Data Centers Are Becoming the New Core Asset Class for CRE and Builders
Forecasts point to up to $1 trillion in North American data-center builds from 2025 through 2030. U.S. Census data suggests data-center construction could overtake office spend as early as next year.
That means commercial real estate capital is increasingly tied to AI-driven compute demand. The upside is clear. So is the concentration risk if AI usage or compute growth cools.
Why money is shifting
Compute needs keep climbing-training, inference, cloud, and streaming all need space, power, and cooling. Hyperscalers and major colocations want speed to market, long-term capacity, and low-risk delivery.
For investors, the draw is long leases, sticky tenants, and infrastructure-like cash flows. For builders, it's repeatable templates, bigger MEP scopes, and premium scheduling fees-if you can secure power and gear.
Hard constraints that decide winners
- Power first: Substation proximity, available MW, interconnection queues, and realistic energization dates. Transformer and switchgear lead times can set your schedule-plan early.
- Cooling approach: Air vs. liquid, water rights, and discharge permits. Heat density is rising; know your WUE targets and what local utilities will approve.
- Fiber and latency: Proximity to major network routes and on-ramps. Dual-path redundancy should be designed in from day one.
- Zoning and community: Diesel gensets, battery systems, and 24/7 operations draw attention. Pre-wire your outreach and noise/traffic studies.
- Incentives: Sales/use tax exemptions, industrial revenue bonds, and job credits can swing your pro forma. Get a specialist to model them early.
- Supply chain: Switchgear, generators, chillers, and PDUs are bottlenecks. Lock capacity with manufacturers and use alternates that are pre-approved.
- Water and sustainability: Many municipalities are tightening water use and emission rules. Design toward dry or hybrid cooling where possible.
Investor notes: underwriting that holds up
- Lease structure: Pre-lease with take-or-pay and cost pass-throughs where you can. Confirm escalation mechanics on both rent and power.
- Demand scenarios: Model delayed AI deployments, tenant consolidation, and slower compute refresh cycles. Stress test DSCR and exit cap rates.
- Capex discipline: Set contingencies for MEP inflation and grid delays. Tie EPC fees to milestones with meaningful liquidated damages.
- Power pricing: Hedge with PPAs or block-and-index strategies. Make sure the lease allows recovery of volatility.
- Phasing: Stage capacity in modular blocks to reduce exposure if take-up slows.
Build tactics that hit dates (and margins)
- Standardize a repeatable "kit of parts" with modular MEP skids and pre-tested controls.
- Secure long-lead equipment before final design-use provisional sums and pre-negotiated alternates.
- Run permitting, site work, and equipment procurement in parallel with a well-defined long-lead strategy.
- Commission in tiers with clear hold points; don't let punch lists bleed into operations.
- Design for future density: raised floors vs. slab, hot/cold aisle containment, and liquid-ready pathways.
Site selection: where the math works
- Established hubs: Northern Virginia, Dallas, Phoenix, Atlanta, Chicago-deep ecosystems and labor pools but tougher power queues.
- Emerging nodes: Columbus, Kansas City, Reno, Montreal/Toronto-friendlier power and land costs, growing fiber routes.
- Energy profile: Grid mix, reliability, curtailment risk, and future rates matter as much as current tariffs.
- Resilience: Floodplains, fire risk, and extreme heat days. Insurance and backup strategies must be priced in.
Risk if AI demand cools
- Shift from single-tenant to multi-tenant colo to broaden base demand.
- Design shells for flexible hall sizes and scalable power paths.
- Protect downside with staged delivery, termination fees, and diversified tenant mix (AI, cloud, network, enterprise).
- Know your secondary use plan: labs, HPC, disaster recovery, or partial industrial conversion.
Quick checklist for your next data-center deal
- Verified interconnection path, energization date, and substation capacity in writing.
- Locked transformer/switchgear slots with penalties for slippage.
- Documented cooling strategy with water use and discharge permits identified.
- Fiber routes, carriers on letterhead, and dual-entry plans approved.
- Entitlements mapped with a community plan for gensets and 24/7 operations.
- Incentive term sheets baked into the capital stack.
- EPC contract with milestone-based payments and performance LDs.
- Commissioning agent retained from schematic design.
- Lease terms that pass through power volatility and include step-ups.
- Downside models and exit strategies reviewed at IC-before money moves.
Data points and further reading
Track construction spend trends directly from the source: U.S. Census Bureau Construction Spending. For grid planning and load growth context, this is useful: U.S. Energy Information Administration.
Build your team's AI fluency
If you lead capital planning, underwriting, or delivery, give your team a working grasp of AI demand drivers, model lifecycles, and how they translate into MW, density, and lease terms. Curated learning for job roles is here: Complete AI Training: Courses by Job.
The next cycle will reward speed, predictable delivery, and clear risk controls. Pick markets where the power is real, standardize your build, and only scale as fast as your interconnection and gear will allow.
Your membership also unlocks:
