Vertiv Models Power, Cooling in Nvidia Omniverse to Speed AI Buildouts

NVIDIA GTC Studio with Insights from Vertiv — NVIDIA

NVIDIA

Surging GPU energy demands require operators to replace bespoke data center construction with pre-simulated, factory-standardized infrastructure to support 600-kilowatt racks.

AI-assisted summary of NVIDIA GTC Studio with Insights from Vertiv — original source. Check the original for full context.

Source published:

Key ideas

  • Vertiv is shifting data center deployment from bespoke, on-site construction to a factory-like model using pre-engineered, standardized building blocks. (5:54)
  • The integration of NVIDIA Omniverse allows Vertiv to create digital twins of power and cooling chains, enabling physics-based simulations before physical construction begins. (3:34)
  • The industry is moving away from general-purpose design toward purpose-built 'AI factories' that are hyper-optimized for specific GPU cluster architectures. (8:59)
  • Vertiv is proactively solving for future physics-based constraints, including the management of racks reaching 600 kilowatts and scaling site capacities to the gigawatt level. (10:22)

Detailed notes

## Background This NVIDIA GTC Studio episode was published by the official NVIDIA YouTube channel on March 30, 2026. The program features Scott Armul, Chief Product and Technology Officer at Vertiv, who discusses the critical role Vertiv plays in delivering power and cooling infrastructure to support the massive scale of modern AI factories.

## TL;DR The discussion focuses on the transition from traditional, bespoke data center construction to a standardized, converged infrastructure model required for high-density AI workloads. Armul explains how Vertiv leverages NVIDIA Omniverse to create digital twins of power and cooling systems, allowing for physics-based simulations that accelerate deployment and ensure performance before physical construction begins. By positioning themselves as the foundational layer of the 'five-layer cake' of AI architecture, Vertiv aims to solve the heat and power challenges of future gigawatt-scale data centers.

## Atmosphere & style The tone is professional, forward-looking, and intellectually engaging, reflecting the high-stakes environment of an industry trade conference. Host Tiffany Jansen creates a warm and accessible rapport with Armul, successfully balancing technical depth with relatable analogies that make complex engineering concepts feel tangible. There is a palpable sense of urgency and excitement throughout the conversation, as both speakers emphasize the rapid, 'compounding acceleration' of AI technology and the critical importance of keeping pace with it.

## Bulletin - [05:54](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=354s) Vertiv is shifting data center deployment from bespoke, on-site construction to a factory-like model using pre-engineered, standardized building blocks. - [03:34](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=214s) The integration of NVIDIA Omniverse allows Vertiv to create digital twins of power and cooling chains, enabling physics-based simulations before physical construction begins. - [08:59](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=539s) The industry is moving away from general-purpose design toward purpose-built 'AI factories' that are hyper-optimized for specific GPU cluster architectures. - [10:22](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=622s) Vertiv is proactively solving for future physics-based constraints, including the management of racks reaching 600 kilowatts and scaling site capacities to the gigawatt level.

## Key takeaways - [00:10](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=10s) Vertiv’s rise in public visibility is directly linked to the rapid expansion of generative AI and their close technical partnership with NVIDIA. - [01:30](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=90s) Modern AI models require massive clusters of GPUs, necessitating dedicated physical infrastructure to handle the resulting power consumption and heat output. - [02:00](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=120s) Vertiv’s cooling strategy spans a full heat dissipation path: from extracting heat at the silicon level to removal from the data hall and final atmospheric rejection or reuse. - [03:34](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=214s) Simulating infrastructure in NVIDIA Omniverse allows Vertiv to verify performance and compatibility digitally, significantly accelerating the 'time to token' for new AI clusters. - [06:12](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=372s) The shift toward converged physical infrastructure eliminates the inefficiencies of integrating piecemeal products from disparate vendors on-site. - [07:15](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=435s) In the 'five-layer cake' analogy of AI architecture, Vertiv occupies the foundational infrastructure layer. - [07:30](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=450s) Standardized building blocks are critical for scaling data centers to gigawatt capacities without the logistical chaos of traditional construction. - [08:59](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=539s) Data center design is no longer 'agnostic'; infrastructure must now be engineered specifically to match the requirements of the computing hardware it supports. - [09:45](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=585s) The industry is evolving to treat the entire data center site as a single, optimized 'computer' rather than a collection of independent server rooms. - [10:22](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=622s) Vertiv maintains a 'maniacal' focus on looking multiple GPU generations ahead to ensure infrastructure capabilities keep pace with hardware performance roadmaps.

