Power Economics for Hyperscale Data Centers: Why LCOE Is a Starting Point, Not the Investment Answer

LCOE is useful for screening generation technologies under a common set of assumptions. A data center investment requires a broader cost bridge that incorporates the grid, interconnection, firming, backup, contract, schedule, and capital structure around the power supply.

This analysis reflects evidence reviewed through August 23, 2026.

Power economics are project economics

Electricity is often one of the largest operating cost exposures for a hyperscale facility. Its actual share of cost varies materially with workload, ownership model, lease structure, utilization, cooling design, local tariff, and accounting treatment. For investors and developers, the durable conclusion is not a universal percentage. It is that relatively small changes in the all-in power price can materially change cash flow at scale.

A one-cent-per-kilowatt-hour difference equals $10 per megawatt-hour. At a 100 MW metered average load operating continuously, that difference is approximately $8.76 million per year. At 500 MW, it is approximately $43.8 million per year, before escalation, load growth, or changes in utilization. This arithmetic is why power diligence belongs in the core investment case, not in a technical appendix.

The illustrative calculation is 100 MW × 8,760 hours × $10/MWh. Actual consumption, ramp, outages, demand charges, losses, and contract terms will change the result.

What LCOE measures

Levelized cost of energy converts the lifetime cost and output of a generation asset into an average cost per unit of electricity, typically expressed in dollars per megawatt-hour. In simplified form, it is the present value of capital, operating, maintenance, fuel, and other modeled costs divided by the present value of electricity generated.

Used carefully, LCOE can create a consistent first-pass comparison across generation technologies and highlight sensitivity to capital cost, cost of capital, capacity factor, fuel, operating expense, tax treatment, and asset life. The U.S. Energy Information Administration's AEO2025 analysis defines LCOE as the cost to build and operate a generator over a specified recovery period. It also states that LCOE does not capture every factor that contributes to investment decisions or the value a plant provides to the grid.

The answer is only as comparable as the assumptions. Capacity factor, construction timing, financing, fuel, asset life, subsidies, regional resource quality, and included cost categories must be aligned before two LCOE figures can be treated as alternatives.

What current benchmarks show—and what they do not

In its June 2025 LCOE+ analysis, Lazard estimated unsubsidized ranges of $37–$86/MWh for onshore wind, $38–$78/MWh for utility-scale solar, $48–$109/MWh for gas combined cycle, and $50–$131/MWh for utility-scale solar plus storage. These ranges are useful market reference points, not customer bills or site-specific offers.

Lazard expressly identifies potentially significant factors outside the core comparison, including transmission, network upgrades, congestion, curtailment, integration, permitting, and certain development or environmental costs. Its analysis also uses standardized financing and fuel assumptions and describes itself as a point-in-time analysis rather than a forecast.

EIA reaches a similar methodological conclusion from a different analytical framework: technology cost must be considered together with regional variation and system value. That distinction is especially important for a data center, which is not purchasing an abstract annual average. It is seeking a specific quantity of power, in a specific location, with a defined ramp, reliability standard, emissions objective, and delivery date.

The LCOE comparison trap

Build the all-in delivered power cost

The relevant underwriting bridge begins with the price of energy or the economics of dedicated generation, then adds the costs and risks necessary to convert that supply into usable campus power. Depending on the structure, it may include generation, utility energy, or contracted PPA pricing and escalation; transmission, distribution, network upgrades, interconnection facilities, substations, and line losses; capacity, demand, standby, minimum-bill, rider, and other tariff charges; and firming, storage, backup generation, fuel logistics, ancillary services, and reserve requirements.

It should also capture congestion, basis, shape, imbalance, curtailment, and settlement exposure; onsite capital, operating expense, maintenance, replacement reserves, insurance, and decommissioning; credit support, deposits, collateral, termination payments, take-or-pay commitments, and guarantees; taxes, incentives, renewable attributes, carbon requirements, and environmental compliance; and delay, availability, outage, ramp, and performance risk translated into schedule and cash-flow scenarios.

Some items are paid through monthly bills, some are embedded in contract pricing, and others are capitalized or contingent. They should be placed on a common basis without double counting. The objective is not false single-point precision. It is a transparent cost stack that shows which assumptions drive value and which party bears each risk.

The labels are less important than complete scope, consistent timing, and clear risk ownership. A low energy price can be outweighed by an expensive interconnection, weak delivery date, high demand charges, or a requirement to procure separate firming and backup capacity.

What would change the investment answer?

• The framework is stable, but the preferred power architecture is conditional. Re-underwriting is warranted when a binding interconnection or utility-service date replaces an indicative schedule or moves materially
• a tariff, special contract, or regulatory order changes minimum demand, collateral, exit, curtailment, or cost-allocation obligations
• a generation, storage, or backup proposal demonstrates a materially different site-specific all-in cost
• the customer load ramp, flexibility, redundancy requirement, or service-level commitment changes.

The answer should also be revisited when fuel, congestion, basis, financing, tax, or equipment assumptions move outside the approved downside range, or when a credible alternative can meet the energization date with lower risk-adjusted cost or greater strategic flexibility.

Do not preserve a preferred solution merely because it was cheapest in an earlier model. Update the cost bridge when binding documents, schedules, tariffs, technical requirements, or counterparties change.

Reliability changes the comparison

A hyperscale campus generally needs a dependable power profile, while generation technologies differ in when and how reliably they produce. Comparing intermittent energy with dispatchable or firm capacity on LCOE alone can therefore compare unlike products. Storage may reshape output over hours, but it does not automatically provide multiday or seasonal firmness. Backup systems may protect critical loads, but their economics, permitting, fuel supply, run-time limits, and emissions profile require separate analysis.

