From megawatts
to cash flow.

Investment view

Scarcity creates value only when the asset can capture it.

A large addressable market does not establish the value of an individual development. The investable proposition is a documented path from site rights and power service to customer acceptance and cash collection, purchased at a price that compensates for the remaining uncertainty.

Our central hypothesis is that, where local power access constrains development, verified delivery and enforceable customer commitments can support an economic advantage. That advantage is conditional. It can be consumed by construction costs, transferred to utilities or suppliers, or eliminated by the acquisition price. A well-positioned asset can still be a poor investment.

This paper combines dated public evidence with an original decision framework and transparent illustrative calculations. It does not estimate a specific asset’s fair value or assert that regional scarcity causes higher investment returns.

01 · Evidence and inference

Demand is material. Delivery remains conditional.

There is also evidence against a uniformly tightening market. PJM’s January 2026 forecast reduced near-term peak expectations through 2032 after revised economic and vehicle assumptions and better vetting of large-load requests. Forecasts should be treated as conditional planning records, with careful attention to duplicate applications and project maturity. [5]

Berkeley Lab reports that only 13% of capacity requested in 2000–2020 had entered operation by the end of 2025, while 75% had withdrawn. For the available subset of regions, completion timelines exceeded five years. These are historical generator outcomes, not a probability or timeline to apply to a new data center. Differences in cohort, region, technology, and unresolved projects limit comparison. [2]

Inference boundary. These records support diligence on delivery risk. They do not identify a causal scarcity premium, establish future rents, or prove that returns will exceed the cost of capital. Evaluating those propositions requires transaction and operating evidence matched by location, contract quality, vintage, and development stage.

02 · Define the asset

One megawatt can describe several different things.

A campus announcement may describe ultimate facility demand. A lease may price contracted IT capacity. A utility agreement may cover a smaller initial block with a staged ramp. Treating these quantities as interchangeable overstates either revenue capacity or power certainty.

Use the same unit and period throughout the model. MW measures power; MWh measures energy. Power usage effectiveness relates total facility energy to IT energy over a measurement period. An annual PUE assumption does not itself establish peak electrical requirements, redundancy, or a firm service entitlement.

Billable capacity is also different from server utilization. A customer may owe capacity rent while using little compute, or occupy a space while rent commencement remains conditional. The contract determines the revenue mechanism; technical utilization informs operating and renewal risk.

03 · Delivery and critical path

Measure readiness with evidence, not milestones alone.

Construct a dependency schedule linking utility works, permits, equipment, construction, commissioning, and customer acceptance. Identify the last unresolved predecessor for each revenue-producing phase. An earlier transformer delivery has no financial benefit if a substation or permit remains the binding constraint.

A useful first-pass representation is: operating date equals the latest required readiness date plus remaining integrated commissioning. This is valid only if the readiness dates share a consistent basis and do not already include commissioning. A full schedule must represent sequential work, available float, resource constraints, and common causes of delay.

Do not multiply independent completion probabilities across workstreams without evidence of independence. A design change can simultaneously reset equipment approval, building works, and customer acceptance. A common regional bottleneck can affect every supplier quote in a shortlist. Build joint downside cases around such mechanisms.

For a staged investment, preserve the ability to defer later phases until the evidence improves. That option has value only if land, supplier, customer, and financing agreements permit a pause at an acceptable cost. Irrevocable deposits and cross-defaults can remove much of the apparent flexibility.

04 · Contract and regulatory allocation

A strong tenant does not eliminate project risk.

Blackstone’s 2026 mid-year perspective emphasizes power and zoning barriers and long-term pre-construction leases with large technology customers. This is a relevant statement of an institutional investment thesis from a commercially interested owner. It is not independent evidence about the merits of a particular asset. [6]

Test that thesis against the documents: identify the contracting entity, parent guarantee, acceptance conditions, termination rights, remedies for late delivery, and the treatment of expansion phases. A tenant name on a presentation does not identify the entity that must pay. A lease term does not reveal how much downside transfers to the landlord.

Map power, fuel, capacity, transmission, and backup costs to the party that ultimately bears them. A nominal pass-through can still create working-capital exposure, collection disputes, tenant affordability risk, or an unhedged mismatch between the site’s tariff and a financial power hedge.