## Detailed walkthrough ### [00:10](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=10s) From Behind-the-Scenes to the Spotlight: Vertiv's Growing Recognition

Tiffany Jansen welcomes Scott Armul, the Chief Product and Technology Officer at Vertiv, to the NVIDIA GTC Studio. Jansen opens the conversation with a lighthearted admission: she has been eagerly anticipating this sit-down because Vertiv is tackling a massive, long-standing challenge in AI infrastructure that the general public is only now starting to fully appreciate. Armul agrees with this assessment, noting that Vertiv has historically transitioned from being a critical but relatively obscure player in the data center industry to a company that is now firmly in the spotlight.

Armul attributes this newfound public awareness and industry recognition to Vertiv's strong financial performance and rapid growth, but most importantly, to their deep partnership with NVIDIA. By jointly developing cutting-edge power and cooling solutions, Vertiv has captured the attention of people who previously paid little mind to the physical infrastructure space. Armul jokes that even his own neighbors finally understand what he does for a living, illustrating how the rise of generative AI has pushed data center infrastructure into the mainstream consciousness.

[01:30](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=90s) Explaining the Core Challenge: Power and Cooling for AI

To explain Vertiv's mission in simple terms, Jansen shares an anecdote about how she described the company to her mother earlier that morning. She explained that the advanced AI models and chatbots people interact with daily require thousands of GPUs to run, and to keep those GPUs functioning, they need two fundamental things: power and cooling. Armul enthusiastically validates this high-level explanation, confirming that every single query, inference, or digital interaction relies on physical GPUs sitting inside a data center.

Armul explains that these high-performance chips consume immense amounts of power and, as a consequence, generate extraordinary amounts of heat. Vertiv's job is to design and deploy critical infrastructure—ranging from traditional air cooling to advanced liquid cooling systems—to safely extract that heat from the chip, pull it out of the server racks, remove it from the data hall, and ultimately reject it into the atmosphere or find innovative ways to reuse it.

As Scott Armul explains: > "To query that chatbot, to run that AI inference, to do all of those things that a normal person would interact with AI or candidly any data center, those servers, those chips, those GPUs have to sit somewhere."

- **Core Infrastructure Elements:** Vertiv designs and delivers the physical systems that power and cool data centers, including liquid cooling, air cooling, and heat rejection systems. - **Heat Dissipation Path:** Heat is extracted directly from the silicon chips, pulled out of the server racks, removed from the data hall, and rejected into the atmosphere or redirected for heat reuse to minimize environmental impact.

[03:34](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=214s) GTC Announcements: Digital Twins and the Omniverse

Turning to the major news of the week at GTC, Armul highlights Vertiv's groundbreaking work within NVIDIA Omniverse, specifically mentioning their converged systems engineering approach. Vertiv is partnering with NVIDIA to pre-define and simulate entire data center environments digitally before any physical construction begins. By utilizing physics-based simulations and emulations within the Omniverse platform, Vertiv can model how their power and thermal chains—including liquid cooling and heat rejection systems—will interact with NVIDIA's chip clusters and servers under real-world workloads.

This digital twin approach represents a massive shift away from traditional, bespoke on-site construction. Instead of solving compatibility and performance issues after equipment has been delivered and commissioned, Vertiv can verify the entire system's behavior virtually. This drastically accelerates deployment timelines, helping customers achieve a faster "time to token."

