How do giant data-center leases become shareholder cash?

Data-center leases are growing in scale. Applied Digital contracted 400MW for CoreWeave; Core Scientific disclosed approximately 590MW of projects for the same tenant; VNET announced a 510MW order from an unnamed leading internet customer in May 2026.^[1][2][5] Customers willing to reserve resources in advance give developers a basis for construction.

A long contract does not automatically turn today’s construction payments into shareholder profit. Developers fund land, equipment and building work before tenants generally begin paying for accepted capacity. The investment value depends on the length of that funding interval, the allocation of costs and the tenant’s rights when delivery fails.

Megawatts alone cannot rank the contracts

Applied Digital’s April presentation described Polaris Forge 1 as 400MW of critical IT load, with an approximately fifteen-year base term and $11 billion of contracted revenue excluding pass-through power. At that date, 100MW was energized and 300MW remained under construction.^[1] It is a dated development snapshot, not evidence that the entire campus was delivered by August.

Core Scientific’s April filing described approximately 590MW of critical IT capacity, with more than 185MW billable as of March 13. Most contracts have twelve-year initial terms, but Austin has seven years.^[2] Both developers depend on CoreWeave. Multiple campuses do not diversify the ultimate tenant’s credit risk.

Digital Realty’s three Northern Virginia facilities each contain 96MW of IT capacity and are leased to three distinct investment-grade hyperscale customers for fifteen years, with 3.6% annual escalators. VNET’s 510MW order came from an unnamed leading internet customer in Greater Beijing.^[4][5] Anonymity does not imply weak credit, but it also provides no basis for assigning an identity.

The four disclosures sum to roughly 1.788GW only as an indication of transaction scale. Their dates, capacity definitions and delivery stages differ. They cannot be treated as contemporaneous operating capacity. Comparing them directly with a power station’s output is also imprecise: IT load requires additional cooling and distribution power, and campuses need not all operate at full load.

Revenue density must be compared with capital required

Dividing Applied Digital’s $11 billion by fifteen years and 400MW gives approximately $1.83 million per MW per year, or $153 per kW per month. This is a simple average of nominal contract revenue across years and ramp periods. It is not the initial rent, current revenue or project internal rate of return.^[1]

The company estimated construction capex at $11 million–$13 million per MW.^[1] Dividing average annual contract revenue by that cost gives a gross revenue-to-construction-cost ratio of about 14.1%–16.7%. Operating expenses, maintenance, corporate overhead, tax, financing and the timing of construction cash flows are absent. The ratio is not a shareholder return.

The rough calculation still identifies useful sensitivities. A 10% construction overrun with unchanged revenue reduces that ratio by about 9.1%. Tenant funding can reduce the developer’s own invested capital. The same rental income can therefore support quite different equity economics.

Comparisons should establish who owns the building, buys the servers and receives cost reimbursement. Colocation revenue per MW generally excludes GPU services. Comparing it directly with a cloud operator that owns accelerators confuses different assets and services. Lower revenue density can reflect less capital at risk rather than an inferior return.

Minimum payments do not remove delivery obligations

Minimum payments can reduce vacancy risk, but the agreement defines the protection. Core Scientific disclosed service credits for specified power and environmental failures, together with termination rights for repeated or prolonged failures and other events.^[2] The developer must continue providing an acceptable service throughout the long contract.

Three situations deserve separate treatment: a tenant receives compliant capacity but does not use it; the developer fails to deliver; or the parties agree to reschedule. Minimum payments are most helpful in the first case. The other cases depend on contractual terms. The phrase take-or-pay cannot substitute for reading them.

Credit support also differs in strength. A project-company promise, parent guarantee and bank letter of credit are not interchangeable. Deposits may cover only part of the remaining obligation. A strong tenant helps, but a longer lease also spans more possible changes in financing conditions, technology and the tenant’s own business.

Concentration influences bargaining power as well as default exposure. One tenant occupying most development capacity may request schedule changes or technical upgrades across several campuses at once. Real diversification concerns paying entities and their cash sources, not merely the number of locations on a map.

