How to Invest in Data Centers: The Physical Home of the AI Boom
.png "How to Invest in Data Centers: AI Infrastructure, REITs, and Energy Stocks Explained (2026 Guide)")
How to Invest in Data Centers: The Physical Home of the AI Boom
If you have read our previous post on the AI investment stack, you know that AI runs on six layers — from chip fabrication all the way up to energy infrastructure. But there is a physical layer that holds everything together that we have not covered in depth yet: the data center.
Every GPU cluster, every AI model, every cloud application lives inside a data center building. Before a single inference is run or a training job is completed, someone has to build the facility, run the power, maintain the cooling, and lease the space. That physical infrastructure is a multi-decade investment opportunity — and one most average investors overlook entirely.
What Is a Data Center — and Why Does AI Need So Many of Them?
A data center is a purpose-built facility that houses servers, networking equipment, storage systems, and the power and cooling infrastructure to keep them running. Think of it as the physical real estate that the entire digital economy rents space in.
AI has dramatically accelerated data center demand for three reasons:
- Training runs are compute-intensive. Training a single large AI model can consume megawatts of power over weeks — requiring purpose-built facilities with dense power delivery and sophisticated cooling.
- Inference at scale requires proximity. Every time someone uses an AI product, the model runs an inference. Billions of inferences per day require distributed data center capacity close to end users to keep latency low.
- Hyperscaler capex is flowing directly into data centers. Google, Microsoft, Amazon, and Meta have collectively announced hundreds of billions of dollars in data center construction for 2026 alone.
Four Ways to Invest in Data Centers
Way 1 Data Center REITs — The Purest Public Market Play
Real Estate Investment Trusts that own and operate data center facilities are the most direct way for retail investors to gain exposure. They earn rent from hyperscalers, cloud providers, and enterprises — and by law must distribute at least 90% of taxable income as dividends.
| Company | Ticker | Scale | 2026 Revenue Guidance |
|---|---|---|---|
| Equinix | EQIX | 300+ data centers, 70+ metros, 3,400+ customers | $10.12B–$10.22B (+10–11% YoY) |
| Digital Realty | DLR | One of the largest global operators; leases to hyperscalers and enterprises | ~$6.6B projected |
| Iron Mountain | IRM | Expanding rapidly from physical records storage into data centers | $7.625B–$7.775B total; $1B+ data center segment |
Equinix reported Q1 2026 revenue of $2.44 billion — double-digit annual growth — with net income jumping 21% year-over-year to $415 million. Adjusted Funds from Operations (AFFO) surpassed $1 billion for the first time in a single quarter. Gross bookings for 2025 were up 27% over 2024, driven by AI-related demand from hyperscalers.
This is not a company benefiting from AI hype. This is a company with signed leases, occupied buildings, and recurring revenue growing at double digits. The AI buildout is showing up directly in its financials.
Way 2 Data Center Builders and Cloud Operators
Beyond REITs, some companies build and operate data centers without the REIT structure. CoreWeave IPO'd in early 2025 and builds GPU-optimized data centers purpose-built for AI workloads. The hyperscalers are the largest data center builders in the world — but data center exposure is embedded inside much larger businesses. Note: DigitalBridge (DBRG), previously a publicly traded digital infrastructure asset manager, has been acquired by SoftBank for $16.00 per share — shareholders approved the deal on April 23, 2026, with the transaction expected to close in the second half of 2026. It is no longer a clean investment option for new buyers.
Way 3 Infrastructure and Components — Inside the Building
Once a data center building exists, it needs to be filled with equipment. These companies supply power systems, cooling, networking, and server racks — and benefit directly from every dollar of data center construction spending.
| Company | Ticker | Role Inside the Data Center |
|---|---|---|
| Vertiv | VRT | Power and thermal management — critical as GPU density increases power load per rack |
| Arista Networks | ANET | High-speed networking switches connecting GPU clusters inside the facility |
| Super Micro | SMCI | Assembles AI server racks housing NVIDIA GPUs — high growth, high volatility |
| Eaton | ETN | Uninterruptible power supply (UPS) and power management for data centers |
| Schneider Electric | SBGSY | Data center power and cooling solutions — French company, accessible via US ADR |
Way 4 Energy and Power Infrastructure — Feeding the Buildings
A modern AI data center can consume 100+ megawatts of power — roughly equivalent to a small city. As GPU density increases, power demand per facility grows with it. The companies that generate and deliver that power are quietly becoming AI infrastructure plays. Constellation Energy signed a 20-year power purchase agreement with Microsoft in September 2024 to restart Three Mile Island Unit 1 — now renamed the Crane Clean Energy Center — with power expected to begin flowing in 2027, a year ahead of the original schedule. In a separate deal, Meta signed an offtake agreement for the entire output of another Constellation nuclear facility, also beginning in 2027. These are not one-off arrangements — they represent a structural shift in how hyperscalers are securing long-term power for AI data centers.