As Scott Armul notes: > "Within NVIDIA's Omniverse, we're trying to drive towards more physics-based and simulated types of designs to speed up a lot of the converged infrastructure approach so that long before a shovel ever hits the ground or a construction company is trying to put up a data center, we understand the performance of our gear."

```mermaid flowchart TD accTitle: Vertiv Digital Twin to Physical Deployment Workflow accDescr: A flowchart showing how Vertiv uses NVIDIA Omniverse to simulate, pre-fabricate, and deploy converged data center infrastructure.

subgraph Digital_Phase ["Digital Simulation (NVIDIA Omniverse)"] A["Physics-Based Simulation"] -->|"Model thermal & power chains"| B["Pre-defined Interfaces & Block Sizes"] B -->|"Verify chip & cluster performance"| C["Digital Twin Validation"] end

subgraph Physical_Phase ["Physical Deployment"] C -->|"Pre-fabricated assembly"| D["Factory-Like Construction"] D -->|"Standardized connection points"| E["Rapid On-Site Commissioning"] E -->|"Faster 'Time to Token'"| F["Operational AI Factory"] end ```

- **Omniverse Integration:** Vertiv uses NVIDIA Omniverse to create digital twins of data centers, allowing for physics-based simulation of thermal and power performance. - **Time to Token:** By simulating the interaction of chips, clusters, and infrastructure digitally, Vertiv accelerates physical deployment and reduces the time required to get AI clusters operational.

[06:12](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=372s) The Foundation of the AI Factory: Converged Physical Infrastructure

Jansen asks Armul to demystify the concept of "converged physical infrastructure," a term that can easily sound like corporate buzzwords to those outside the industry. Armul explains that converged infrastructure is about moving away from piecemeal, "point products" designed by isolated companies that must be awkwardly integrated on-site by consulting engineers and construction crews. Instead, Vertiv designs pre-engineered, standardized building blocks with predefined interfaces and connection points, fully simulated in a CAD and Omniverse environment.

When discussing Vertiv's position in the broader AI ecosystem, Armul enthusiastically embraces the popular "five-layer cake" analogy of AI architecture, proudly claiming the bottom layer. Jansen calls this the most important, foundational layer, and Armul agrees, explaining that a robust, purpose-built foundation allows operators to scale data centers to gigawatt levels without the chaos of multiple trades clashing on-site. By treating data center deployment more like a standardized factory assembly line, Vertiv helps companies unlock superior performance, achieve faster time-to-market, and lower their overall total cost of ownership (TCO).

As Scott Armul puts it: > "In the five layer cake analogy that has become so fun to dissect this week, we're the infrastructure layer."

- **Five-Layer Cake Analogy:** Vertiv represents the foundational physical infrastructure layer upon which the rest of the AI stack (hardware, software, models, applications) is built. - **Scale Challenges:** Standardized, converged building blocks are essential for scaling data centers to gigawatt levels without the complexity of traditional, bespoke construction.

### [08:59](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=539s) Shifting Paradigms in AI Infrastructure Scott Armul explains that the era of "agnostic" data center design has ended. Previously, data centers were built around general-purpose CPU-based servers with relatively predictable energy and cooling requirements, where the specific workload was rarely a primary concern for infrastructure engineers. AI has fundamentally altered this by moving toward a "unit of compute" model where the hardware is hyper-optimized for specific performance parameters.

By understanding the architecture of an NVIDIA "super pod," Vertiv can work backward from these purpose-built clusters to determine exactly what the power and cooling infrastructure must look like to achieve maximum efficiency. Armul notes, “We can scale that same analogy and that same approach out to the entire site so that that site is optimized to work as a single computer.”

