
We often think of Artificial Intelligence as something ethereal—code floating in the “cloud.” But the reality of 2026 is grounded in steel, concrete, and energy supply. With Big Tech “hyperscalers” like Microsoft, Amazon, Meta, and Google projected to spend between $600 and $650 billion this year alone on infrastructure, we are witnessing an industrial buildout that economists argue now rivals the railroad expansion of the 1850s. While the first phase of the AI revolution was defined by who could design the smartest chip, the next phase will be defined by who can physically power it. The opportunity for investors is no longer just about the model; it is about the massive machine required to run it.
What is an AI Data Center?
Unlike traditional data centers that store files and run websites, AI data centers are designed for massively parallel computing. This creates unprecedented heat and power density. (Geek alert) While a standard server rack consumes 5–10 kW, a modern rack equipped with NVIDIA Blackwell or H100 GPUs demands 40 kW to over 100 kW. This physical constraint is forcing a complete redesign of the data centers needed in the AI world.
The Builders & Locations
The build-out is led by the “Hyperscalers”—Microsoft, Amazon (AWS), Google, and Meta—who are racing to secure land and power. Because these facilities consume gigawatts of electricity (enough to power small cities), they are moving away from congested hubs like Northern Virginia to regions with available power and land.
- The Power Belt (Ohio, Indiana, Midwest): Emerging as a prime location due to grid capacity and land availability.
- Texas: Attracting massive investment due to its independent energy grid and renewable abundance.
- Nuclear Sites: Data centers are increasingly being built “behind the meter” at nuclear plants (e.g., Pennsylvania) to guarantee 24/7 zero-carbon power.
Investment Opportunities: The “Picks & Shovels”
Investors are looking beyond the chipmakers to the companies building the physical infrastructure. This includes thermal management (cooling), power distribution, and real estate.
| Ticker | Company | Primary Role | Investment Thesis |
| VRT | Vertiv Holdings | Liquid Cooling | Market leader in thermal management; essential for cooling NVIDIA’s hot chips. |
| ETN | Eaton Corp | Power Mgmt | Critical electrical infrastructure (transformers, switchgear) for grid upgrades. |
| CEG | Constellation Energy | Nuclear Power | Owner of largest US nuclear fleet; prime partner for baseload AI power deals. |
| DLR | Digital Realty | Data Center REIT | Global landlord for data centers with strong hyperscaler relationships. |
| MOD | Modine Mfg | Thermal Systems | Pivot from automotive cooling to high-growth data center thermal solutions. |
| ANET | Arista Networks | AI Networking | Provides the ultra-fast ethernet switches needed to connect GPU clusters. |
Investment Note: This list is for informational purposes only and does not constitute financial advice. Stocks in this sector, particularly pure plays like Vertiv, have seen significant volatility.
As the AI race accelerates, the primary bottleneck has shifted from silicon availability to energy reliability. With modern GPU racks consuming over 100 kW of power—ten times the standard of a decade ago—the digital economy is colliding with the physical limits of the power grid. For the astute investor, this signals a clear pivot: the most durable long-term value may not be found in the volatile software wars, but in the “picks and shovels” of the infrastructure supply chain—the liquid cooling systems, the copper interconnects, and the baseload nuclear power that keep the lights on. The code is written, but the concrete is just beginning to pour.
Please comment or send an email if you have any questions.
Joe Baker, CFP®
Baker Asset Management, LLC
joebaker@bakerassetmanagement.com







