Powered Shell Data Centers: What's Included, What Isn't, and Why Power Is the Product

A powered shell is a secured utility connection with a building around it, and the speed advantage it offers comes almost entirely from the power having already been arranged.

It sits between raw land and turnkey data centre capacity, which makes scope control the whole game.

The conventional definition is specific. A powered shell is a secure facility with the physical structure, power to the site and connectivity options, excluding major IT infrastructure such as UPS systems, generators and cooling systems. The tenant funds and builds the interior, typically taking on 60 to 70 percent of total fit-out cost that a turnkey landlord would otherwise carry.

That split is the product. Everything else follows from it.

This guide covers what is in and out of scope, why the model exists, what it costs, what to specify, and where tenants get caught.

What's In, What's Out

The scope boundary is the single most consequential term in a powered shell transaction, and market practice varies enough that it must be written down rather than assumed.

White space is the raised floor area where IT equipment sits. Grey space is everything supporting it: electrical rooms, mechanical plant, generator yards.

The cooling ambiguity, and why it matters

Market practice genuinely varies on cooling plant. Some powered shell deals deliver a cooling plant with the building. Others deliver only the structural and spatial provision for the tenant to install one.

That single item can move several million dollars per megawatt, which is why it belongs in the term sheet rather than in a conversation.

The demarcation point

The demarcation point is where the landlord's electrical scope ends and the tenant's begins. In a powered shell, most of the electrical spend above that point sits with the tenant.

Define it by physical location and by equipment, not by description. "Power to the building" means different things to a developer and to a tenant's electrical engineer.

Why the Product Exists: Power

A powered shell is faster than a ground-up build because someone else has already secured the utility capacity, and in the current market that is the only part of the programme that cannot be accelerated with money.

Interconnection has become the binding constraint on data centre development. PJM waits approach seven years for large campuses, and ERCOT has been tracking roughly 474 GW of large-load interconnection requests, around 90 percent of them data centres.

Equipment compounds it. Power transformers, breakers and large conductor are running at 90 to 130 weeks, roughly two to two and a half years.

Against that, construction is the easy part. A building can be built in eighteen months. A utility connection cannot be bought in eighteen months at any price.

What you are actually paying for

The asset is the kW of secured utility capacity. The walls are the delivery mechanism.

That reframes the evaluation entirely. A powered shell with an unsigned utility service agreement is a warehouse, however complete the building looks.

What to verify before signing

Is the utility service agreement executed, and for what capacity? Is the capacity energised, or is it a queue position with a study date? Has the transformer been ordered, and is there a purchase order reference? What happens to the lease if the connection slips?

Those four questions separate a real powered shell from a speculative one, and none of them is about the building.

For the full picture on interconnection queues and equipment availability, see our analysis of power generation trends and the supply constraint. For how AI demand is reshaping generation requirements, see our guide to AI and data centre power needs.

Speed, and What Actually Drives It

A powered shell typically reaches operation in 6 to 12 months from occupancy, against 18 to 36 months for a full ground-up build.

That range assumes the power is secured and energised. It describes the fit-out period, not the development period, and a shell whose utility connection is still in a queue does not deliver it.

The saving comes from three things: the building already exists, the utility capacity is already arranged, and the long-lead electrical equipment on the landlord's side has already been procured.

It does not come from the fit-out being quick. Installing UPS, generators, mechanical plant and white space is substantial work, and it is work the tenant is funding.

The Economics

A powered shell moves 60 to 70 percent of total fit-out cost from the landlord to the tenant, and turnkey delivery requires roughly three to five times the capital per megawatt.

The benchmark numbers

Turnkey data centre construction runs roughly $8 to 12 million per megawatt. For AI-optimised facilities designed for rack-scale density, published benchmarks put it at $15 to 20 million or more per megawatt.

The 2026 institutional benchmark for a hyperscale shell-and-core build, assuming standard cloud-compute load, 1.20 PUE design and air-cooled or hybrid air-evaporative cooling, runs $11 to 14 million per IT megawatt per JLL and Cushman & Wakefield pricing outlooks.