Data Center REITs vs. AI Hardware Stocks — Key Differences
| Factor | Data Center REITs (EQIX, DLR) | AI Hardware (NVDA, MU) |
|---|---|---|
| Volatility | Lower — backed by long-term leases and real assets | Higher — exposed to semiconductor demand cycles |
| Income | Dividend-paying by law (90%+ payout required) | Minimal dividends — return through price appreciation |
| AI upside | Indirect — occupancy rates and rent growth | Direct — every AI model needs their chips |
| Downside protection | Stronger — long lease terms, sticky customers | Weaker — sentiment-driven, fear cycles are sharp |
| Tax treatment | REIT dividends taxed as ordinary income — less favorable | Long-term capital gains treatment available after 1 year |
The CPA Angle: REIT Dividends Are Not the Same as Regular Dividends
Regular stock dividends from companies like Apple or Microsoft may qualify for the 15% or 20% qualified dividend rate. REIT dividends do not. Because REITs distribute income rather than retained earnings, most REIT dividends are taxed as ordinary income — up to 37% federally for higher earners, plus state taxes.
The exception: Under Section 199A, individual investors may deduct up to 20% of qualified REIT dividends, effectively reducing the rate. This deduction has income limitations — consult a tax professional to confirm eligibility.
Because REIT dividends are taxed at ordinary income rates, holding REITs in a taxable brokerage account is the least tax-efficient structure. The better options:
- Roth IRA: All dividends and growth compound tax-free. Qualified withdrawals are never taxed — the most powerful structure for REIT income over time.
- Traditional IRA / 401(k): Dividends grow tax-deferred. You pay ordinary income tax on withdrawal — but defer the tax hit, potentially decades into the future.
- Taxable account: Least efficient for REITs. You owe ordinary income tax on dividends each year, compounding the tax drag on returns.
Some REIT distributions include a return of capital component — a portion of the dividend that is technically a return of your original investment, not income. Return of capital distributions are not taxable in the year received — instead, they reduce your cost basis in the shares, which affects your capital gain calculation when you eventually sell.
This can create a large taxable gain at sale if you are not tracking basis carefully. Your 1099-DIV will classify distributions — always review it before filing and share it with your tax preparer.
Why Data Center Demand Is Structural — Not Cyclical
- AI training and inference demand is compounding. Each new generation of AI models is larger and requires more compute. Every new AI application adds inference volume. The demand curve is not linear — it is accelerating.
- Hyperscaler capex commitments are locked in for years. Data center leases are signed years in advance — the revenue pipeline for REITs like Equinix is highly visible and contracted.
- Supply constraints are real. Data center construction takes 18–36 months. Power permitting is a bottleneck in many markets. The gap between demand and available supply keeps occupancy rates high and supports rent growth.
- The colocation market is growing at 16.5% annually. Projected to reach $72.37 billion by 2030, up from $46.84 billion in 2026 — a five-year expansion of nearly $26 billion in a single market segment.
- Energy constraints are creating moats. Data centers with secured long-term power agreements — particularly nuclear — have a structural advantage over new entrants who face multi-year waits for grid connections in high-demand markets.
How Average Investors Can Get Started
- Start with REITs for income and stability. Equinix and Digital Realty are the two most established publicly traded data center REITs — long operating histories, investment-grade credit ratings, and growing dividends backed by long-term hyperscaler leases.
- Hold REITs in a Roth IRA or IRA — not a taxable account. The ordinary income tax treatment of REIT dividends makes tax-advantaged accounts the right structure for long-term REIT positions.
- Add infrastructure exposure for operating leverage. Vertiv, Arista, and Eaton have more earnings volatility than REITs but more upside as data center construction accelerates.
- Consider energy plays as the contrarian angle. Constellation Energy and Vistra are not obvious data center stocks — but Constellation's 20-year nuclear power deals with both Microsoft and Meta, and Vistra's direct data center energy contracts, make them structural beneficiaries of the same buildout lifting EQIX and DLR.
- Watch the fear cycles here too. Data center stocks sold off sharply during the DeepSeek scare in January 2026, when investors feared AI compute demand would collapse. It did not. Equinix went on to report its best quarter ever in Q1 2026. Fear cycles in data center stocks follow the same pattern as semiconductor stocks — sentiment-driven drops in fundamentally sound businesses.
Final Thought: The Physical Layer of the AI Boom Is Just Getting Started
Every AI model you interact with, every cloud application you use, every piece of data you store — all of it lives in a building somewhere. Those buildings are being built faster than at any point in history, powered by the largest capital expenditure programs in the history of technology.
Data center REITs are one of the few places in the AI investment landscape where you can collect a growing dividend while the AI buildout happens around you — without needing to pick which AI model wins or which chip generation dominates.
The physical infrastructure of the AI economy is being built right now. Average investors have a seat at that table too.
Read next: When Will the AI Boom Be Over? Not Any Time Soon — The Full AI Stack for Average Investors
About the author: Jenny is a CPA with experience in the wealth and asset management industry, valuation, and financial reporting. She writes about practical investing strategies, tax optimization, and long-term wealth building.
Disclaimer: This content is for educational purposes only and not financial advice. Always consult a qualified professional before making investment decisions. The author is a CPA and not a registered investment adviser. CPA credentials relate to accounting and tax matters only. Nothing in this post constitutes advice from a licensed investment professional. References to past stock performance, including specific percentage returns discussed in this post, are historical facts only and are not indicative of future results.
