Hyperscale Data Center Leasing

AI Infrastructure & Compute

Definition

Hyperscale data center leases structure dedicated facilities (20MW-200MW power capacity, 100K-1M+ square feet) for single-tenant AI infrastructure deployments using triple-net lease frameworks: (1) Base rent $80-$150 per kW per month reflecting land, building shell, mechanical systems, (2) Operating expense pass-throughs covering utilities (excluding power), property taxes, insurance, maintenance, and (3) Capital improvement obligations where tenant funds build-out (raised floors, cooling systems, electrical distribution, fiber connectivity) amortized over lease term or absorbed upfront. Lease terms: 10-15 year initial with 2-3 renewal options (5-year increments), creditworthy tenant requirements (investment-grade rating preferred, otherwise guarantees/LC), rent escalations (2-4% annually or CPI-linked), early termination penalties (6-12 months base rent). Power components separate: Tenant executes power purchase agreement directly with utility securing capacity allocation (50MW-200MW) and locking electricity rates ($0.04-$0.12/kWh varying by market), landlord provisions substations and backup generators but tenant pays ongoing power consumption. Development timelines: 18-36 months from contract to delivery (permitting, construction, utility interconnection).

Why it matters

Hyperscale leasing creates $20B+ annual transaction volume as AI infrastructure scaling demands exceed ownership capital. Key economics: Meta/Microsoft/Google each require 1-2GW AI capacity (2025-2027) = 10-20 facilities at 100MW+ each, leasing 40-60% (4-12 facilities per hyperscaler) preserving $5B-$15B capital for GPUs/software versus real estate. Landlord perspective: Data center REITs (Equinix, Digital Realty) invest $300M-$1B per hyperscale facility earning 6-9% unlevered returns on long-term leases to creditworthy tenants, development risk (tenant defaults pre-completion, utility delays) mitigated through pre-leasing (break ground only with signed 10-year lease). Tenant perspective: Locking long-term capacity at predictable costs (rent escalations capped 2-4% annually) hedges AI infrastructure expansion while avoiding technology risk (if workload requirements change years 8-10, can choose not to renew rather than owning obsolete facility). Understanding leasing mechanics critical for: Infrastructure investors evaluating data center REIT valuations (enterprise value multiples 15-25x EBITDA driven by hyperscaler tenant credit quality and lease duration), AI companies planning capacity (owned versus leased trade-offs, lease negotiation leverage points), and developers timing facility construction (oversupply risk if multiple hyperscale facilities deliver simultaneously in same market compressing rents 20-40%).

Common misconceptions

  • Triple-net doesn't mean tenant pays all costs—landlord retains: building structure/roof repairs (major capital), property management overhead, lease commissions, debt service. Tenant pays: interior systems, utilities, operational expenses. Allocation disputes common—HVAC replacement capital or expense? Depends on specific lease terms.
  • Power capacity isn't guaranteed in perpetuity—utility commits to deliver specified capacity but reserves right to interrupt during extreme events (heatwaves, grid emergencies). Critical loads require on-site generation (diesel/natural gas backup) adding $50M-$150M capital costs. Some markets (Texas ERCOT) facing reliability concerns reducing facility valuations 10-20%.
  • Lease renewals aren't automatic—at option exercise (years 10, 15, 20), landlord can adjust rent to market rates if current rent below market by 20%+. Prevents tenants locking below-market rent indefinitely. Creates renegotiation risk—tenant invested $200M+ in facility improvements, landlord has leverage demanding rent increases or tenant faces costly relocation.
  • Announced capacity is not the same as economically usable capacity. Power availability, interconnection timing, chip allocation, cooling design, customer utilization, and contract duration all affect whether an AI infrastructure asset converts technical capacity into durable cash flow.

Technical details

Lease structure and financial terms

Base rent calculations: Per-kW pricing: $80-$150/kW/month typical range. Calculation: 50MW facility × 1,000 kW/MW × $100/kW = $5M monthly base rent = $60M annual. Geographic variance: Northern Virginia $120-$150/kW (premium market, high demand). Phoenix/Dallas $90-$120/kW (growth markets, moderate demand). Iowa/Nebraska $70-$90/kW (tier-2 markets, lower costs). Rent escalations: Fixed 2-4% annually, or CPI-indexed with caps (e.g., min 2%, max 5%), prevents hyperinflation exposure while providing landlord protection against costs.