Who the model suits

Developers want to advance a site beyond raw land without funding a full turnkey facility before tenant requirements are certain.

Operators may lease the shell, complete the fit-out and sell downstream capacity.

Tenants get control over redundancy, density and cooling in exchange for carrying the capital and the execution risk.

The model works because different parties want different forms of control, not because it is cheaper in absolute terms.

PUE, the Lease and Where Risk Actually Sits

Power Usage Effectiveness is a lease term before it is an efficiency metric, because the power pass-through in a data centre lease is frequently larger than the base rent.

PUE is total facility power divided by power delivered to IT equipment. A PUE of 1.5 means half a watt of overhead for every watt of computing.

The benchmarks

The Uptime Institute's 2024 survey reported an average PUE of 1.58 across surveyed data centres, down from roughly 2.0 ten years earlier. Hyperscale operators such as Google report fleet-wide PUE around 1.10. Most enterprise facilities operate between 1.4 and 1.8.

Below 1.4 is good for a traditional enterprise facility. Below 1.2 is excellent and typically requires free cooling or liquid cooling.

Water Usage Effectiveness (WUE) is the equivalent water metric, and it matters increasingly where cooling water is constrained or regulated.

The PUE cap

A lease may include a PUE cap: if the facility's PUE exceeds, say, 1.4, the tenant is not responsible for the inefficiency above that threshold. That shifts operational risk back to the landlord and gives the landlord a reason to run the building efficiently.

This is where landlord and tenant incentives collide, and it is the term most worth negotiating in a lease where the power pass-through exceeds the rent.

Lease structure

Powered shell lease rates run roughly $60 to 110 per kW per month on a shell basis, on terms of 10 to 20 years with escalators. Deal sizes typically run from 10 MW to over 100 MW. Cap rates for powered shell have been quoted around 7.0 to 8.0 percent, against 6.0 to 7.0 percent for wholesale colocation, reflecting the risk the tenant carries.

Note the unit. Data centre space is priced per kilowatt, not per square foot. The product is power.

If PUE exceeds the cap

Without a cap, the tenant pays for the landlord's inefficiency for the full term. On a twenty-year lease where the pass-through exceeds the rent, a 0.2 difference in PUE is a material sum, and it is fixed at signature.

Shell Specification: What to Write Down

A shell that misses the tenant's cooling load, redundancy standard or equipment density forces redesign, and that risk sits with the tenant.

Six specifications determine whether a shell works for the intended load.

Secured power capacity, in MW, with the utility service agreement status and energisation date.

Structural floor loading, expressed in pounds per square foot or kN/m². High-density racks, liquid cooling distribution units and battery rooms all impose loads a conventional warehouse slab was not designed for.

Clear height and column spacing, which determine whether hot aisle containment, overhead busway and cooling distribution can be installed as planned.

Space provision for mechanical and electrical plant, including generator yard, fuel storage, chiller or CDU space, and electrical rooms.

Design rack density, in kW per rack, because it determines the cooling architecture and therefore the space and the structure.

Expansion provision, in both floor area and power, with the utility capacity headroom stated rather than assumed.

For phased growth planning, see our guide to scalable data centre solutions.

Power Infrastructure

In a powered shell, the landlord delivers the kW of utility capacity, the structural building and the basic distribution. The tenant builds out the UPS, the generators and the white space fit.

Electrical distribution on the landlord side covers the utility feeders, the substation and medium-voltage switchgear up to the demarcation point.

Uninterruptible power supply is tenant scope in a conventional powered shell. The landlord provides the space and the structural and electrical provision; the tenant selects topology, runtime and redundancy.

Backup generation is tenant scope. Diesel generators remain the standard; gas turbines and reciprocating engines appear on larger sites and where fuel supply and emissions permitting allow. NFPA 110 governs emergency and standby power system classification by type, class and level.

