3 Stocks Powering Data Center Growth: The Overlooked AI Infrastructure Play

Applied Digital designs, builds, and operates data center facilities engineered specifically for AI compute workloads from the ground up. That distinction matters more than it sounds. This is not a traditional co-location provider slapping some liquid cooling onto a legacy server farm and calling it AI-ready. Applied Digital's infrastructure is purpose-built for the extreme power density and cooling demands of NVIDIA Blackwell and similar high-density GPU clusters - rack densities that standard data centers flat-out cannot support without expensive, time-consuming retrofits.

The business model is straightforward: long-term, contracted revenue from hyperscalers and large AI training customers who need dedicated compute infrastructure but don't want to own it. Applied Digital acts as the capital intermediary - building the expensive physical plant, locking down power agreements, managing the facilities - while its customers focus on compute and software. It's a capital-heavy model, and there is real execution risk during construction phases. But once facilities go operational, the revenue is high-visibility and recurring. That's the trade-off.

What jumped out at me is the company's focus on sourcing power in markets with abundant, low-cost electricity - regions with available grid capacity and (increasingly important) proximity to renewable energy sources. While hyperscalers are scrambling to find sites in saturated markets like Northern Virginia and Silicon Valley, Applied Digital is quietly securing power agreements in underserved geographies. That makes them a supplier of something genuinely scarce right now: shovel-ready, high-power, AI-ready data center capacity. Revenue grew substantially year-on-year in fiscal 2025, driven by completed facility ramp-ups and expanding customer commitments.

Pure Storage might be the most overlooked name in the entire AI infrastructure conversation. And the reason is almost comically simple: when investors picture an AI data center, they think GPUs and power. Storage? That's the unglamorous plumbing nobody wants to talk about at dinner parties. But the physics here are unforgiving. A GPU cluster is only as productive as the speed at which you can shovel data into it. Bottlenecked storage turns expensive compute into expensive idle compute. That's not a nuance - it is a fundamental constraint.

Pure Storage's FlashBlade and FlashArray product lines are built for the high-throughput, low-latency data access that AI training clusters demand. The all-flash architecture delivers data to GPU clusters without the input-output bottlenecks you get from traditional hard-disk-drive (HDD) storage - bottlenecks that become crippling when training runs involve petabytes of data across thousands of GPUs running concurrently. The company has said publicly that AI-related storage demand is already a meaningful revenue driver, with customers deploying Pure Storage arrays specifically to eliminate storage as the chokepoint in their AI training pipelines.

So the product story is strong. But what makes Pure Storage especially interesting on a quant basis is the financial profile underneath. Revenue growth has exceeded 25% year-on-year in recent quarters, and gross margins keep improving as the software subscription layer, Evergreen, takes a bigger share of total revenue. Evergreen subscriptions generate recurring annual contract value - revenue visibility that extends well past any single product cycle - and the subscription gross margins are substantially higher than hardware. This mix shift toward software is structurally lifting earnings quality over time, which is exactly what quantitative screens pick up through rising EPS revision trends and improving profitability scores.

Johnson Controls is a 140-year-old industrial company whose primary business - building automation, HVAC systems, fire and security infrastructure - intersects directly with the data center boom through one critical technology: cooling. As AI rack power density has escalated from 5–15 kilowatts for traditional server racks to 50–120 kilowatts for the latest GPU cluster configurations, thermal management has emerged as a fundamental constraint on data center expansion. A data center that cannot adequately cool its compute is a data center that must throttle its GPUs - or risk hardware failure. Cooling, in this context, is not a background utility cost but a core enabling technology.

Johnson Controls supplies both traditional air-side cooling infrastructure and the advanced liquid cooling systems that AI data centers increasingly require. Liquid cooling - in which coolant is circulated directly past compute components rather than relying on air movement - is the only viable thermal management approach for the highest-density AI racks currently being deployed. JCI's portfolio spans chillers, precision air cooling, fluid dynamics systems, and the building automation software that orchestrates these systems in complex data center environments. Its OpenBlue digital platform provides the analytics and control layer that allows operators to optimise cooling efficiency, predict failures before they occur, and reduce power usage effectiveness (PUE) - a key operational metric for data center sustainability.

What makes Johnson Controls an interesting quantitative pick rather than simply a beneficiary of a long-term trend is its valuation context. Unlike Vertiv - whose stock has risen significantly and now trades at roughly 46 times forward earnings - Johnson Controls arrives at this data center growth story from a lower valuation starting point, with a large installed base of existing maintenance and upgrade contracts providing baseline revenue stability. The company's data center exposure is growing as a percentage of its overall building solutions business, creating a re-rating opportunity as the market recognises an industrial holding company increasingly operating as a critical data center infrastructure provider. Its Q3 2025 results and raised full-year guidance reflected strong order momentum tied specifically to AI-related data center construction projects.