### [10:22](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=622s) Future-Proofing for the Gigawatt Era Looking ahead, Armul describes the industry’s state as one of "compounding acceleration." Because the roadmap for GPU performance is unlocking new capabilities at such a rapid pace, Vertiv is forced to look multiple generations ahead to solve looming physics-based constraints. The primary challenge involves managing massive energy density, such as powering racks that reach 600 kilowatts or higher, and scaling entire "AI factories" to reach gigawatt-scale capacity.

To meet these demands, the company works "somewhat maniacally" to ensure their technology investments and development roadmaps stay ahead of the hardware. The goal is to ensure the physical infrastructure is ready and tested well before the next generation of GPU clusters arrives for deployment. As Armul puts it, “We're in a race to try to not only keep up with that pace, but we're trying to look multiple GPU generations ahead to say, what are the physics-based problems we're going to need to solve?”

- **Primary Challenges:** - Managing high-density racks (600 kilowatts+). - Scaling total site capacity to the gigawatt level. - Synchronizing infrastructure readiness with GPU release cadences.

## Notable quotes > [01:30](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=90s) To query that chatbot, to run that AI inference, to do all of those things that a normal person would interact with AI or candidly any data center, those servers, those chips, those GPUs have to sit somewhere. — Scott Armul

> [10:22](https://www.youtube.com/watch?v=MyNEr7tf-WQ&t=622s) We're trying to look multiple GPU generations ahead to say, what are the physics-based problems we're going to need to solve? — Scott Armul

## Mentioned Scott Armul, Tiffany Jansen, Vertiv, NVIDIA, NVIDIA GTC, NVIDIA Omniverse

## Video description Scott Armul, Chief Product and Technology Officer at Vertiv, explains how Vertiv meets AI infrastructure demands at scale—delivering the power and cooling systems behind modern data centers and developing the solutions to streamline the design and deployment of next-generation AI factories.

[Source](https://www.youtube.com/watch?v=MyNEr7tf-WQ)

Reading summary

As generative AI workloads intensify physical demands on data centers, Vertiv is co-developing power and cooling infrastructure with NVIDIA to transition facilities from fragmented, bespoke builds into standardized, pre-simulated compute factories. Power and Cooling as Foundational Constraints (1:30) Every AI query and inference task relies on high-performance GPUs that consume vast amounts of electricity and generate extreme heat. Scott Armul, Chief Product and Technology Officer at Vertiv, explains that physical infrastructure operates as the bottom layer of the AI technology stack, tasked with extracting thermal energy directly from silicon chips, routing it out of server racks and data halls, and expelling it into the atmosphere or redirecting it for heat reuse. Accelerating Deployment Through Omniverse Digital Twins (3:34) To shorten the deployment cycle—often framed as 'time to token'—Vertiv is integrating physics-based modeling into NVIDIA Omniverse. Instead of diagnosing integration issues on-site during commissioning, engineers digitally model complete power and thermal chains under live workloads before ground is broken. This virtual validation allows teams to verify cluster behavior, standardize connection interfaces, and transition data center delivery toward factory-style pre-fabrication. From Workload-Agnostic Facilities to Single Computers (6:12) Armul states that traditional, workload-agnostic data centers built around general-purpose CPUs are being replaced by hyper-optimized 'units of compute.' Moving away from disparate point products integrated on-site, Vertiv designs converged physical infrastructure around architectures such as the NVIDIA super pod. By reverse-engineering power and cooling systems directly from the cluster's compute requirements, operators can optimize an entire physical site to function as a unified, cohesive machine. Engineering for 600-Kilowatt Racks and Gigawatt Sites (10:22) Anticipating rapid GPU performance gains, Vertiv designs infrastructure multiple hardware generations in advance to