challenges around performance, governance, security, and operational complexity. So, to talk about how infrastructure is evolving to address these challenges, I'm joined by DD Dasgupta, VP of product marketing for Equinix. DD, great to have you join us.

DD Dasgupta

>> Bob, thanks for having me, and it's great to be back at theCUBE.

Bob Laliberte

>> Yeah, CUBE alumni, right? So, it's fantastic. To get things started, I wanted to talk about the fact that there's so many organizations now that are moving from these isolated AI projects, and they're turning to these multi-agent, multi-model AI environments. And really, we're seeing that spanning really that distributed environment, right? Multiple different clouds, platforms, and even more importantly, data sources when you think about inferencing and the edge. So, I'd love to get your take. How are you seeing that enterprise AI architecture evolve? What kind of challenges are organizations running into as they try and scale this distributed AI workloads?

DD Dasgupta

>> Yeah. No, it's a great question. There's a lot to unpack there. So, let's start with the multi-agent, multi-mode. I think at the heart of it is the choice and flexibility. We've seen this with any technology. We have the first couple of solutions that folks gravitate to, but then they look for choice and flexibility. They don't want to get locked in. They're thinking about it from the long-term. And so, that's with any application, AI is no different. And so, because of all these choices now, customers are wanting that choice and freedom. But also, there's specific models and from specific technology, specific companies that do one thing really well. Let's say it's financial services, the model they're wanting is very different from, let's say a retail or a healthcare is needing. So, it goes to the choice and flexibility and also the specialization of what that technology can enable. Absolutely paramount. So, that's point number one. The second thing, and at the heart of why we call this distributed AI, is AI is built on data. And guess what? Data is distributed. It's in the clouds, it's at the edge, it's on our watch, on our refrigerators, in our cars. And data tends to be heavy. We know this term data gravity. And so, it's so much easier and more economical to move the model, the technology, the inferencing to the data versus doing the other way around. And so, that's why we're seeing the growth of distributed architectures, distributed intelligence because it's all driven by distributed data.

Bob Laliberte

>> Yeah. No, I couldn't agree with you more. I mean, and the whole drive behind that, there's really that real time intelligence, that real-time insight that organizations want. So, collecting the data at the edge and then shipping it all the way back to a cloud or to a data center, you're not going to have that ability to move as quickly. So, completely agree with you there. We're seeing a lot of that distributed, especially now that we're moving out of just the training idea of AI and moving into that inferencing, especially as things like physical AI and all that take hold, we're really starting to see that evolve and organization look to figure out now how do we accomplish this, right? And how do we access the data that's in those remote locations as well?

DD Dasgupta

>> Yeah. Yeah. 100%. I talk to a lot of clients and the realization is nobody's going to make money on training. Every company is going to make money on inferencing. That's a company's competitive advantage. That's how businesses will grow faster. And again, when 70% to 80%, actually, maybe it's closer to 90% of the data on this planet is being created at the edge, well, that's where you want your technology to reach, not at one centralized location. So, totally agree. That's where inferencing is taken off. That's what companies are interested in. The fun is at the edge.

Bob Laliberte

>> Yeah, absolutely. And so, I wanted to touch upon, you recently had an announcement, you have the Equinix Distributed AI Hub, really a framework for connecting these AI ecosystems. I wonder if you could take a few moments and explain what it is, how it's going to help organizations simplify all that complexity that's being created by trying to connect all of these distributed models, data, and infrastructure.

DD Dasgupta

>> Right. So, turns out it's actually not a new concept for us, it goes back to the origin story of Equinix, where every data center that was ever built was meant to do one thing and really one thing really well. And it was not a place where we wanted different technologies to integrate. It all changed with the internet because you had traffic on different network service providers all wanting to connect into one place. And that was our first hub. It was our first multi-vendor hub, built to have internet traffic come together. And then, over the years, we've gone through the cloud revolution, the mobile revolution, the social. And at every step of the way over the past 27 years, that's what we've done. We've created more than 280 data centers, which are also 280 neutral platform hubs around the globe. And so, now with AI, you're needing new services, but also, an evolution of the existing services, like security, like governance, like sovereignty. So, it's not very different from what we've done, but we're calling it the Distributed AI Hub because this is specifically for the AI workloads, and in particular inferencing. So, that's basically the new piece, which is we're bringing the AI vendors, the neoclouds, of course the hyperscalers, as well as more than 3,000 of different IT services in one place. But as a concept, it's not new for our company.

Bob Laliberte

>> Correct. But that's really one of the true values that you have is that ecosystem that you've created and that infrastructure that you've built out is now tailor-made for a highly-distributed environment. If I remember the stats correctly, 10 milliseconds from 90% of the world's population.

DD Dasgupta

>> You got it.