Lazard's 2025 report treats the cost of firming intermittency as a separate system question and notes that the value of firm capacity rises as intermittent penetration increases. For a campus-level investment case, the analysis should specify required availability, duration, redundancy, restoration, and operating mode, then cost the architecture that satisfies those requirements. Those technical judgments should be developed with qualified engineering, utility, and operating specialists.

Location and schedule can dominate nominal cost

A theoretically attractive supply source has limited value if it cannot be interconnected within the development schedule. The Federal Energy Regulatory Commission's 2025 State of the Markets report estimated more than 50 GW of in-service U.S. data center capacity at year-end 2025 and described increasingly large loads seeking faster interconnection. FERC noted that facilities entering service in 2025 averaged almost 80 MW, while projects expected in 2030 averaged more than 400 MW in the database it reviewed.

The scale and timing of those requests can require new generation, transmission, substations, or other network investment. This makes the promised energization date, milestone schedule, cost responsibility, and remedy package as important as the modeled steady-state power price. A cheaper solution arriving two years late may be economically inferior to a higher-priced solution that supports a contracted delivery obligation.

Underwriting should connect power milestones to land obligations, equipment orders, construction spend, tenant commitments, financing availability, liquidated damages, and carrying cost. Schedule risk belongs in the same model as price risk.

Commercial structure reallocates risk; it does not eliminate it

Grid service, utility special contracts, physical PPAs, financial or virtual PPAs, retail supply agreements, onsite generation, and hybrid arrangements allocate different combinations of price, volume, shape, basis, curtailment, credit, construction, fuel, and regulatory risk. Two structures with similar expected prices can produce very different downside cases.

Large-load tariffs are also evolving. Lawrence Berkeley National Laboratory's 2025 rate-design brief identifies issues including fair allocation of system costs, stranded-asset protection, resource adequacy, minimum demand commitments, collateral, contract term, and exit fees. These provisions can improve project credibility by clarifying who pays for infrastructure, but they can also create long-duration fixed obligations that remain even if the campus ramps more slowly than planned.

A financeable structure should identify the party best able to manage each risk, align contract duration with the asset and customer obligations, and preserve enough flexibility for phased growth. The analysis should test both expected cost and the cost of being wrong.

A transaction-focused diligence framework

Power diligence should produce decision evidence, not only a technical description. For a proposed investment or financing, the workstream should reconcile load by phase; supply source, quantity, profile, fuel, environmental attributes, and replacement options; delivery counterparties, facilities, upgrades, milestones, and enforceable service dates; and all-in economics, fixed and variable charges, escalation, capitalized costs, incentives, losses, and downside sensitivities.

It should also reconcile contractual term, credit support, minimum purchases, curtailment, change in law, force majeure, termination, step-in rights, and remedies; required reliability, backup and storage roles, outage exposure, insurance, and service obligations; capital sequencing, contingency, cost-overrun support, refinancing, reserves, and alignment with customer commitments; and credible fallback supply, phased energization, load flexibility, and the value or cost of delay.

The key investment assumptions should be mapped to documents, counterparties, milestones, and remedies. Unverified savings should not be capitalized at the same value as contracted, executable savings.

Why scenario analysis matters

Power demand is rising rapidly, but the outcome is not a single forecast. The International Energy Agency's Energy and AI analysis emphasizes the dependence of AI deployment on electricity supply. In the United States, LBNL's June 2026 update estimates a 2030 reference case of 649 TWh for data center electricity use, with compounded uncertainty scenarios spanning 521 to 843 TWh.

A project model should be equally explicit about uncertainty. At minimum, it should test load ramp, utilization, energy and fuel prices, congestion, escalation, construction delay, interconnection cost, financing cost, capacity factor, equipment replacement, tariff change, tax treatment, and customer credit. The purpose is not to predict every outcome. It is to identify which variables can impair debt service, equity returns, or delivery commitments before capital is exposed.

What should drive the decision

A disciplined power strategy balances five dimensions: expected all-in cost, reliability, delivery schedule, contractual risk, and strategic flexibility. Sustainability objectives and regulatory obligations should be treated as design requirements and priced transparently, not appended after the commercial structure is selected.

The preferred answer may be a portfolio rather than one source: grid service for scale, contracted generation for price or attribute objectives, storage and backup for defined reliability functions, and phased capacity to match the customer ramp. The right mix will depend on site-specific engineering and utility conditions. From a capital and transaction perspective, the objective is to make that mix legible, financeable, and resilient under downside scenarios.

The power case should answer with evidence: What is contracted, what is indicative, and what remains an assumption? Who controls the delivery date, and what remedies exist if it is missed? Which costs are excluded from the headline price or LCOE? Which party bears price, volume, performance, credit, and change-in-law risk? What is the executable alternative if the load ramp or power plan changes?

Bottom line

LCOE is a useful screening metric, but hyperscale investment decisions turn on the all-in, risk-adjusted cost of power delivered to the campus when it is needed.

The analysis becomes investable only after generation cost is reconciled with the grid, interconnection, reliability architecture, contract terms, schedule, and capital structure. In a market where a $10/MWh difference can move annual cost by millions of dollars, precision about scope and risk ownership is not a technical detail. It is central to value.

Robert Dizon is Nistar's Chief Financial Officer for Capital Markets & Transactions. His perspective is grounded in 19 years of M&A and transaction advisory experience across 120+ transactions, with emphasis on diligence, transaction execution, operating improvement, capital structure, and risk identification. If you are evaluating hyperscale power economics, a campus power strategy, or a related capital structure, contact our team to discuss your project.