prevent physical bottlenecks. According to Armul, the primary engineering hurdles involve managing unprecedented energy densities—including server racks reaching 600 kilowatts or more—and scaling aggregate campus capacity to gigawatt levels. Keeping development roadmaps synchronized with fast GPU release cycles is critical to ensuring cooling and power delivery systems are fully validated prior to chip deployments. While digital twin modeling and pre-engineered modular blocks provide a pathway to manage extreme rack densities and gigawatt-scale sites, maintaining synchronization between rapid GPU development cycles and long-lead physical infrastructure remains a compounding technical hurdle.

Source quotes

[1:30] Scott Armul: To query that chatbot, to run that AI inference, to do all of those things that a normal person would interact with AI or candidly any data center, those servers, those chips, those GPUs have to sit somewhere. [10:22] Scott Armul: We're trying to look multiple GPU generations ahead to say, what are the physics-based problems we're going to need to solve?

Original transcript

[0:10] 1 Hi, everyone. [0:10] 2 Welcome to the NVIDIA GTC Studio. [0:13] 3 My name is Tiffany Jansen, and today I [0:15] 4 am joined with a very special guest, Scott [0:18] 5 Armel from Vertiv. [0:19] 6 Scott is the Chief Product and Technology Officer. [0:22] 7 Scott, how are you today? [0:23] 8 I'm doing well. [0:24] 9 Thank you very much for having me. [0:25] 10 Thank you for being here. [0:26] 11 Scott, I have to admit something. [0:28] 12 When I saw that we were having a [0:30] 13 sit-down conversation, I had been really looking [0:33] 14 forward to this because what you are doing [0:36] 15 at Vertiv is solving a really big challenge [0:39] 16 with AI and infrastructure that I think, I [0:42] 17 mean, I know you know this for many, [0:44] 18 many years now, but I think the general [0:46] 19 public is starting to become aware of. [0:48] 20 I think that's fair. [0:49] 21 From a Vertiv perspective, I think we've gone [0:52] 22 from important to the data center in the [0:55] 23 industry, but relatively unknown to folks who are [0:58] 24 actually paying attention to who we are and [1:00] 25 what we do now. [1:01] 26 Vertiv, we're the power and cooling and overall [1:04] 27 solutions [1:05] infrastructure folks from a data center level. [1:07] 28 We're helping to actually build and fabricate and [1:11] 29 design the data centers of the future. [1:12] 30 I would say a lot of the awareness [1:13] 31 and the industry recognition comes back to some [1:16] 32 of our performance, our growth, and all of [1:17] 33 those things, but the partnership with NVIDIA and [1:20] 34 kind of jointly developing a lot of the [1:22] 35 things we're talking about, I think has really [1:23] 36 put us on a map to a lot [1:25] 37 of folks that weren't [1:26] previously paying attention to [1:27] 38 this space or to us in general. [1:29] 39 Absolutely. [1:30] 40 Before we started filming, I know we were [1:32] 41 speaking about just that and you mentioned now [1:34] 42 my neighbors actually know what I do. [1:36] 43 How do you explain to your neighbors, for [1:39] 44 example, what does Vertiv solve? [1:41] 45 What challenges does it solve? [1:43] 46 Yeah. [1:43] 47 So Vertiv, like I said, we're really kind [1:45] 48 of the critical infrastructure folks. [1:47] 49 So when we think of powering data centers, [1:49] 50 we think of cooling data centers, and we [1:51] 51 think of how we actually deploy and bring [1:54] 52 up the physical infrastructure of the data center [1:56] 53 building and [1:57] all of the interconnective pieces themselves. [1:59] 54 That is what Vertiv does. [2:00] 55 We design it. [2:02] 56 We think through kind of the future state [2:03] 57 of where products and solutions need to go [2:05] 58 to enable the AI solutions and the chip [2:08] 59 architectures that Jensen and the NVIDIA team are [2:11] 60 talking about. [2:12] 61 We help as kind of the foundational infrastructure [2:14] 62 level bring that to reality. [2:17] 63 Absolutely. [2:17] 64 When I was speaking to my mom this [2:20] 65 morning and [2:20] sharing about Vertiv, she's like, well, [2:22] 66 what do they do? [2:23] 67 And here's how I explained it. [2:24] 68 Tell me if