Bob Laliberte

>> So, no matter where you are in the globe, you've got the ability to connect up and not just to the infrastructure of a compute, but that ecosystem of partners that's available. So, as we all know, there's never going to be any one company that enables you to do everything. So, that ecosystem play has become so much more important. People are recognizing, even though there's so many more things to do, you just see that ecosystem playing such a bigger role in today's environment, I think.

DD Dasgupta

>> 100%.

Bob Laliberte

>> Yeah, absolutely. So, I wanted to touch upon some of the concerns that organizations have. And you brought this up a little bit at the beginning, and that's about lock-in, as they now want to experiment with different models, find specific models for their specific industries and so forth. So, talk to us a little bit more about how the Distributed AI Hub enables organizations to maintain that architectural flexibility and vendor neutrality as their AI strategies evolve. I wonder if you could share like, "Hey, I'm starting out. I've got maybe one model that I'm going to, but then as I evolve my model and I want to expand," how is the AI Hub going to enable them to do that maybe more quickly, securely, et cetera?"

DD Dasgupta

>> Right. Right. Now, what we've seen is customers wanting to be on this journey on their terms, right? So, we know data is difficult to move, it's called data gravity, but then outside of data, when you think about the different services, whether it's network services, security services, observability services, customers do not want to get locked into any one vendor, any one technology or any one cloud. Now, even in the initial days of AI, what we saw was CIOs that had already been locked into a cloud, when it came to AI, they're still stuck there because they've got all of their data, all of their systems, all of their tooling tied to that one vendor. And unfortunately, a lot of them are not able to take advantage of models from another cloud, a neocloud, or maybe their own homegrown models because 90% of the data is locked up, trapped up in one cloud. And so, it goes back again to the philosophy of Equinix and neutrality. Equinix, the N stands for neutrality, is in the name of the company and that has been our fundamental philosophy from the start. And what we're seeing now with the Distributed AI Hub announcement, if a customer wants to go with a security solution from Palo Alto or Fortinet or Zscaler, whoever it is, they want to be able to do that at a place which is behind their corporate firewall at an Equinix data center, but be able to mix and match these technologies. And then, going back to a point I made earlier around we're seeing a divergence in the requirements by industry vertical. A lot of these things started out as horizontal solutions, but turns out the competitive edge is in the model specific to a financial services, specific to healthcare, and then specific to what they're getting from a certain security vendor, an observability vendor. And so, that's what the whole concept is to enable that choice and flexibility, but then have it delivered closest to where the data is generated, so that it's not expensive for the customer to move that data around.

Bob Laliberte

>> Yeah. So, one of the questions I wanted to follow up on you with is that given the global events that are happening right now, sovereignty is a big concern. 280 data centers globally?

DD Dasgupta

>> Yep.

Bob Laliberte

>> So, I'm wondering if you could touch upon how you're able to help satisfy the sovereignty requirements for global organizations operating in certain parts of the world that require not only the data, but maybe other things to remain in that country?

DD Dasgupta

>> Right. No, 100%. So, look, let me unpack sovereignty into the way I think about it, it's kind of three levels. It starts with data sovereignty. Data cannot leave a country, a boundary, whatever. So, there's data sovereignty. Then, you have network sovereignty, which is data can move from UK to France, but it can't go through Germany, as an example. So, you've got network sovereignty. And then, what we're seeing now is AI sovereignty, which is yes, the first two are okay, but a model being trained in Germany cannot be trained by data from America or France or somewhere else. So, it's really that next level of sovereignty that countries, communities are needing. So, we are not a sovereignty provider. Equinix is not a sovereignty provider, but what we do enable our customers is we give them the knob of like, do you want just data residency? Do you want network sovereignty? Do you want network sovereignty? Or do you want all the way to AI sovereignty? And the way we're able to do that is because in our data centers, we've got the hyperscalers, right? 40% of everything that goes in and out of any cloud is happening at an Equinix data center. So, we've got the hyperscalers. We've got the sovereign clouds from the sovereign states, whether it's France or Japan or Germany. So, we've got the sovereign clouds. We've also got the companies' own data centers, and so we are the interface. And so, we can give the dial in the hands. And so, what I call this is really customer controlled sovereignty.

Bob Laliberte

>> Perfect.

DD Dasgupta

>> There are 172 privacy laws in the world. In US alone, 32 states have different privacy laws, so this is absolutely a situation where one solution is not going to satisfy everyone. The best thing we can do is give the control back to the customer.

Bob Laliberte

>> Got it. Got it. And that makes a lot of sense. And I like the way you positioned that with those different tiers of sovereignty because most people think data sovereignty, that's been the most common one, but that was great to be able to explain that and how you're able to give them the control to be able to do that. Distributed environments, a lot more tack points, a lot more potential risk and so forth. You had talked about earlier, you rattled off a few of the vendors that you're working with. I saw as part of the launch, you actually have an integration with Palo Alto Networks and their Prisma AIRS for AI security. I wonder if you could touch upon and explain how that's going to help enterprises deliver and implement a consistent security and governance across these distributed environments.