Regulatory treatment also remains a moving input. PJM’s 13 August 2026 filing proposes an Interim Resource Adequacy Service and a Large Load Registry, including an emergency reduction framework for certain new loads without qualifying supply coverage. This paper treats that filing as a proposal, not an enacted nationwide rule or a prediction of FERC’s decision. [7] FERC’s June 2026 actions separately directed six regional operators to justify or reform large-load integration rules. [8]

Run a curtailment case against the actual customer service obligation. Backup generation may address some outages but introduces fuel, emissions, maintenance, and runtime constraints. A financial hedge does not deliver physical power during a local interruption. Regulatory diligence must resolve the current tariff and retail service agreement before commitment.

05 · Transparent financial sensitivity

Small changes can compound across the capital structure.

Illustrative assumptions only. Consider 100 MW of IT capacity, $800 million of project cost, and $480 million of interest-only debt. The cases below are constructed sensitivities, not market estimates, supplier quotes, forecasts, or Blackstone data. No probability is assigned to any case.

Annual rent equals IT capacity × 1,000 kW/MW × billable fraction × monthly rent per kW × 12. Operating cash before tax and debt service equals rent × operating cash margin, less sustaining capital expenditure. Power reimbursements and the corresponding expense are excluded from both revenue and margin. The cash margin is an explicit case assumption, not a model of fixed and variable costs.

The combined downside does not cover interest from the modeled cash available. Its break-even billing fraction rises to 72.0%, compared with the assumed 50%. That is a funding question before it is a terminal-valuation question. Break-even is conditional on each case’s rent, margin, sustaining capex, and interest cost.

Coverage here is deliberately called pre-tax cash interest coverage. It is not lender-defined debt-service coverage. A transaction model must add taxes, working capital, fees, amortization, replacement events, lender reserve requirements, and refinancing. It must also represent monthly construction draws and contractual revenue ramps.

The interest example assumes $480 million is already drawn at 7%, with no operating receipts or offsetting recovery, and uses simple interest. The present-value example assumes a 10% annual discount rate and shifts the entire future cash-flow stream without changing its amounts or lifespan: reduction = 1 − (1.10)^{−delay in years}. These are separate sensitivities and must not be added together.

A fixed lease-expiry date, unrecoverable revenue, penalties, or a changed exit value can produce a different result. Conversely, phased delivery, delayed capital spending, or enforceable damages can reduce exposure. Model those mechanisms explicitly rather than applying a universal delay haircut.

Capital overruns can also concentrate in equity. At the illustrative $800 million cost and $480 million debt, initial equity is $320 million. A 20% cost increase with unchanged debt requires another $160 million—50% more equity. This arithmetic does not establish the resulting return; it identifies a funding sensitivity.

Download assumptions, formulas, and calculated results (JSON) ↗

06 · Countercase and falsification

State what would weaken the thesis.

The opposing case is that supply response, process improvements, compute efficiency, geographic substitution, and customer bargaining power reduce the economic benefit of scarce sites. These mechanisms can coexist with aggregate demand growth. Rising usage does not guarantee rent growth if the supply response is faster, and a secured site does not guarantee pricing power against a concentrated tenant base.

Better interconnection processing is already visible in the Berkeley Lab record. That does not remove all delivery risk, but it argues against treating historical congestion as permanent. [2] Changes in PJM’s forecast also show that project vetting can reduce near-term assumed demand. [5]

Assign probabilities only when evidence supports them. A scenario range is not a confidence interval. A regional observation is not a parcel-level forecast. The appropriate response to an unresolved material variable may be staged commitment, a contractual condition, a lower price, or no investment.

07 · Investment committee

Require a chain of evidence to cash.

An investment memorandum should allow a reviewer to trace each material return assumption to a contract, a technical record, an observable market comparison, or a clearly labeled judgment. Organize approval around the following questions.

1. What asset is being acquired? Reconcile site control, facility power, IT capacity, development phases, and the revenue perimeter.
2. What proves service will be available? Obtain the current service agreement, utility upgrade plan, equipment evidence, remaining conditions, and curtailment obligations.
3. Who must pay, and when? Review the actual obligor, guarantee, acceptance test, rent commencement, pass-throughs, and relief rights.
4. What funds the joint downside? Quantify delayed revenue, committed spending, interest, reserves, and access to incremental equity.
5. Which risks survive stabilization? Examine tenant concentration, renewal, cooling density, equipment replacement, tariff exposure, and refinancing.
6. What would cause us to stop? Define evidence-based conditions for repricing, deferral, or termination, with a named owner for each unresolved item.

Our conclusion is selective: favor evidence of deliverability and enforceable cash collection, while paying only for risks the investment can bear. This is a framework for testing an asset, not a categorical recommendation to buy data centers, utilities, or equipment suppliers.