SPEAKER_1: Alright, so last episode we discussed the critical role of memory in AI data centers. But let's shift focus to what happens after the chips are purchased, packaged, and shipped. SPEAKER_2: That's the right question. Running a modern AI data center—dense with high-powered GPUs—requires significant electrical infrastructure, often overlooked by investors. SPEAKER_1: We're talking about companies beyond chip design or software, focusing on the essential infrastructure layer. SPEAKER_2: Exactly. Vertiv and Eaton are key players in the power, thermal, and electrical infrastructure essential for AI data centers. Every hyperscale facility relies on their products. SPEAKER_1: What do Vertiv and Eaton specifically supply? Listeners might picture a generic contractor, not an AI infrastructure leader. SPEAKER_2: Vertiv specializes in critical power and thermal management, including uninterruptible power supplies and liquid cooling systems. Eaton focuses on electrical distribution, such as switchgear and UPS systems. AI-driven power density changes are described as a 'tsunami effect' by Eaton's executives. SPEAKER_1: Wait—tsunami effect. That's a strong phrase. What does that actually mean in practice? SPEAKER_2: It means upgrades aren't isolated. When AI workloads push rack-level power density higher, you can't just swap one component. You need to rethink everything from the utility interconnect through distribution, UPS systems, and busways. The whole chain has to be redesigned. In that context, Eaton's management increased its multi-year data center growth outlook—revising its compound annual growth rate projection for that end market from 16% to roughly 25%, covering 2022 to 2025. SPEAKER_1: And Eaton's CEO Craig Arnold was specific about the order book. What did he actually say? SPEAKER_2: He stated that orders tied specifically to AI data center demand more than doubled on a trailing-twelve-month basis, and that negotiations in the U.S. increased over fourfold. That's not a gradual uptick—that's a step change in demand that reprices the business. SPEAKER_1: Mm-hmm. Now Vertiv's numbers are even more granular. Walk through what the financials actually showed. SPEAKER_2: [short pause] So in Q3 2024, Vertiv reported net sales of approximately $2.074 billion—up 19% year over year. Operating profit rose about 48%. The adjusted operating margin hit 20.1%, expanding 310 basis points versus the same quarter in 2023. And trailing twelve-month organic orders through September 2024 were up around 37%, driven by hyperscale and colocation data centers. SPEAKER_1: So not just revenue growth—margin expansion at the same time. That's the signature of pricing power, not just volume. SPEAKER_2: Correct. And it continued into Q4. Adjusted operating margin reached 21.5% that quarter, with organic net sales up 27% year over year—every region posting above 20% organic growth. For the full year 2024, net sales reached about $8 billion, adjusted operating profit approximately $1.552 billion, and the adjusted operating margin expanded roughly 410 basis points to 19.4%. Backlog hit around $7.4 billion by Q3 2024, up about 47% year over year. SPEAKER_1: That backlog number is striking. $7.4 billion means projects are being awarded well before capacity is needed. SPEAKER_2: [emphasis] That's the key idea for infrastructure suppliers specifically. Large power and cooling projects get contracted years in advance. Early-mover suppliers lock in visibility and pricing power that commodity vendors don't get. Vertiv also guided for 2025 revenue of roughly $9.125 to $9.275 billion, with 15 to 17% organic growth expected—management's signal that AI-driven demand continues accelerating. SPEAKER_1: But here's something worth flagging. Vertiv beat Q4 2024 earnings expectations—adjusted EPS of $0.99 against a consensus of $0.82—and the stock still declined on the news. How does someone tracking this series make sense of that? SPEAKER_2: It's a real tension. Rerated infrastructure names can face volatile sentiment even when fundamentals are strong. The market had already priced in a lot of the good news. That's the same dynamic we've seen at every link in this chain—the obvious winner gets priced, and then the question becomes what's still underpriced. For Vertiv and Eaton, the underappreciated piece may be liquid cooling. Vertiv's management described that market as still immature in 2024, even while reporting acceleration in liquid cooling revenue. SPEAKER_1: So liquid cooling is the next sub-bottleneck within the infrastructure layer. SPEAKER_2: Potentially. And the exposure isn't limited to hyperscalers. Colocation and enterprise facilities adapting to AI workloads are also upgrading. Eaton even launched a modular data center solution in North America in March 2024 specifically targeting edge computing, machine learning, and AI deployments. The