1 Common development play‑books
-
Ground‑up “merchant” or powered‑shell build. Developer secures land, power and zoning, then delivers a shell with enough critical infrastructure to attract pre‑leases. Construction budgets in the U.S. now run ≈ US $10‑16 million per delivered MW depending on the market and density.
-
Build‑to‑suit / forward funding. A hyperscale or colocation tenant signs a long lease before or during construction; the lease supports cheaper debt and lowers risk.
-
Sale‑leaseback (corporate carve‑out). Enterprises unlock capital by selling their owned facilities and leasing them back; investors get a “bond‑like” income stream. Current exit pricing is around 5 % cap rates for powered‑shell assets and mid‑6 % for turnkey facilities.
-
Platform roll‑up. Private‑equity or infrastructure funds buy regional operators, add capacity, and target an IPO or strategic sale once the platform reaches scale.
-
Edge / micro‑DC pods. Small modular units placed near population clusters, often with 3‑ to 5‑year flip horizons once revenue density is proven.
2 Hold‑period and exit assumptions investors will accept
-
Merchant ground‑up: underwrite a 5‑7‑year hold—long enough to finish the build, lease to ±70 % occupancy and sell into a core fund; exits are usually modelled at US $9‑11 m per MW in tier‑one markets.
-
Build‑to‑suit: assume 7‑10 years so the first rent escalations are visible; valuation is on forward NOI or $/MW benchmarks similar to merchant deals.
-
Platform strategy: GPs often plan 10 + years and, if LPs want liquidity, move stabilized assets into a continuation fund.
-
Edge pods: 3‑5 years is typical because value is tied to achieving revenue run‑rate rather than long‑term lease cash‑flows.
(Institutional LPs recognize these ranges as “market” but will still stress‑test exit cap‑rates ±100 bp.)
3 Where the money comes from
-
Senior construction loans: still available at ≈ 65‑80 % loan‑to‑cost (LTC) with spreads of SOFR + 200‑275 bp when there is a credit‑rated anchor tenant.
-
Hold‑co / portfolio term debt and ABS: the asset‑backed‑securities market is expanding fast—outstanding data‑center ABS is projected to reach ≈ US $50 bn by 2027, double 2024 levels.
-
Preferred equity / mezzanine: 10‑12 % cash‑pay coupons plus 2‑3 % PIK tails are common for bridging big equity cheques.
-
Common equity: infrastructure funds, sovereign wealth and insurers look for 12‑15 % unlevered IRRs on core‑plus risk.
-
Tenant “capacity deposits.” Some AI builders now pre‑pay for power reservations—effectively giving the developer an interest‑free slice of the capital stack.
4 What makes a good location in 2025
-
Power you can actually energize. Interconnection queues stretch 4‑8 years in advanced economies and may delay ≈ 20 % of planned projects; a site with interim on‑site generation or a fast‑tracked utility agreement is worth a premium.
-
Robust fiber and low latency. Proximity to long‑haul and metro fiber paths (e.g., Northern Virginia or Chicago) still dictates where hyperscalers sign first.
-
Climate & cooling economics. Cooler climates (Nordics, Dublin) and access to non‑potable water cut operating expenses and improve ESG scores.
-
Renewable‑energy access and government incentives. States that combine large renewable portfolios with tax abatements—for example, Virginia sales‑tax relief on equipment—stay on short‑lists.
-
Regulatory stability & data‑sovereignty rules. EU GDPR, China MLPS and similar regimes require in‑country storage; paired “urban + rural” campuses are an emerging workaround.
-
Natural‑hazard profile and land economics. Investors discount high‑risk zones. In the U.S. the average land price is about US $5.59 / sq ft, but parcels of ≥ 50 acres now cost 23 % more and average 224 ac in size as developers build multi‑building campuses.
-
Local labour and construction ecosystem. A deep pool of specialized trades can shave months off the schedule and limit cost escalation.
-
Demand/supply balance. North‑American colocation vacancy hit a record 2.6 % in late 2024, with 72 % of the 6.5 GW pipeline pre‑leased—evidence that new capacity in power‑constrained markets fills quickly.
5 Why all of this matters now
AI workloads are exploding. The IEA projects that data‑center electricity demand will more than double to ~945 TWh by 2030, driven largely by AI training and inference. That demand is colliding with power‑grid bottlenecks, tight vacancy and rising construction costs—so investors, lenders and tenants have become extremely picky about underwriting assumptions.
Use this checklist when you build your next model:
-
Confirm energization timeline and price of power.
-
Stress‑test a grid‑delay scenario (+24 months) in the cash flow.
-
Cap senior debt at 70 % LTC unless you have a credit‑wrapped lease.
-
Underwrite exit cap‑rates 50‑100 bp wide to show resilience.
-
Score sites on power 30 % / fiber 20 % / incentives 15 % / climate 15 % / land cost 10 % / hazard 10 %.
Keeping the story this clear—and tied to current market evidence—will resonate with credit committees and LP investment boards alike.
You can hire me to build a custom financial model here.
Check out more financial model templates.
Article found in General Industry.