DD Dasgupta

>> Yeah. No, 100%. So, I have a philosophy which is centralize what you can and distribute what you must. Because let's face it, it's much simpler to do things centrally versus doing this in a distributed way. And we've learned that through distributed applications, distributed data, and now distributed intelligence. When you think about security, and to your point, the attack surface is absolutely growing at the same pace at which this whole AI revolution is growing. You don't even need a physical hacker anymore. You've got bots that can do that. So, that's an example of where the enforcement of the policy needs to be distributed, but trying to do the governance in a distributed fashion is not for the faint-hearted. I would not recommend it for anyone.

Bob Laliberte

>> Absolutely.

DD Dasgupta

>> So, our partnership, and we're starting with Palo Alto Networks and we're working with several other security and privacy and sovereignty vendors in the industry, it builds on that same philosophy. We want the governance to be centralized, so that we're not adding additional management complexity, regulatory complexity. All of that stays centralized, but the enforcement, we want these attacks to be prevented closest to where they enter from, which is at the edge. And that's why those 5,000... actually, it's more like 10,000 performance hubs that we've already built over the last 27 years, we're now enabling that to do that AI security at the edge. And so, again, it's the distributed model when it comes to enforcement, centralized governance.

Bob Laliberte

>> Yep. No, I love that. Centralized management for efficiencies and distributed enforcement for efficacy.

DD Dasgupta

>> There you go.

Bob Laliberte

>> Perfect.

DD Dasgupta

>> Yeah, absolutely.

Bob Laliberte

>> Yeah, that's great. So, let's take a step back, maybe look at the bigger picture for a minute. When you're talking to CIOs and other infrastructure leaders and they're trying to evaluate this type of an architecture, what business outcomes should they be thinking about? That's really become topical with AI. It's not about the technology, but about the outcome. So, when they're thinking about pulling together this infrastructure for a distributed, what kind of outcomes should they be thinking about?

DD Dasgupta

>> Yeah. I mean, it's one of those things like it is absolutely a board discussion now. Security, cloud and now AI, every CIO is asking that question to their CIOs and also to the lines of businesses. And we are in an AI race. Let's be very clear about that. This is how companies are going to compete. This is how the lines are going to move in terms of market share, their customers. So, it all boils down to time to market. And so, if you have a solution which helps you accelerate, move at the pace at which the CEO is asking without risking, of course, whether it's your data, your IP, that's at the heart of what every CIO asks us. "How can you help me accelerate?" And that's where having 280 data centers in over 72 locations, if there isn't space and capacity in London, we can stand up a customer in Manchester literally in a matter of days. We don't have to wait for the new data center to be built or our space. So, we start with our ability to deliver that time to market with just physical space, power and network connectivity. But when they get started with us... No data center is an island, this is a team sport, we'll need all the vendors. But guess what? In each of these data centers, we have thousands of ecosystem partners. So, they have their partners right next to them, literally, maybe even in the same rack or the rack right next to them. And that partner could be an AWS, it could be a Google, or it could be Salesforce, the list goes on. And so, that's the other way we can give them the time-to-market advantage. And so, time to market is at the heart of, I think what you're asking, Bob, and we're very happy to be able to deliver that turnkey solution and get a customer up and running.

Bob Laliberte

>> No, I think that's great. So, not only flexibility and choice, but also agility to be able to rapidly move to where you need to when you need to. No, I think that's great. Well, this has been great. We're running out of time, so I have one last question for you so put on your visionary hat. As we're looking ahead and enterprises move deeper into distributed and agentic AI, how do you see the infrastructure architectures evolving over the next couple of years?

DD Dasgupta

>> Yeah, I think it is going to be a case of specialization and hyper-specialization. You think about what the clouds have done for the entire industry, they've been able to solve a lot of what I call the horizontal challenges. But now, again, going back to what a bank needs, actually, even within a bank. High-frequency trading versus capital banks versus insurance versus mortgage, very different models, very different risk security profiles, time to market, things like that. So, I think the job of the infrastructure is to mirror the application and the job of the application is to mirror what the business is trying to do. We're just going to see more and more hyper-specialization and these requests coming to us infrastructure vendors is, "Make this top-to-bottom integrated for the one thing that I'm trying to do." So, hyper-specialization is where I'm going to end that on.

Bob Laliberte

>> Excellent. That sounds great. DD, thank you so much for sharing all your perspectives. I think it's clear that the AI infrastructures are becoming a lot more distributed. So, the ability to connect, secure, orchestrate across that AI ecosystem is really going to be a critical part of what enterprise have for their infrastructure strategy. So, thank you again. This was great. For everyone else, for more information about this announcement and what Equinix is doing, please visit their website. And thank you all for joining us.