Powered Shell vs Turnkey Data Center: Scope, Cost Boundaries and Acceptance

A powered shell provides a building and an agreed level of site infrastructure for a customer to complete. A turnkey data center includes a more extensive fitted infrastructure scope, delivered to a defined readiness condition. The labels alone do not establish how much power is usable, whether cooling and backup systems are installed, who owns the fit-out, or which tests must pass before handover.

The practical comparison is therefore between two delivery boundaries. Identify what exists at each boundary, what work remains and who must prove that the combined system supports the intended IT. This matters whether the project uses a conventional building, prefabricated modular data centers, or a combination of factory-built systems and site construction.

Our AI data center business-model guide examines who owns infrastructure and who sells compute. Here the question is narrower: what state of facility readiness are you purchasing or leasing, and how do you accept it? Ownership and delivery scope can vary independently.

TL;DR

• Powered shell and turnkey describe market offerings, not a universal inclusion list. Compare the actual drawings, equipment schedules, exclusions and readiness definitions.
• Separate utility capacity, installed facility capacity and supported IT load. An allocated power figure is not proof of energized, tested rack capacity.
• Cooling must meet the tenant's hardware interfaces, including residual air loads and the outdoor heat-rejection boundary.
• Factory tests, site checks and integrated demonstrations answer different questions. Passing one does not automatically complete the others.
• Normalize cost to the same completed scope and operating state, including owner-side integration, commissioning and remaining site work.
• Choose the model your team can deliver and operate. A shell can offer control; a turnkey package can consolidate defined responsibilities. Neither removes all project risk.

What the terms establish—and what they leave open

Published developer descriptions show why the distinction is useful. Trammell Crow Company describes powered shells in terms of the building and exterior improvements, including specified power, fiber and security infrastructure; its turnkey description extends into interior mechanical, electrical and telecommunications fit-out. That is an example of a supplier's boundary, not an industry-wide specification.

Digital Realty's 2025 annual report similarly distinguishes its fitted offerings from Powered Base Building capacity for customers able to build and operate their own infrastructure. These products illustrate different customer responsibilities. They do not imply that every shell includes the same substation, or that every turnkey facility includes the customer's servers.

Treat a proposal's label as the beginning of the scope discussion. Ask the supplier to place its handover boundary on a physical diagram and in an equipment list. Then reconcile that boundary with the lease, construction or supply documents and the operating arrangement. Technical scope and the commercial document should describe the same facility.

Compare delivery states with a boundary matrix

The following is a comparison worksheet, not a list of guaranteed inclusions. Complete it for each offer before comparing headline prices.

Use an interface register alongside the table. Every unresolved item needs a responsible party, evidence required, decision date and dependency. An entry marked “by others” names neither the other party nor its completion date. Close those entries before design or procurement decisions rely on them.

For the separate question of which manufacturer, provider or integrator should coordinate the work, use the delivery-model selection guide. A turnkey outcome can involve several contractors; its accountability depends on the agreed scope, authority and remedies, rather than the number of brands on a drawing.

Define exactly where power becomes usable

“Power available” can refer to land near a transmission route, a reserved utility allocation, an energized connection, installed electrical equipment or tested capacity at the rack. Each represents a different stage of readiness and a different remaining investment.

Require four separate statements:

1. Connection basis: delivery point, voltage, capacity rights, conditions, milestones and curtailment arrangements where applicable.
2. Installed path: equipment and conductors between that point and the IT, including all interfaces that another party supplies.
3. Supported operating state: IT capacity with the agreed equipment unavailable for maintenance or under a specified fault scenario.
4. Demonstrated result: test conditions, measurements, deviations and evidence that the assembled path meets the requirement.

Do not compare an upstream MVA rating directly with a downstream MW IT promise. Power factor, losses, auxiliary demand and operating topology must be addressed in the electrical design. Nor does annual PUE convert a utility allowance into a safe peak connection size. Our power-distribution guide explains the chain of equipment; the AI load-profile guide examines why changing compute demand matters to that chain.

A shell can have a satisfactory connection while leaving extensive downstream work to the tenant. A turnkey package can include downstream systems while excluding a utility upgrade. Both arrangements can be valid. Neither should be presented as complete rack capacity until the remaining boundary is explicit.

Make the cooling boundary specific to the tenant

An air-temperature commitment is insufficient for equipment that also requires a liquid interface. A cooling capacity expressed in megawatts is incomplete without temperatures, flows, ambient conditions and the supported operating cases.

For air-cooled servers, establish the inlet envelope, airflow, containment, exhaust path and service constraints. Rear-door heat exchangers move heat from server exhaust into a water loop while the IT remains air cooled. Direct-to-chip systems introduce a different equipment interface, while some heat can still require air cooling. These configurations are explained in the liquid-cooling guide.

For a liquid-cooled tenant, record at least the connection points, supply/return limits, fluid specification, water quality, pressure, flow, filtration, leak response and responsibility for flushing and filling. Establish who supplies the CDU and who maintains both sides of its boundary. The CDU sizing guide develops those interface questions.