Critical load is the power delivered to IT equipment. IT load is the same quantity viewed from the tenant's side. Every capacity figure in a lease should state which it refers to and at which measurement point.

For generation technology options, see our guide to power generation equipment compared, and for data centre supply specifically, our guide to data centre power sources. Where storage supports ride-through or peak management, see our guide to battery energy storage systems.

Density and Cooling: Where the Design Decision Actually Sits

Rack density determines cooling architecture, and the threshold where air cooling stops working is a question of airflow rather than of temperature.

What density actually looks like today

The Uptime Institute's 2025 survey puts the average of operators' most common rack densities at nearly 9 kW, with 4 to 5 kW the single most frequent answer, and 82 percent of facilities peaking below 30 kW even in their densest racks.

Current AI racks draw 125 to 140 kW and ship liquid-cooled by design.

Those two facts describe different buildings. A shell specified for a 9 kW estate cannot host a 125 kW AI deployment without redesign, and a shell specified for AI is overbuilt and oversized for conventional enterprise load.

Where air cooling stops

Air cooling runs out at around 40 to 50 kW per rack, and the constraint is airflow rather than any thermal limit.

ASHRAE TC 9.9 makes the argument in cubic feet per minute: a 40 to 50 kW rack needs up to 5,000 cfm, while a best-of-breed raised floor delivers roughly 1,900 cfm per floor tile. A 60 kW rack needs roughly 10,000 cfm at a 20°F temperature differential.

Published thresholds vary from 25 to 50 kW per rack, and the variation is real rather than disagreement. It depends entirely on how much air the floor can actually deliver and how well the containment works. With good hot and cold aisle containment, air cooling reliably covers up to about 25 to 30 kW per rack. Beyond that, the required air volume becomes impractical.

The cooling options

CRAC and CRAH units are the traditional raised-floor approach, computer room air conditioners and air handlers respectively, effective in moderate climates and at conventional densities.

In-row and rear-door cooling move the heat exchange closer to the rack, extending the usable air range.

Direct-to-chip liquid cooling places a cold plate on the processor, with a coolant distribution unit (CDU) between the facility water loop and the technology cooling loop. It adds $200 to $500 per kW of IT capacity in infrastructure cost and reduces cooling energy costs by 30 to 50 percent at high densities.

Immersion cooling submerges server components in dielectric fluid. It achieves the highest densities and imposes the most specific requirements on structure, fluid handling and serviceability.

Free cooling and economisers use ambient conditions to reduce or eliminate mechanical cooling, and they are the primary route to a PUE below 1.2.

Liquid cooling adoption is rising quickly, with AFCOM's 2025 State of the Data Center reporting it nearly doubling year over year, from 19 to 36 percent of facilities.

The shell consequence

The cooling choice determines the building. Liquid cooling needs pipework routes, CDU space and floor loading for fluid-filled equipment. Immersion needs fluid handling, containment and very different serviceability provision. Air cooling needs floor plenum depth and airflow path.

Specifying a shell without naming the design density is specifying the wrong building.

ASHRAE TC 9.9 and the Thermal Envelope

ASHRAE Technical Committee 9.9 Thermal Guidelines for Data Processing Environments is the industry reference for environmental classes and thermal operating envelopes.

The 2021 fifth edition establishes air-cooled classes A1 through A4 and introduces class H1 for liquid-cooled high-density racks.

ASHRAE TC 9.9 recommends 18 to 27°C at the equipment inlet for air-cooled classes A1 through A4. Progression toward A4 allows wider allowable envelopes, up to 45°C in the widest class, which enables free-air economisation and improved PUE at the cost of reduced thermal margin during transient events.

High-density class H1 equipment requires a narrower 18 to 22°C, with coolant supply typically specified in that range.

One caveat worth knowing. The thermal density assumptions behind the 2021 guidelines were roughly 40 to 60 kW per rack. For current liquid-cooled AI silicon, OEM cold-plate specifications increasingly govern in practice.

