From Grid MW to Live IT Capacity

A data center can be described as 500 MW and still leave the most important commercial question unanswered: 500 MW of what, available when, and supported by which evidence?

The distinction affects valuation, site comparison, sales pipelines and grid planning. A utility connection envelope, a campus electrical design, a building's critical load and an operating IT load are related, but they are not interchangeable.

PowerlandMap uses an evidence ladder to keep those quantities separate. The objective is not to reduce every project to one headline number. It is to show what each number represents, how far the project has progressed and what still has to happen before capacity can support customers.

Start with the measurement boundary

Power Usage Effectiveness compares total facility energy with energy delivered to IT equipment. The U.S. Department of Energy's measurement guidance distinguishes whole-site demand from IT equipment input. Cooling, power conversion, lighting and other supporting systems consume part of the energy entering the site.

Google's published PUE methodology illustrates the point operationally. Its fleet-wide figure includes overhead such as substations, transformers and water-treatment systems. A PUE of 1.09 does not mean a 100 MW grid feed automatically supports 100 MW of IT. It means that, under the stated boundary and operating conditions, facility energy is 1.09 times IT energy.

That relationship is useful, but it is not a universal conversion factor. PUE varies by site, climate, load, design and operating state. The Uptime Institute's 2025 survey reported a weighted average PUE of 1.54 across respondents, showing why a generic assumption can materially distort capacity.

Level one: grid opportunity

The first level is an electrical opportunity: a utility study, landowner statement, network-development plan or indicative connection envelope.

This is not secured power. Evidence should identify the network operator, connection point, voltage, requested capacity, study status, reinforcement scope and target date. It should also state whether the number is import capacity, generation, a queue position or future system potential.

This is the starting point for grid-connection diligence, not the end. A regional generation surplus cannot be allocated to a site without project-level proof.

Level two: contracted electrical capacity

The second level requires a firmer document: an executed connection agreement, accepted offer, capacity reservation or equivalent utility commitment.

Even then, the date matters as much as the megawatts. The project may depend on a substation, transmission reinforcement, phased energisation, securities or milestone payments. Capacity can be contracted yet unavailable for years.

This is why powered-land assessment should combine electrical evidence with land control, permitting and delivery obligations. The useful output is a phase-by-phase schedule with conditions, not one campus total.

The PJM registry approved by U.S. regulators is relevant because it is expected to capture service dates, ramp schedules, backup generation and locations. Reuters reported the decision on 12 October 2026. Better visibility still does not turn a regional demand forecast into live IT supply.

Level three: facility electrical design

The third level concerns what the campus and buildings are designed to receive and distribute.

A project may have a 100 MW grid connection and several buildings, each with transformers, UPS systems, cooling plant and a redundancy arrangement. The gross electrical envelope covers the entire facility, not only servers.

Design capacity remains future capacity until equipment is procured, installed, commissioned and energized. A masterplan showing six buildings is not six operational facilities. The same anti-double-counting rule applies to phases: list them separately only when they are genuinely identifiable, and never add both the campus total and its components.

Evidence belongs in a development-readiness plan and should align with permit diligence.

Level four: designed IT capacity

Designed IT capacity is the critical load intended for compute, storage and networking under a defined configuration.

Its basis must still be explicit. Is it nameplate capacity, design load, initial fit-out, maximum fit-out or a modeled result? Does it apply to one building, one phase or the whole campus? What resilience arrangement supports it?

A PUE assumption can reconcile IT design with facility power only when the boundary and operating case are documented. The same applies to rack density. A 150 kW rack architecture is a design characteristic, not proof that every rack has been installed, sold or loaded.

This distinction extends the framework in Data Center Capacity Is Not One Number.

Level five: commissioned and available IT capacity

Commissioned capacity has passed defined testing and can support equipment under agreed operating conditions. It should be tied to a building, room or phase and to a commissioning date.

Availability adds another test: is the capacity free for a customer, reserved, fitted out, powered shell or already occupied? Those commercial states cannot be inferred from engineering completion alone.

Evidence can include commissioning certificates, operational announcements and dated operator disclosures. Acceptance criteria belong in commissioning readiness.

Level six: contracted and live IT load

Contracted capacity is allocated to a customer under an agreement. Live load is operating. The two can differ materially during deployment and ramp-up.

A lease, pre-lease or cloud-services agreement can provide demand evidence, but the scope, conditions and start date matter. A reported negotiation is not a contract. On 13 October 2026, Reuters reported that NTT Data was in final talks with Anthropic over an Osaka facility described as taking up to 500 MW of power. Until the parties confirm the agreement, site, phasing and measurement basis, the figure is an electrical-envelope signal rather than delivered IT capacity.

The six fields every number needs

For every capacity figure, record:

• measurement boundary
• asset and phase
• status
• evidence source and date
• earliest supported availability
• remaining conditions

Unknown values should remain unknown. Estimated IT capacity should be labeled and kept separate from sourced IT capacity. Electrical, grid, generation and campus figures should retain their original meaning.

Contradictions are useful. If a utility document, planning application and operator announcement describe different capacities, do not average them. Identify their boundaries and dates, then determine whether they describe separate phases or a genuine conflict.

What buyers should compare

A credible market comparison separates present supply from future potential. It lets an investor see whether value comes from land, power rights, permits, construction progress, customer commitment or operating infrastructure.

The PowerlandMap coverage is built around that separation. The public demo illustrates the decision workflow without presenting demonstration data as a live commercial dataset. The market-study catalogue provides structured analysis for specific geographies and questions.

The headline number is useful only when its evidence is visible. The commercial advantage comes from knowing which milestone has actually been reached, which risks remain and what can realistically be delivered next.

Teams comparing projects can request access to PowerlandMap to review capacity, power and development signals with their sources and dates.

*Analysis published on 13 October 2026. Project figures remain attributed to their original sources. Hero image: original conceptual illustration.*