this is accurate at a [2:26] 69 very high level. [2:27] 70 I said, you know, models we use that [2:30] 71 run, you know, different chatbots that we interact [2:33] 72 with, they need thousands of GPUs to run. [2:36] 73 But for those GPUs to run, they need [2:38] 74 two things [2:39] effectively, which is power and cooling. [2:41] 75 And those two things can bring challenges. [2:44] 76 And basically that is what you are solving. [2:46] 77 100%. [2:47] 78 To query that chatbot, to run that AI [2:51] 79 inference, to do all of those things that [2:54] 80 a normal person would interact with AI or [2:56] 81 candidly [2:57] any data center, those servers, those chips, [3:00] 82 those GPUs have to sit somewhere. [3:02] 83 They sit in a data center. [3:03] 84 That data center has to deliver power to [3:06] 85 that chip so that it can run. [3:08] 86 And effectively that chip creates heat. [3:10] 87 And a lot of the solutions that Vertiv [3:12] 88 deploys, whether [3:13] it's liquid cooling or air cooling [3:14] 89 or other types of approaches is about removing [3:17] 90 that heat from the chip, getting it out [3:19] 91 of the racks, getting it out of the [3:20] 92 data hall in the building and figuring out [3:22] 93 how to get it to the atmosphere or [3:23] 94 ideally to figure out ways to reuse it [3:26] 95 to make all of these data centers we're [3:28] 96 driving more [3:29] efficient, more effective and maybe less [3:31] 97 impactful to the world as well. [3:33] 98 Absolutely. [3:34] 99 You had some big announcements this week here [3:36] 100 at GTC. [3:37] 101 Could you share some of them with me? [3:38] 102 Yeah. [3:39] 103 The biggest announcement I think was the creation [3:42] 104 of the Omniverse Rubin DSX, our one core [3:45] 105 solution for that. [3:47] 106 Really what that means from a Vertiv perspective [3:49] 107 is an end-to-end, pre-designed, pre [3:52] 108 -fabricated, [3:53] converged systems engineering approach to a full [3:56] 109 data center. [3:57] 110 So we think of all of the elements [3:59] 111 of powertrain [4:00] and thermal chain, the liquid cooling [4:02] 112 we talked about, the heat rejection of a [4:04] 113 data center. [4:05] 114 We have partnered with NVIDIA and others to [4:09] 115 pre-define what all of those interfaces look [4:11] 116 like, what that [4:12] site solution ultimately looks like [4:14] 117 and to do it in a way that [4:16] 118 is maybe changing how we would have done [4:18] 119 data center design previously and that we're doing [4:21] 120 it digitally. [4:22] 121 So within [4:23] NVIDIA's Omniverse, we're trying to drive [4:25] 122 towards more physics-based and simulated types of [4:28] 123 designs to speed up a lot of the [4:30] 124 converged infrastructure [4:32] approach so that long before a [4:34] 125 shovel ever hits the ground or a construction [4:35] 126 company is trying to put up a data [4:37] 127 center, we understand the performance of our gear. [4:40] 128 We understand how the chips and the clusters [4:42] 129 and all of the servers working together will [4:44] 130 perform in that environment and we can launch [4:46] 131 that to the industry or to all of [4:49] 132 the customers so that we can help speed [4:51] 133 deployments and we can help enable us to [4:54] 134 get time to token a little quicker. [4:56] 135 Absolutely. [4:57] 136 And that's one of the biggest things that [4:58] 137 comes to mind for me anyways when I [5:00] 138 think about digital twins is really how they [5:04] 139 enable companies to move so much faster now. [5:06] 140 Do you have any examples, maybe not even [5:08] 141 specific numbers, but examples of how you've seen [5:12] 142 using digital twins speed up the process? [5:15] 143 I think it's critical that digital twins are [5:19] 144 really an enabler for us to get out [5:20] 145 of this bespoke on-site custom construction world [5:25] 146 that we've previously lived in to something that [5:28] 147 is truly more pre-designed and thoughtful around [5:31] 148 the building [5:32] blocks that have predefined interfaces and [5:34] 149 connection points [5:35] and how ultimately our gear needs [5:37] 150 to be deployed on-site in a way [5:38] 151 that doesn't [5:39] require the complexity of figuring out [5:41] 152 after it's already been