Tenant funding reduces the initial burden and future collections

Core Scientific’s project disclosure describes customer-funded construction accompanied by partial future revenue credits capped at $1.5 million per MW.^[2] The arrangement can reduce cash that the developer must raise during construction. Future rent, however, is reduced by those credits. Customer funding cannot be counted as free capital while all nominal rent is treated as distributable cash.

Consider a simplified $100 million tenant contribution repaid through $10 million of annual rent credits for ten years. The developer raises $100 million less initially but collects $10 million less each year. The arrangement’s value depends on funding costs, the start of credits and contractual risk, rather than the increase in assets alone.

A completion guarantee may return project risk to the parent. Core Scientific’s filing requires the parent to provide additional funds for specified projects if available funding proves insufficient.^[2] Project-level financing does not necessarily isolate the parent’s shareholders from construction overruns.

A useful lease model therefore identifies the total budget, secured funding, remaining gap, tenant contributions and future offsets. Debt and preferred-capital priority further determine the residual available to common equity. A large revenue stream describes scale; the capital structure determines who receives it.

A quarter of delay can outweigh years of rent escalation

Assume a 100MW facility charges $150 per kW per month in base rent. Fully billable capacity produces $180 million annually. A three-month delay with no billing reduces that year’s rent by $45 million. This is an illustrative contract, excluding expense savings and any compensation, not a forecast for a named company.

If $800 million of debt has already been drawn at an 8% annual rate, the same three-month interval also carries approximately $16 million of interest. Forgone rent and financing expense are different effects and should not simply be added as an accounting profit loss. Both nevertheless lengthen the shareholder’s wait for cash.

Digital Realty’s disclosed contracts contain 3.6% annual rent escalators.^[4] With other terms unchanged, year-ten rent would be approximately 1.375 times year-one rent, and year-fifteen rent about 1.641 times. Escalation helps nominal income grow, but cannot offset every construction overrun or guarantee a discounted return.

Timing, cost and financing belong in one model. Prompt energization can support good economics at a moderate initial rent; repeated delays can erode an apparently attractive contractual yield. Early phases are especially important because they determine when operating cash can begin supporting subsequent construction.

Platform backlog must retain the ownership share

Digital Realty reported a nine-month weighted-average interval between new lease signing and contractual commencement in the second quarter. Signed but uncommenced leases represented $1.9 billion of annualized GAAP base rent at 100% ownership, or $1.4 billion at the company’s share.^[3] Citing only the first figure risks assigning partners’ economics to the listed company.

Neither number becomes revenue in a single subsequent quarter. Annualization describes the scale after relevant leases commence, while actual recognition follows phased delivery. GAAP rent may also use straight-line accounting and differ from cash collected. Annualization, ownership and cash accounting must remain distinct in an earnings model.

Larger platforms have different customer and market exposures from a single-campus developer, but scale does not eliminate construction or financing risk. Equinix’s interconnection and xScale mix, and GDS’s area and delivery measures, require their own definitions.^[6][7] A single per-MW multiple cannot capture every operator.

VNET should likewise be followed from orders into usage. The 510MW contract indicates willingness to reserve future resources; delivery, customer installation and realized revenue provide the next evidence.^[5] A gap between orders and utilization may reflect normal phasing or a delay. The project timetable distinguishes them.

Which contracts are more likely to leave cash for shareholders?

An attractive contract combines achievable construction cost, enforceable payments, reasonable allocation of rising expenses and financing that covers development and ramp-up. Weakness in any of those conditions reduces the protection offered by a long lease term.

Positive evidence is concrete: billable capacity rises on schedule, cash rent arrives, construction costs remain controlled and the remaining funding gap shrinks as phases open. New projects are more valuable when those capabilities can be repeated without creating excessive customer concentration.

The adverse evidence includes recurring schedule revisions, tenant deferrals, reliance on expensive funding and accounting revenue that outruns collections. A contract need not be canceled for common-equity returns to deteriorate. Cash-flow timing should change in the model when the facts change.

Long leases provide clearer revenue visibility, but transaction size is the starting point of analysis. Investors ultimately own residual cash after the full cost of capital. Following each agreement through construction, acceptance, billing and collection reveals how much shareholder value the large order can actually create.