ASHRAE 90.4, Energy Standard for Data Centers, is the energy compliance standard, and it uses mechanical load component and electrical loss component rather than PUE. PUE is an operational benchmark, not a compliance metric.

Redundancy and Tier Classification

Tenants specify redundancy against a tier standard, and the shell must provide the space, power paths and structural provision for the level they intend to build.

The Uptime Institute Tier classification, Tiers I through IV, defines data centre infrastructure by capability rather than by component count. Tier III introduces concurrent maintainability, meaning any capacity component can be removed from service without affecting IT operation. Tier IV adds fault tolerance, meaning a single unplanned failure does not affect IT operation.

N+1, 2N and 2N+1 describe the redundancy configuration of a specific system. N is the capacity required to carry the load. N+1 adds one spare unit. 2N duplicates the entire system.

The shell decides what is achievable. Dual utility feeds, separated generator yards and independent distribution paths are structural and spatial decisions, not fit-out decisions. A shell with a single incoming path cannot host a 2N electrical design.

Commissioning

Redundancy claims are proven at commissioning, not at design. Integrated systems testing under full simulated load is where N+1 and 2N designs either demonstrate their capability or do not. The shell lease should state whose scope that is and what access the tenant has.

For what each redundancy level actually requires to build and to prove, see our guide to data centre power redundancy, N, N+1, 2N and 2N+1.

Safety, Fire and Governing Standards

Five standards govern data centre fire protection, power systems and energy performance in North America, and a shell specification should name them.

Fire detection and suppression, including clean agent systems and very early smoke detection, are design drivers rather than add-ons, and the structural and spatial provision for them is shell scope.

Physical security, covering perimeter, access control and surveillance, is a shell design consideration because retrofitting it into an occupied facility is substantially harder.

Insurance

Insurers assess fire protection design, redundancy configuration and commissioning evidence when pricing data centre risk. A shell delivered without documented compliance evidence becomes the tenant's problem at underwriting, which is after the lease is signed.

Outside the United States

NFPA standards are North American. ASHRAE TC 9.9 thermal guidelines and the ISO/IEC 22237 series for data centre facilities and infrastructure apply more widely, and European installations work to EN standards. Local building codes, seismic requirements and water use regulation all vary and all affect shell design.

Network and Connectivity Readiness

Planning network integration at the shell stage reduces deployment delays and supports operational scalability, without expanding the project scope beyond core infrastructure.

Four items belong in a shell specification:

Predefined entry points for fibre and network cabling, with diverse physical routes into the building.

Structured pathways that simplify future cabling installation without core drilling an occupied facility.

Dedicated space for telecom and network equipment, usually a meet-me room.

Redundant routing options supporting uptime requirements, with the number of distinct carriers and physical entry paths stated.

Carrier diversity is a shell attribute, not a fit-out choice. A building served by one physical route cannot be made diverse after construction without civil works.

Who Buys Powered Shells

Six buyer types use the model, and each is solving a different problem.

Data sovereignty is a recurring driver. Regulations requiring data to remain within a jurisdiction force local build, and a powered shell is the fastest route to local capacity that the tenant still controls.

Where Tenants Get Caught

Powered shell is not always lower risk, and four failure modes recur.

The shell misses the cooling load. Designed for a conventional estate, leased to a tenant planning high-density deployment. The redesign cost and the delay sit with the tenant.

Density assumptions change mid-term. A tenant signing for 9 kW racks and deploying AI hardware three years later finds the building cannot carry the load, the airflow or the floor weight.

The redundancy standard differs. A shell built with one incoming utility path cannot host a 2N design, and that is discovered during tenant design rather than during lease negotiation.

The power is not energised. A shell marketed on secured capacity where the capacity is a queue position, not a connection. This is the one that cannot be engineered around.

How to avoid all four

Specify the design density, the redundancy standard and the cooling architecture in the lease, and verify the utility service agreement independently. Those four items, confirmed before signature, eliminate most of the risk the model carries.