Operating expense pass-throughs: Property taxes: $5-$15/kW/month (varies dramatically by jurisdiction). Insurance: $2-$5/kW/month (all-risk property, earthquake where applicable). Common area maintenance: $3-$8/kW/month (landscaping, security, access roads). Reconciliation: Annual true-up comparing actual expenses to estimates, tenant pays or receives credit for variance. Caps: Some leases cap expense increases at 5-7% annually protecting tenants from cost explosions.

Tenant improvement allowances: Landlord contribution: $50-$150 per square foot toward tenant build-out (electrical distribution, cooling systems, security). Amortization: Added to base rent over 10-15 year lease term at 6-9% interest rate. Example: $100M TI allowance, 10-year amortization, 7% rate = $14.2M annual rent add ($1.18M monthly). Tenant pays upfront: Advanced cooling (liquid vs air), premium finishes, redundant systems beyond landlord minimums. Typical total tenant investment: $200M-$500M for hyperscale facility.

Security deposits and guarantees: Cash deposits: 3-6 months base rent held as security. For $5M monthly rent = $15M-$30M deposit. Alternatives: Letter of credit from bank (annual fee 1-2% of LC value, preserves cash). Parent guarantee: Corporate parent guarantees subsidiary tenant obligations (common for startup AI infrastructure companies backed by large tech firms). Step-down provisions: After 3-5 years of timely payments, reduce LC from 6 months to 3 months, or release entirely if tenant achieves investment-grade rating.

Power and utility considerations

Utility capacity allocation: Dedicated substation: For 50MW+ facilities, utility builds dedicated substation and transmission lines. Cost: $20M-$80M (sometimes tenant-funded, sometimes utility-funded passed through rates). Timeline: 18-36 months from approval to energization. Capacity commitment: Utility reserves power capacity for facility—if tenant delays/cancels, may owe liquidated damages ($5M-$20M) compensating utility for stranded investment.

Power purchase agreement structures: Rate structures: Fixed rate: $0.06-$0.10/kWh locked for 10-20 years providing cost certainty. Index rate: Tracks utility published rates with negotiated discount (10-20% below retail). Hybrid: Fixed base + commodity pass-through (separate energy from distribution). Minimum take obligations: Tenant must purchase 60-80% of reserved capacity annually or pay shortfall penalties. Prevents speculative capacity hoarding. Renewable energy options: Physical delivery (utility builds dedicated solar/wind farm delivering to facility). Virtual PPAs (financial hedge—tenant pays utility variable rate, receives renewable energy credits offsetting costs).

Backup power and reliability: On-site generation: Diesel or natural gas generators providing 2N redundancy (double capacity needed, either can fail). Cost: $30M-$100M for 50MW facility. Uninterruptible power supply (UPS): Battery systems bridging 10-30 second gap between grid failure and generator startup. Cost: $10M-$30M. Fuel storage: 24-72 hour on-site diesel storage (500K-2M gallons), permits and environmental compliance expensive in urban markets. Annual testing: Monthly generator tests, annual full-load exercises, maintenance contracts $1M-$3M annually.

Utility interconnection risks: Delivery failures: If utility cannot deliver promised capacity (transformer shortages, permitting delays, grid constraints), tenant may terminate lease without penalty or receive rent credits during delay period. Historical precedent: Northern Virginia 2022-2024 saw 12-24 month delays on multiple projects as Dominion Energy struggled with demand surge. Mitigation: Tenants negotiate detailed milestone schedules with liquidated damages for utility delays ($100K-$500K per month), right to terminate if delay exceeds 6-12 months.

Lease obligations and risk allocation

Maintenance and capital responsibilities: Landlord obligations: Structural integrity (roof, exterior walls, foundation). Common infrastructure (access roads, parking, perimeter security). Major capital replacements (HVAC chillers end-of-life after 15-20 years). Tenant obligations: Interior systems (electrical distribution, cooling within facility, backup power systems). All operational maintenance (filters, repairs, testing). Technology upgrades and capacity expansions. Gray areas: HVAC replacement—capital (landlord) or maintenance (tenant)? Disputes common, requires detailed lease definitions.

Insurance and liability: Property insurance: Tenant maintains all-risk coverage on tenant improvements ($200M-$500M value). Landlord maintains coverage on building shell. Business interruption: Tenant carries business interruption insurance covering lost revenue if facility inoperable (12-24 month coverage typical). Liability insurance: General liability $10M-$50M per occurrence, cyber liability $25M-$100M (covering data breaches), umbrella coverage $50M-$250M (excess of underlying). Indemnification: Tenant indemnifies landlord for tenant operations, landlord indemnifies tenant for building defects/landlord negligence. Mutual waiver of subrogation—insurers cannot sue opposing party after paying claim.