ASHRAE TC 9.9's water-cooled-server white paper explains the separation between facility and technology cooling loops in a liquid-to-liquid CDU arrangement. Both loops need compatible materials and controlled water quality, and the technology loop needs a clear maintenance owner. Use current OEM requirements for the selected equipment.

Continue the boundary outside. The cooling duty must reach an outdoor rejection system that works under the project's design conditions and agreed degraded states. Water storage, replenishment, adiabatic assistance and any utility-water requirement need their own stated assumptions. A closed internal loop alone does not establish zero site water use.

This is where a fitted facility can still leave substantial tenant work. The supplier may have installed a facility-water loop, but the selected OEM rack might need a CDU, different operating temperatures or separate residual-air provision. Require a jointly accepted interface schedule before calling that facility ready for the tenant's AI system.

Separate construction completion from operational acceptance

A signed delivery note can confirm that equipment arrived. It cannot demonstrate that the building supports the required IT during a utility transition or a maintenance event. Acceptance needs requirements, test procedures, measurement points and a party empowered to accept the result.

ASHRAE's commissioning resources describe a process organized around the owner's project requirements, with documented roles, procedures, reporting and training. Its published catalogue lists Standard 202-2024 for new buildings and systems. The applicable standard and local requirements should be selected for the actual project; naming a standard does not establish that the completed facility complies with it.

Organize the evidence into three useful layers:

Factory acceptance testing verifies the manufactured equipment or integrated package against its agreed specification before shipment. Record the actual assembly, configuration, instruments, test loads and limitations. Factory testing can reduce uncertainty about internal wiring and controls, but cannot prove a site connection or an outdoor system that was absent from the test.

Site acceptance and functional testing verify installation, connections and equipment behavior at the destination. The plan should address inspections, settings, protection/controls interfaces, system start-up and tests under the stated conditions. Replacing “commissioning included” with a priced test schedule makes it possible to see what the package actually covers.

Integrated demonstrations test interactions between systems against the intended service states. The exact scenarios follow the design: for example, loss of the normal supply, transitions between sources, cooling response, isolation for maintenance, alarm escalation and restoration. Agree safety controls and test authority; this is planned specialist work, not an improvised outage exercise.

Uptime Institute's certification programme distinguishes assessment of design, constructed facilities and operations. Where that programme is a project requirement, identify the exact stage, capacity and object being assessed. A vendor design reference, equipment certificate or stated Tier objective is not a constructed-facility award.

Product certification does not settle the whole project

Prefabricated systems need documentation suitable for the installation and its approving authorities. UL Solutions' current modular-data-center guidance describes UL 2755:2025 and separately manufactured functional modules that operate together. The relevant lesson is traceability: know which product or assembly was evaluated and how it connects to the site.

Request the actual applicable evidence rather than collecting logos. Identify its issuer, holder, scope, configuration and conditions. Then list remaining local design submissions, inspections and site acceptance obligations. A certificate on an electrical assembly does not answer whether tenant piping has been cleaned correctly, or whether an operating team can execute the agreed failure-response procedure.

For a shell, the tenant needs access to the base-building records required for fit-out and integration. For turnkey delivery, the customer needs a consolidated, controlled handover pack rather than disconnected manuals from each equipment supplier. Neither model should leave critical settings, drawings or operating instructions available only in a contractor's private archive.

Normalize the cost of reaching the same endpoint

The shell price and the turnkey price usually cover different work. Build a common endpoint—such as a defined IT capacity, rack schedule and operating state—and reconcile every missing item to it.

For the shell route, total completion cost includes the shell commitment, remaining fit-out, design and integration, equipment procurement, site testing, owner management and contingency for the retained scope. For the turnkey route, add the package's exclusions, customer-side work and the contingency associated with those retained obligations. Lease cash flows and an owned construction package also need a common comparison period and financial treatment.

Avoid assuming the larger invoice is automatically the more expensive route. Conversely, a single-point package can carry a premium without transferring the risk the customer actually wants transferred. Test the boundary and the proposed delivery capability together.

Use the modular data center cost guide for facility cost drivers. Keep the server fleet and AI operations separate unless they are explicitly included. The useful commercial output is a normalized inclusion/exclusion schedule with dated quotations, rather than an unsupported price per megawatt.

Put schedule dependencies beside the handover date

Request a programme that shows design approval, utility milestones, long-lead components, site works, factory work, transport, installation, testing and operational handover. Mark dependencies controlled by the supplier, customer, utility and other contractors.

Factory integration can allow some activities to run alongside site preparation. That advantage depends on an early enough interface freeze and a site ready to receive the package. A late change to rack geometry, coolant conditions or incoming voltage can affect equipment already in production. The site-selection guide provides the earlier site and utility questions; the completion programme should connect those answers to delivery.

Define several dates rather than one ambiguous “delivery” date: ready to receive equipment, equipment delivered, ready for energization, ready for agreed testing, ready for IT installation and ready for operational service. Some can coincide; they need not. State the conditions that move each milestone and how a delayed owner-side dependency affects the sequence.

