AIccelerate Deep Dive · February 26, 2026 · 25 min read

Data Centres, Energy
& REITs

The $7 Trillion AI Infrastructure Supercycle

AI's insatiable demand for compute is triggering the largest infrastructure buildout in history. We identify 10 winners across data centre REITs, nuclear energy, and power infrastructure positioned to capture decades of structural growth.

Infrastructure Spend
$7T
projected investment by 2030
McKinsey / Morgan Lewis
Power Demand 2030
134.4GW
from 50.5 GW in 2024
S&P Global / 451 Research
Consumption Growth
165%
power consumption rise by 2030
Goldman Sachs
Hyperscaler Spend
$700B
AI build-outs in 2026
Kavout / Morningstar

US Data Centre Power Demand Trajectory

Grid demand projected to nearly triple by 2030

0 35 70 105 140 GW 50.5 GW 2024 61.8 GW 2025 75.8 GW 2026 108 GW 2028 134.4 GW 2030 +166% growth
Source: S&P Global / 451 Research

The Power Bottleneck

How AI demand explosion creates investment opportunities across the value chain

AI Demand Explosion 122% CAGR inference Power Grid Strain 165% demand growth 4+ Year Wait Times 40% DC constrained '27 Solutions ☢ Nuclear Power 92.5% capacity factor 24/7 zero-emission baseload ⚡ Renewables Clean energy PPAs ESG + scale advantage 🔋 Fuel Cells Speed to Power Deploy months, not years 🏢 Metro REITs Inference advantage Low-latency, interconnected CEG TLN VST NEE BE EQIX DLR Infrastructure Layer — Powers Everything Above VRT Cooling & Power ETN Electrical Mgmt

Energy Source Capacity Factors

Why nuclear is the gold standard for 24/7 AI workloads

Nuclear 92.5% Gas 56% Wind 35% Solar 25%

Why nuclear wins: AI workloads run 24/7/365. Nuclear's 92.5% capacity factor means near-constant output — no intermittency, no storage needed. It's the only source that matches AI's always-on demand profile at scale.

The Inference Shift

AI workloads are moving from training to inference — reshaping who wins

🏋️

Training Era

Massive GPU clusters, rural campuses

  • Hyperscale campuses in remote locations
  • Latency tolerance (batch processing)
  • Few massive clusters
  • Hyperscaler-dominated demand
🚀

Inference Era NOW

Low-latency, metro-market, enterprise

  • Metro locations near population centres
  • Ultra-low latency (real-time responses)
  • Distributed across many sites
  • Enterprise AI adoption accelerating
60%

of Equinix's Q4 largest deals were driven by AI workloads — and half came from enterprise customers, not hyperscalers. The inference shift massively favours metro-market REITs like EQIX and DLR.

10 Winners Across the Value Chain

Positioned for decades of structural growth in the AI infrastructure supercycle

EQIX
Equinix
HIGH
Price
$960.52
Mkt Cap
$94.4B
Revenue
$9.22B +5.4%
EBITDA
$4.14B +20.4%

Dominant metro-market moat with 273 data centres across 77 markets. AI inference requires exactly what EQIX offers — low latency, cloud proximity, and network interconnection at scale.

KEY CATALYSTS
  • ▸ 3 GW powered land banked in constrained markets
  • ▸ Q4 bookings up 42% YoY; 52 projects underway
  • ▸ Enterprise AI adoption accelerating
View on Perplexity Finance →
DLR
Digital Realty
HIGH
Price
$179.59
Mkt Cap
$61.7B
Revenue
$6.11B +10%
Net Income
+117.2%

Cloud zone markets becoming top priority for hyperscalers. DLR uniquely positioned with capacity + power in constrained markets, taking more development risk for faster delivery.

KEY CATALYSTS
  • ▸ 5 GW powered land banked
  • ▸ Record earnings trajectory
  • ▸ Rate cut tailwinds for REIT valuations
View on Perplexity Finance →
IRM
Iron Mountain
MEDIUM
Price
$111.78
Mkt Cap
$33.1B
Revenue
$6.90B +12.2%
EBITDA
$2.08B +9.3%

Underfollowed transformation from document storage to AI-era data centre powerhouse. Massive global real estate footprint converting to data centres — higher risk, higher reward.

KEY CATALYSTS
  • ▸ $2.27B CapEx — aggressive DC buildout
  • ▸ Data centre segment fastest-growing
  • ▸ Transformation narrative gaining traction
View on Perplexity Finance →

Risk / Reward Matrix

Positioning our 10 picks on risk vs. upside potential

RISK LEVEL → UPSIDE POTENTIAL → HIGH UPSIDE / LOWER RISK HIGH UPSIDE / HIGHER RISK MODERATE / LOWER RISK MODERATE / HIGHEST RISK EQIX CEG NEE ETN VRT VST TLN DLR BE IRM REITs Energy Infrastructure

Key Risks

The bear case: what could derail the supercycle thesis

Grid Constraints

Power grid infrastructure may not expand fast enough. 4+ year wait times could slow deployment significantly, particularly in primary markets like Northern Virginia.

📋

Regulatory Pushback

Growing political pressure over electricity costs for consumers. The GRID Act and similar legislation could cap data centre power consumption or impose additional costs.

🧠

AI Efficiency Gains

Breakthroughs in AI model efficiency (like DeepSeek) could reduce power demand growth. More efficient chips and architectures may lower infrastructure needs over time.

📈

Interest Rate Risk

Rising rates pressure REIT valuations through higher cost of capital and reduced attractiveness of dividend yields relative to bonds. Affects EQIX, DLR, IRM most.

🏗️

Construction Overruns

$60B+ in data centre construction starts in 2025 alone. Supply chain bottlenecks, labour shortages, and cost escalation could compress margins and delay timelines.

🏢

Hyperscaler Self-Builds

Meta, Google, and others are building their own data centres. If hyperscalers vertical-integrate further, demand for third-party REITs and infrastructure could soften.

Key Takeaways

1

The supercycle is structural, not cyclical

$7 trillion in infrastructure investment is committed. With 122% CAGR in AI inference demand and $700B in hyperscaler spend in 2026 alone, this is a multi-decade buildout — not a hype cycle. Power demand tripling by 2030 creates irreversible momentum.

2

Power is the new moat

With 4+ year grid connection wait times and 40% of AI data centres potentially constrained by 2027, companies with secured power access — nuclear (CEG, VST, TLN), renewables (NEE), and speed-to-power solutions (BE) — hold the strategic advantage.

3

The inference shift changes the winners

AI moving from training to inference structurally favours metro-market REITs (EQIX, DLR) over rural hyperscale campuses. Low latency, multi-cloud interconnection, and enterprise demand make urban network hubs irreplaceable.

4

The picks-and-shovels layer is pure leverage

VRT (+168.8% NI growth) and ETN (Nvidia 800V partnership) don't depend on any single customer or geography. Every data centre built — by anyone — needs cooling, power distribution, and electrical management. They are the infrastructure tax on the entire supercycle.

Sources

Disclaimer: This analysis is for informational purposes only and does not constitute financial advice. All data and projections are sourced from publicly available research and may be subject to change. Stock prices and financial metrics reflect the most recent available data at time of publication (February 26, 2026). Past performance is not indicative of future results. Always do your own research before making investment decisions.