Which Model Fits

Choose powered shell when you have the capital, the capability and a specific technical requirement. Choose turnkey when you have none of those.

Powered shell suits tenants with in-house engineering capability, a specific density or redundancy requirement, a long horizon, and the capital to fund a fit-out.

Turnkey suits tenants who want capacity rather than a building, who do not want an operating role, and who are willing to pay a higher rate per kW for the landlord to carry the capital and the service level.

Retail colocation suits smaller requirements where the operator provides everything including the operational service.

The staffing consequence is real. A powered shell tenant is building and then operating a facility. That is a department, not a contract.

What Prismecs Does

Prismecs delivers power infrastructure for mission-critical facilities, covering engineering, procurement, construction management, installation, commissioning and operations.

Delivered project scope includes eight TM2500 dual-fuel units totalling 260 MW at Birr, Switzerland, built as a fast-track plant online in six months on a compact site with a new 220 kV interconnection and engineered noise controls; four TM2500 units totalling 110 MW at Duqm, Oman, kept grid-ready with resident O&M crews, CMMS and parts support; an LM2500XPRESS plant at Miaoli, Taiwan delivered in ten months; DC-coupled battery energy storage for solar and hybrid projects; and owner's engineering on a 7 MW, 28 MWh battery energy storage system including the utility interconnection application.

Birr is the relevant reference for anyone whose constraint is time. Two hundred and sixty megawatts, a new 220 kV interconnection, six months from award to operation. That is the same problem a powered shell exists to solve, approached from the generation side.

Capability spans data centre solutions for the vertical, EPCM services for project delivery, power generation asset services for on-site and backup generation, distributed energy solutions for microgrids and storage, I&C services for electrical systems, controls and commissioning, owner's engineering for independent review, and O&M services for the operating phase.

Apply this article's criteria to any shell you are offered, including anything we propose. Ask whether the utility service agreement is executed and the capacity energised. Ask what design rack density the building was specified for. Ask where the demarcation point sits, by location and by equipment. Ask whether a 2N electrical design is physically achievable in the building as built.

To discuss power availability, shell specification or on-site generation for a data centre project, send your target capacity in MW, site location, design rack density, redundancy standard and required in-service date to sales@prismecs.com or call +1 (888) 774-7632.

Frequently Asked Questions

What is a powered shell data center?

A secure facility with the physical structure, power to the site and connectivity options, excluding major IT infrastructure such as UPS systems, generators and cooling systems. The tenant funds and completes the interior fit-out. It sits between raw land and turnkey capacity, and the scope boundary is the most consequential term in the transaction because market practice varies on what the landlord actually delivers.

What is included in a powered shell and what is excluded?

Included: the building, utility power to the site, usually the substation and basic distribution, and connectivity pathways. Excluded: UPS systems, generators, tenant-specific power distribution, final cooling configuration, server racks, cabling and IT equipment. Cooling plant varies by deal, with some shells including it and others providing only the space and structural provision, so it must be written into the term sheet.

What is the difference between powered shell and turnkey?

Who funds the fit-out and who carries the operational risk. In a powered shell the tenant funds the interior, taking on roughly 60 to 70 percent of total fit-out cost, and gains control over density, redundancy and cooling. In turnkey the landlord funds a complete redundant facility to a contracted service level and charges a higher rate per kilowatt. Turnkey requires roughly three to five times the capital per megawatt.How fast can a powered shell be deployed?

Typically 6 to 12 months from occupancy to operation, against 18 to 36 months for a ground-up build. That range assumes the power is secured and energised, and it describes the fit-out period rather than the development period. The saving comes from the building existing, the utility capacity being arranged, and the landlord's long-lead electrical equipment being procured, not from the fit-out itself being quick.

Why do powered shells exist?