For a phased project, reserve space and interfaces for expansion without representing future capacity as installed. Show whether the next phase interrupts the operating first phase. A shell that supports a later larger build and a turnkey first phase may be compared together, provided the future works and their disruption risk remain visible.

Worked procurement example: the same 2 MW request, different offers

Consider a hypothetical tenant planning 2 MW of IT load. Its rack schedule and cooling requirements are still being finalized. Offer A is a powered shell with a stated upstream electrical allowance. Offer B is a fitted package described as “2 MW ready.” Neither description is enough to establish a winner.

The team first asks whether A's upstream allowance can support the desired IT plus auxiliaries through the required power topology. It asks B whether 2 MW means connected equipment capacity, usable IT capacity in normal operation, or capacity retained in the specified maintenance state. No annual efficiency ratio can substitute for that electrical review.

Next it compares cooling. A reserves plant space but requires the tenant to design and procure the complete thermal path. B includes a cooling system, but the team must verify its interfaces against the final OEM schedule. A nominal liquid-cooling provision without the required temperatures, flow or residual-air capacity remains an open item.

Finally the team names the acceptance owner. Under A, its fit-out integrator may coordinate the completed system, with access to the landlord's base-building test records. Under B, the package provider may coordinate the defined facility scope, while the tenant still accepts the computing system. In both cases, a common test matrix identifies shared demonstrations and unresolved dependencies.

The result can be a shell, a turnkey package or a revised hybrid. The decision follows evidence of completed scope, retained work and team capability. It does not require changing the tenant's business model or pretending that either contract supplies an unspecified AI service.

Choose the responsibility you can actually carry

A powered shell can suit a tenant with a mature design, procurement and operating organization that wants control of its fit-out. The retained scope still needs budget, authority and competent delivery resources. Buying an incomplete facility is a deliberate choice when those capabilities exist.

A turnkey route can suit a customer seeking coordinated delivery of a defined infrastructure outcome. Verify that the prime has authority over the interfaces it is being asked to own. The customer still needs a capable acceptance function and an operating plan, even when more delivery work sits with the supplier.

For either route, close five questions before award: What exact service state is being purchased? Which interfaces remain outside it? Who coordinates the whole design? What evidence triggers acceptance? Who operates and maintains each system afterward? These answers form a practical starting point for the ModulEdge configuration discussion.

ModulEdge's products provide defined infrastructure routes around selected equipment and sites. Project scope still matters: for example, the issued MDC2000 base reference excludes upstream MV switchgear, transformers and standby generators unless added for the project. A broader supply package must state those inclusions explicitly. The MDC2000 page describes its current public configuration; a project proposal establishes the applicable delivery boundary. Tenant introductions, where discussed, do not replace tenant commitments or guarantee occupancy.

Frequently asked questions

Can IT installation begin before the whole facility is accepted?

Use an agreed partial-release plan identifying the area, permitted activities, prerequisites and outstanding work. Receiving racks, connecting them and energizing them can require different releases. Record who controls access, protects installed equipment and authorizes each next step. Permission to install equipment in one completed zone should not be treated as acceptance of the site's full operating capacity.

What if the tenant changes its rack after shell handover?

Open a controlled change against the accepted interface schedule. Check electrical input, cooling conditions, mass, geometry, cabling and maintenance access, then identify affected designs, equipment and tests. Agree the cost and schedule consequences before dependent work proceeds. A rack with the same nominal kilowatts can still require different hydraulic or physical interfaces; the shell's prior handover does not validate the new configuration.

Which records need to follow a change of operator?

Transfer a controlled baseline: as-built drawings, asset and settings registers, test results, operating procedures, maintenance history, warranties and open defects. Reassign access and escalation responsibilities, and record who accepts the operating handover. A walkthrough and representative procedure demonstration can reveal missing knowledge. Equipment ownership, delivery acceptance and responsibility for day-to-day operation should remain separately identifiable.

Should equipment warranties start at the operating milestone?

Check the actual commercial terms for each package. Equipment shipment, installation, supplier acceptance and operational service can occur on different dates, with different warranty or support triggers. Put those dates and responsible parties in the handover schedule. If utility or tenant delays consume part of a coverage period, resolve the required protection explicitly rather than assuming the final facility milestone resets every supplier's warranty.

What if the design-day weather cannot be reproduced during testing?

Agree the evidence method in the test plan. It may combine witnessed measurements at available conditions, supported performance data and a defined later seasonal test. Record assumptions, outstanding demonstrations, operating limits, responsible parties and completion dates. A deferred test needs a tracked obligation; a favorable cool-day result alone should not become an unqualified claim of design-day capacity.

How should remaining defects be handled at handover?

Maintain a defect register with severity, affected operating states, corrective owner, due date and verification evidence. Distinguish items compatible with a specifically agreed limited release from items preventing the required service state. Document temporary controls and their expiry. Closing an item should require the relevant inspection or test result, rather than only a statement that the contractor has finished working on it.