rerating opportunity spans multiple data center classes, not just the largest cloud providers. SPEAKER_1: Now, what are the risks? Because the thesis sounds compelling, but there have to be pressure points. SPEAKER_2: Several. Supply-chain bottlenecks for specialized components. Labor constraints on installation and commissioning. Competition from other industrial players entering the space. And the biggest one: if hyperscaler capex slows—say, a macro downturn or a pause in AI investment—order books thin quickly. Eaton also noted that utilities historically modeled data center loads as relatively flat, and AI is breaking those assumptions. That creates a new risk: grid connection delays that slow deployments even when the equipment is ready. SPEAKER_1: That grid connection point is exactly where I want to go next. Because Eaton's own projection puts worldwide data center capex surpassing $1 trillion by 2029. That's an enormous amount of power demand hitting a grid that wasn't designed for it. SPEAKER_2: And that's the handoff. The rerating of Vertiv and Eaton shows that AI demand can reprice industrial infrastructure companies far from the original software narrative. But once you follow the power chain out of the data center and toward the grid itself, a new set of companies enters the picture—GE Vernova and the broader power-generation layer. That's where the constraint migrates next, and that's the subject of the next episode. SPEAKER_1: And that partnership with Nvidia is worth naming explicitly. Eaton isn't just selling switchgear into a generic industrial market anymore—they're co-developing solutions with the dominant AI chip company. That's a different kind of relationship. SPEAKER_2: It signals something important about where value is concentrating. When Nvidia needs an electrical infrastructure partner to ensure compute deployments aren't constrained by power and cooling, that tells you the bottleneck has genuinely migrated. The chip vendor is now dependent on the infrastructure vendor. SPEAKER_1: The key idea for someone tracking this series: this part of the AI trade isn't a semiconductor story. It's a picks-and-shovels story. SPEAKER_2: Exactly. Think of it this way—during a gold rush, the people selling shovels often capture more reliable returns than the miners themselves. Vertiv and Eaton don't bet on which AI model wins or which hyperscaler dominates. They sell into the physical requirement that every data center shares regardless of who's running it. SPEAKER_1: Mm-hmm. And the exposure is broader than most people assume. SPEAKER_2: Right—it's not just the largest cloud providers. Colocation facilities, enterprise data centers adapting to AI workloads, edge deployments. Eaton launched a modular data center solution in North America in March 2024 specifically targeting edge computing, machine learning, and AI. The rerating spans multiple data center classes. SPEAKER_1: Now, what about the service and maintenance side? Because there's a difference between one-time equipment demand and recurring revenue. SPEAKER_2: [short pause] That's the right pressure test. Large power and cooling installations require ongoing maintenance, software monitoring, and eventual replacement cycles. Eaton has been moving toward software-enhanced power management—they collaborated with Red Dot Analytics on predictive maintenance and anomaly detection for data centers. That's recurring revenue layered on top of the equipment sale. SPEAKER_1: So not purely a capital-equipment story. There's a services layer building underneath it. SPEAKER_2: And that matters for how investors should value it. One-time equipment demand is cyclical. Recurring service contracts tied to installed base are more durable. The question for someone tracking Vertiv and Eaton is how quickly that services mix grows relative to new equipment orders. SPEAKER_1: Wait—but Eaton also flagged something about utilities that I want to make sure we don't skip. Because it connects directly to the next episode. SPEAKER_2: Yes—utilities historically modeled data center loads as relatively flat. AI is breaking those assumptions entirely. Operators and utilities are being forced to rethink grid connections. Eaton projects worldwide data center capital expenditure surpassing one trillion dollars by 2029. That's an enormous amount of new power demand hitting infrastructure that wasn't designed for it. SPEAKER_1: And that's where the constraint migrates. Out of the data center, toward the grid itself. SPEAKER_2: [emphasis] That's the handoff. The rerating of Vertiv and Eaton demonstrates that AI demand can reprice industrial infrastructure companies far from the original software narrative. Now the question becomes: who generates and transmits the power those data centers need? That's GE Vernova and the broader power-generation layer—and that's where the next episode picks up.