delivered or after it's [5:44] 153 already been commissioned. [5:45] 154 That's really been the name of the game [5:47] 155 for us is figuring out an enablement path [5:49] 156 for us to do it digitally with physics [5:53] 157 -based [5:54] capability within the products themselves so that [5:56] 158 we know those simulations and those emulations of [5:58] 159 how the infrastructure will behave are true. [6:01] 160 Exactly. [6:02] 161 Yeah, I like that. [6:03] 162 The word thoughtfully. [6:05] 163 I think that's such a powerful word to [6:08] 164 describe what digital twins are doing. [6:09] 165 It's simple, but it's very powerful. [6:11] 166 Absolutely. [6:12] 167 Scott, [6:13] when it comes to converged  physical infrastructure, [6:16] 168 how do you explain this topic to folks [6:21] 169 listening who maybe aren't as familiar with it? [6:23] 170 It's a lot of big words. [6:24] 171 Potentially a lot of big words and a [6:26] 172 lot of buzzwords as well. [6:27] 173 We hope it's not just corporate speak, but [6:30] 174 really converged [6:31] infrastructure is trying to get out [6:32] 175 of this idea of individual point products that [6:36] 176 weren't [6:37] designed together that were  potentially designed by [6:40] 177 a bunch of independent companies all trying to [6:42] 178 come together on-site and it becomes either [6:44] 179 a consulting [6:45] engineer's problem, a  construction company's problem. [6:48] 180 Converged [6:49] infrastructure for us is really about, I'll [6:51] 181 use the word thoughtfully [6:52] again, the predefined block [6:54] 182 sizes, [6:55] the interface points, the deployment model for [6:58] 183 that infrastructure. [6:58] 184 We do it in a digital simulated environment. [7:02] 185 We do it in a CAD environment so [7:03] 186 that we can actually see how all of [7:05] 187 the piece parts will exist in Omniverse or [7:08] 188 in a digital environment long before it's ever [7:11] 189 put out into the marketplace. [7:13] 190 Those key interface points help us to actually [7:16] 191 design better total solutions as opposed to better [7:19] 192 point products. [7:20] 193 Exactly. [7:21] 194 Total key solutions it really comes down to. [7:23] 195 When you think about that, the total key [7:26] 196 solution, we spoke about this a little bit [7:28] 197 throughout the conversation, [7:29] but where would you place [7:31] 198 Vertiv throughout the entire AI infrastructure? [7:35] 199 In the five layer cake analogy that has [7:39] 200 become so fun to dissect this week, we're [7:42] 201 the infrastructure layer. [7:43] 202 The most important layer. [7:45] 203 It's foundational. [7:46] 204 Now you said it, I didn't, but I [7:47] 205 will run with that. [7:48] 206 The most [7:49] important layer because it's the foundation [7:50] 207 of everything that eventually gets built on top. [7:53] 208 When that foundation is correct and when that [7:56] 209 foundation is defined and deployed in a way [8:00] 210 that is thoughtful and purpose built and with [8:03] 211 these end solutions in mind, it allows us [8:05] 212 to take that infrastructure layer, pair it with [8:08] 213 how energy comes into data centers, pair it [8:10] 214 with a thoughtful and more efficient way in [8:13] 215 which the DSX deployments need to be run. [8:17] 216 It allows us to think of the entire [8:19] 217 chain. [8:19] 218 When we think of that entire chain, that [8:22] 219 purpose built design, [8:23] those converged infrastructure points, we [8:25] 220 can unlock better performance. [8:26] 221 We can drive faster time to market. [8:28] 222 We can drive lower overall TCO for the [8:32] 223 companies we work with in a way that [8:34] 224 when you think about scaling to gigawatt levels, [8:36] 225 we can't be fumbling over ourselves and multiple [8:40] 226 parties and different trades all trying to work [8:42] 227 at the same time. [8:43] 228 These [8:44] converged building blocks  and this converged approach [8:46] 229 to design ideally makes us look and feel [8:49] 230 a lot more like a factory. [8:50] 231 Absolutely. [8:51] 232 And that's [8:51] why it's the foundational, aka important, [8:55] 233 most important layer because it really needs to [8:59] 234 all come together. [8:59] 235 Scott, how are you seeing, and I know [9:02] 