Because utility power cannot be accelerated with money. Interconnection waits approach seven years in some US markets, ERCOT has tracked roughly 474 GW of large-load requests, and transformers run 90 to 130 weeks. A building can be constructed in eighteen months; a utility connection cannot be bought in eighteen months at any price. The asset is the secured capacity, and the walls are the delivery mechanism.

What does a data center cost per megawatt?

Turnkey construction runs roughly $8 to 12 million per megawatt, and $15 to 20 million or more per megawatt for AI-optimised facilities designed for rack-scale density. The 2026 institutional benchmark for a hyperscale shell-and-core build, assuming standard cloud-compute load and 1.20 PUE design, runs $11 to 14 million per IT megawatt per JLL and Cushman & Wakefield pricing outlooks.

What is a typical powered shell lease rate?

Roughly $60 to 110 per kilowatt per month on a shell basis, on terms of 10 to 20 years with escalators, in deal sizes from 10 MW to over 100 MW. Cap rates for powered shell have been quoted around 7.0 to 8.0 percent against 6.0 to 7.0 percent for wholesale colocation, reflecting the risk the tenant carries. Data centre space is priced per kilowatt, not per square foot.

What is PUE and what is a good value?

Power Usage Effectiveness is total facility power divided by power delivered to IT equipment. The Uptime Institute's 2024 survey reported an average of 1.58 across surveyed facilities, down from roughly 2.0 a decade earlier, with hyperscale operators such as Google reporting around 1.10. Below 1.4 is good for enterprise; below 1.2 is excellent and typically requires free cooling or liquid cooling.

What is a PUE cap in a data center lease?

A lease term stating that if facility PUE exceeds a defined threshold, say 1.4, the tenant is not responsible for the inefficiency above it. It shifts operational risk to the landlord and gives the landlord a reason to run the building efficiently. It matters because the power pass-through in a data centre lease is frequently larger than the base rent itself.

What rack density do data centers actually run at?

Lower than the AI headlines suggest. The Uptime Institute's 2025 survey puts the average of operators' most common rack densities at nearly 9 kW, with 4 to 5 kW the single most frequent answer, and 82 percent of facilities peaking below 30 kW even in their densest racks. Current AI racks, by contrast, draw 125 to 140 kW and ship liquid-cooled by design.

At what density does air cooling stop working?

Around 40 to 50 kW per rack, and the constraint is airflow rather than any thermal limit. A 40 to 50 kW rack needs up to 5,000 cfm while a best-of-breed raised floor delivers roughly 1,900 cfm per tile, and a 60 kW rack needs roughly 10,000 cfm at a 20°F differential. With good hot and cold aisle containment, air cooling reliably covers up to about 25 to 30 kW.

What does liquid cooling cost and when is it justified?

Direct-to-chip liquid cooling adds roughly $200 to $500 per kW of IT capacity in infrastructure cost and reduces cooling energy costs by 30 to 50 percent at high densities. It is justified above the air cooling ceiling, where the required airflow becomes impractical regardless of design quality. Adoption is rising quickly, nearly doubling year over year from 19 to 36 percent of facilities per AFCOM's 2025 survey.

What does ASHRAE TC 9.9 specify?

Environmental classes and thermal operating envelopes for data processing environments. The 2021 fifth edition establishes air-cooled classes A1 through A4, recommending 18 to 27°C at equipment inlet, and introduces class H1 for liquid-cooled high-density racks requiring a narrower 18 to 22°C. The thermal density assumptions behind it were roughly 40 to 60 kW per rack, so OEM cold-plate specifications increasingly govern current AI hardware.

What goes wrong with powered shell deals?

Four things. The shell misses the tenant's cooling load, forcing redesign at tenant cost. Density assumptions change mid-term and the building cannot carry the load, airflow or floor weight. The redundancy standard differs, and a single incoming utility path cannot host a 2N design. And the power is not actually energised, which is the only one that cannot be engineered around.

Tags: Powered Shell Data Centers Data Center Design Data Center Power Rack Density Cooling Systems