Default and remedies: Tenant defaults: Non-payment (30-day cure period), bankruptcy (automatic default in most leases), covenant violations (operating facility outside permitted uses). Landlord remedies: Termination and eviction (3-6 month process), damages (unpaid rent through lease term discounted to present value), mitigation duty (landlord must attempt to re-lease reducing damages). Landlord defaults: Failure to deliver facility on time, utility capacity not delivered, major building defects. Tenant remedies: Rent abatement (reduce payment during issue), termination rights (if defect uncured for 6-12 months), damages (tenant relocation costs, business interruption).

Assignment and subletting restrictions: Assignment restrictions: Tenant cannot assign lease without landlord consent (not to be unreasonably withheld). Landlord evaluates: Assignee creditworthiness (must meet or exceed tenant credit quality), use (cannot change from data center to general warehouse), strategic concerns (landlord may block assignment to competitor). Subletting prohibitions: Most leases prohibit subleasing entirely or cap at 10-20% of facility. Prevents tenant from becoming de facto data center operator competing with landlord. Exceptions: Transfers to affiliates (parent/subsidiary) permitted, assignments in merger/acquisition approved if surviving entity creditworthy.

Market dynamics and investment considerations

Supply-demand dynamics: Current supply: US data center market 5-10GW capacity, adding 1-2GW annually. Hyperscale demand: Meta, Microsoft, Google, Amazon each planning 1-2GW AI capacity 2025-2027 = 4-8GW combined. Market tightness: Constrained markets (Northern Virginia, Silicon Valley) seeing 2-4 year waitlists, pre-leasing driving speculative development. Oversupply risk: 2026-2027 delivery wave of 15-20 large facilities in Phoenix/Dallas could exceed demand compressing rents 20-40% if macro slowdown reduces AI capex.

Landlord competitive positioning: Public REITs (Equinix $70B market cap, Digital Realty $50B): Investment-grade balance sheets enabling $1B+ facility development, established hyperscaler relationships, diversified portfolios (100+ facilities) reducing single-tenant concentration. Private developers (QTS, CyrusOne, EdgeCore): Lower cost of capital through institutional LP backing (Blackstone, KKR, GIC), faster decision-making, customization flexibility. Owner-operators (Aligned, Stream): Vertical integration (land, development, operations), margins 10-20% higher but concentrated risks.

Development returns and exit strategies: Pro forma returns: Development cost: $800M-$1.2B for 100MW facility ($8M-$12M per MW). Annual NOI: Rent $60M-$90M - OpEx $10M-$15M = $50M-$75M stabilized NOI. Unlevered yield: 6-9% ($50M-$75M / $800M-$1.2B). Levered returns: 60% LTV debt at 6% = 12-18% equity returns. Exit multiples: Stabilized facilities with creditworthy 10-year leases trade at 15-25x EBITDA ($750M-$1.8B on $50M-$75M NOI) providing developers 50-100% profit on cost after 3-5 years.

Tenant negotiation leverage: Strong tenant leverage: Creditworthy (Meta, Microsoft, Google investment-grade rated), multi-facility pipeline (leasing 3-5 facilities worth $3B+ over 5 years creates volume discounts), competitive markets (landlords bidding for tenant creating 10-20% rent reductions). Weak tenant leverage: Non-investment grade startups (CoreWeave pre-debt raise), single facility (no volume discounts), monopolistic markets (Northern Virginia single dominant landlord controlling best sites), urgent timeline (needed capacity in 12 months versus typical 24-36 month development preventing shopping alternatives).

How capacity becomes underwritable revenue

AI infrastructure investors generally underwrite three linked layers: physical readiness, customer demand, and technology refresh risk. A site, GPU cluster, or hosting contract may be valuable only if power, cooling, networking, chip delivery, customer onboarding, and utilization arrive in the same window.

Monitoring therefore combines operating metrics with contract review: megawatts energized, GPUs delivered and accepted, utilization by customer, committed versus burst usage, counterparty credit, termination rights, upgrade obligations, and residual value assumptions. The strongest structures make those dependencies explicit instead of relying on broad AI-demand narratives.

Related Terms

See in context