236 you just kind of gave a few examples [9:03] 237 of this, but how are you seeing emerging [9:04] 238 AI [9:05] workloads and system architectures  changing traditional assumptions [9:08] 239 about power density and cooling in data centers? [9:12] 240 I think in previous years, data centers candidly [9:15] 241 didn't really care what the workload was. [9:18] 242 It was going to run on a CPU [9:19] 243 -based server. [9:20] 244 It was going to have a particular heat [9:22] 245 draw or energy draw, and you didn't really [9:25] 246 need to know or understand. [9:26] 247 The difference with AI is it is kind [9:29] 248 of a unit of compute type of an [9:31] 249 approach where it's a very optimized for that [9:35] 250 [9:36] either performance parameter  or for that workload, and [9:39] 251 we know what it will look like. [9:41] 252 When we think of kind of a cluster [9:43] 253 size or a super pod in NVIDIA's nomenclature, [9:46] 254 we can then back up from those purpose [9:48] 255 -built pods to what does the powertrain and [9:51] 256 the infrastructure need to look like to best [9:53] 257 support that from a size and optimization and [9:56] 258 efficiency perspective, and we can scale that same [9:59] 259 analogy and that same approach out to the [10:01] 260 entire site so that that site is optimized [10:03] 261 to work as a single computer. [10:05] 262 Wow. [10:06] 263 It's incredible. [10:07] 264 It's so exciting the times we're in, and [10:10] 265 I know before we were filming, again, you [10:12] 266 were mentioning that you've been with Veritiv for [10:15] 267 a while now, so getting to see the [10:17] 268 evolution and to where we are now, what [10:20] 269 do you think about when we look ahead [10:21] 270 a little bit? [10:22] 271 I know this can be a vague question, [10:23] 272 but when we look ahead a little bit, [10:25] 273 I need to know. [10:26] 274 I need to pick your brain. [10:27] 275 What does it look like when it comes [10:28] 276 to this stuff? [10:29] 277 Yeah. [10:30] 278 I think [10:30] we're seeing compounding acceleration, which is [10:34] 279 extremely exciting for us. [10:36] 280 From [10:38] our perspective, our engineers, our service folks, [10:40] 281 we're tied into the direction that GPUs are [10:43] 282 going. [10:43] 283 The pace of technology is accelerating. [10:45] 284 The roadmap that NVIDIA is putting out there [10:47] 285 for GPUs and performance improvement that is being [10:50] 286 unlocked is truly mind-blowing, and from our [10:54] 287 perspective, we're in a race to try to [10:55] 288 not only keep up with that pace, but [10:57] 289 we're trying to [10:58] look multiple GPU generations ahead [11:00] 290 to say, what are the physics-based problems [11:01] 291 we're going to need to solve? [11:02] 292 How do I get this much power into [11:04] 293 a rack that's 600 megawatts or beyond? [11:07] 294 How do I scale to a data center [11:09] 295 site or an AI factory that is a [11:11] 296 gigawatt [11:12] in capacity and work backwards somewhat maniacally [11:15] 297 to say, all right, here's what the roadmap [11:17] 298 needs to look like. [11:17] 299 Here's what the development needs to look like. [11:19] 300 Here are our technology investments that are going [11:21] 301 to be needed for us to make sure [11:22] 302 that we're not tripping over those physics-based [11:24] 303 problems, and we're actually able to turn on [11:26] 304 data centers well in advance of when the [11:29] 305 GPUs and the clusters will be ready to [11:31] 306 be deployed there. [11:32] 307 Absolutely. [11:32] 308 I mean, it's easy for me to say, [11:34] 309 but although [11:35] these are different challenges that you're [11:37] 310 identifying and [11:38] will solve and are solving already, [11:40] 311 they're exciting ones. [11:41] 312 Scott, thank you so much for your time [11:43] 313 today. [11:43] 314 It was really great to sit down with [11:45] 315 you and learn more about what Vertiv is [11:47] 316 doing, especially in partnership with NVIDIA. [11:49] 317 Absolutely. [11:49] 318 It's my pleasure. [11:50] 319 Thank you. [11:51] 320 Thank you. [11:51] 321 Thank you so much, Scott. [11:51] 322 And for those of you tuning in, make [11:52] 323 sure to go check out more sessions on [11:54] 324 the GTC catalog.