In partnership with
Every ChatGPT query needs to run somewhere. Every AI model training session needs physical server space. And right now, there isn't nearly enough of it.
The AI boom isn't just about chips, it's also about where you put them, how you cool them, and whether you can get enough electricity to keep them running. Data center REITs like Equinix (EQIX) and Digital Realty (DLR) have become compelling infrastructure plays because of this.
Goldman Sachs projects data center power demand will jump 165% by 2030. But the bottleneck isn't money or demand—it's electricity. You can't build if you can't secure hundreds of megawatts, and that's reshaping the sector.
AI Broke the Old Data Center Model
Traditional data centers handled cloud storage, email, and predictable loads. But AI training requires a bit more. And by ‘bit more’ we mean as much power as a small city would use, to run thousands of GPUs at full blast for weeks. AI workloads consume 10-20x more power per square foot than traditional computing.
Modern AI data centers need cooling systems that barely existed five years ago, electrical setups handling 150 kilowatts per rack (vs. the old 5-10kW), and direct power substation connections. DCREITS are more than mere real estate assets—they're energy infrastructure with buildings attached.
The Numbers Are Staggering
Data center power use is projected to hit 92 gigawatts by 2027, up from 35 GW in 2023—roughly South Korea's current power capacity.
Big tech has tripled AI infrastructure spending since 2021. The Magnificent Seven is expected to spend $400 to $600 billion on data center capex in 2025. Occupancy rates sit at 90-95% despite new construction everywhere. Facilities opening in 2026-2027 are already pre-leased with 10-15-year commitments.
That's why these data centers trade more like utilities than REITs now.
Power Is the Bottleneck
Land and capital are real constraints. But electricity tops them. Modern AI data centers need 500+ MW, equivalent to a mid-sized power plant. Location scouting now prioritizes power availability, then anything else.
If you can't secure 300+ MW of reliable power, the project doesn't happen. This split the sector fast. Established operators with utility relationships dominate. New players face years just getting grid approvals. Here is what the key players in the DCREITs are up to
Table 1: Key Strategic Investments and Commitments to Digital Infrastructure Enhancement
Company | Focus Area | Investment / Capacity | Timeline | Strategic Objective |
|---|---|---|---|---|
Equinix (EQIX) | Data Center Expansion | $4–$5B CAPEX through 2029 | 2025–2029 | Double data center capacity and strengthen AI hosting readiness. |
AI Infrastructure | 150 kW/rack density | Ongoing | Support high-density AI and HPC workloads. | |
Renewable Power | 1.2 GW PPAs (≈96% of total) | Since 2015 | Achieve 100% renewable energy use across facilities. | |
Digital Realty (DLR) | Capacity Growth | $3–$3.5B CAPEX; 499 MW added | Ongoing | Expand hyperscale and colocation footprint globally. |
Advanced Cooling | AALC / DLC systems deployed | Active | Improve cooling efficiency for AI workloads. | |
Green Financing | $7.2B raised via Green Bonds (1.5 GW projects) | Ongoing | Fund low-carbon and energy-efficient operations. | |
American Electric Power (AEP) | Grid Modernization | $1.7B investment | Through 2029 | Upgrade transmission lines to enhance grid reliability. |
On-Site Power | Up to 1 GW fuel cell rollout | Accelerated | Provide resilient power for data center clusters. | |
NextEra Energy (NEE) | Renewable Development | 6 GW active; 10+ GW pipeline | Ongoing | Supply sustainable power to data center and AI hubs. |
Infrastructure Buildout | $120B CAPEX plan | 2025–2029 | Expand grid and generation capacity to meet AI-driven demand. |
Strong Through 2027, Then Watch Out
The near-term case is pretty clear. Demand is high, supply is tight, occupancy is almost full, and pricing is strong. Growth is about 17% annually through 2028.
But massive new capacity comes online between 2026 and 2028. If hyperscaler spending slows or AI efficiency improves faster than expected, the market could flip from undersupply to oversupply around 2027. That means shifting from scarcity-driven pricing to normal competition where power access matters more than capacity additions.
Best-positioned companies have long-term leases with solid tenants, their own power generation or utility partnerships, and presence across power-constrained markets. The risk is operators who grabbed land but can't deliver power, or built in weak markets—those projects stall or launch right when pricing gets tough.
Key Performance Indicators to Watch
What | Where It Is Now | Why It Matters |
|---|---|---|
Occupancy Rate | 90-95% | Shows how tight supply is; pricing power stays strong |
Pre-Leasing | Record high through 2027 | Forward visibility on demand |
Power Access | Critical bottleneck | Limits new supply; protects existing players |
Hyperscaler Spending | $385-$598 billion in 2025 | Shows sustained demand |
New Supply Coming | ~30% capacity increase by 2027 | Could moderate pricing after 2027 |
The sector is still growing structurally, but timing matters. 2025-2027 looks like the sweet spot where supply is tight. After that, you want to focus on operators with real competitive advantages, not just ride the overall trend.
Who Wins Long-Term
The sector will split. Winners are established operators with power locked down, hyperscaler relationships, and technical capability—companies like Equinix and Digital Realty that can execute at scale.
Losers are late entrants without power, operators in weak markets, and anyone just betting on building more without differentiation.
Wild card: new players like Fermi America targeting an 11 GW power campus for AI. If they execute, they disrupt. If not, a cautionary tale.
The shift is from real estate to energy infrastructure. Companies that recognize this will capture value. Those treating it like normal real estate will struggle as the market matures. AI demand is real, but the gold rush phase comes with an expiration date.
Crypto’s Most Influential Event
This May 5-7 in 2026, Consensus will bring the largest crypto conference in the Americas to Miami’s electric epicenter of finance, technology, and culture.
Celebrated as ‘The Super Bowl of Blockchain’, Consensus Miami will gather 20,000 industry leaders, investors, and executives from across finance, Web3, and AI for three days of market-moving intel, meaningful connections, and accelerated business growth.
Ready to invest in what’s next? Consensus is your best bet to unlock the future, get deals done, and party with purpose. You can’t afford to miss it.
