Reach Data Center Decision MakersAdvertise

What Is a Powered Shell Data Center? The Emerging Model for AI Infrastructure

AEO Summary: A powered shell data center is a facility that provides building infrastructure (structural envelope, utility power feed, electrical distribution to demarcation point, fire suppression, and physical security), but leaves cooling, networking, and IT fit-out to the tenant. It bridges the gap between raw land (high customization, 18-24 month deployment) and turnkey colocation (ready in weeks, less customization). Powered shell economics: $80-120/kW/month, 3-6 month fit-out timeline. Ideal for AI neoclouds, hyperscalers, and enterprises with 1MW+ deployments requiring custom infrastructure.

Introduction: Why Powered Shells Are Becoming the Default for AI Infrastructure

Five years ago, data center deployment had two primary paths: turnkey colocation (rent space in someone else’s facility) or ground-up construction (buy land, build from scratch).

In 2026, there’s a third option gaining traction: powered shell colocation. It’s neither fully managed nor fully custom. The operator provides the building and electrical infrastructure; you bring in your own cooling, networking, and IT equipment.

This model is emerging specifically because of AI infrastructure demand. AI workloads demand custom cooling solutions (liquid cooling for GPU clusters), custom power distribution (to handle extreme density), and custom networking (neoclouds manage their own routing and carrier relationships). Turnkey colocation can’t accommodate these requirements. But ground-up construction is too slow. Neoclouds need to deploy in 6-12 months, not 24 months.

Powered shell fills that gap. It’s fast enough for neoclouds, customizable enough for extreme AI workloads, and economical for large deployments. By 2027, powered shells will likely be the dominant deployment model for AI infrastructure operators.

What’s Included in a Powered Shell Data Center

A powered shell facility provides:

Building & Structure

  • Weather-resistant building envelope (roof, walls, floors)
  • Seismic design appropriate to geographic region
  • Loading capacity for heavy equipment (GPUs, cooling systems, power distribution)
  • Fire-rated construction where required by code

Utility Power Feed

  • Connection to utility grid (one or more feeds, depending on design)
  • Transmission from utility hand-off point to facility boundary

Electrical Distribution

  • Primary transformer and switchgear
  • Distribution panels and subfeeders to demarcation point (where the tenant’s infrastructure begins)
  • Backup power infrastructure (if included; see below)
  • Metering for power consumption

Fire Suppression & Safety

  • Fire detection system (sprinklers or gaseous suppression)
  • Emergency egress (emergency exits, emergency lighting)
  • Building control systems (HVAC, lighting, access control)

Physical Security Shell

  • Perimeter fencing or walls
  • Entry gates and access control
  • Basic lighting
  • Structural security (locked roof hatches, no unauthorized entry points)

What’s NOT Included (Tenant Responsibility)

IT Cooling The tenant is responsible for deploying cooling infrastructure capable of handling their equipment’s heat load. Options include:

  • Liquid cooling (for GPU clusters, extreme density)
  • In-row air cooling units
  • Precision air conditioning (PAC) systems
  • Custom cooling architecture

Networking Infrastructure

  • Network cable plant (fiber and copper)
  • Network equipment (routers, switches, cross-connect infrastructure)
  • Carrier connections (fiber runs from carrier hand-off points to network equipment)
  • Internet gateway / DDoS mitigation

IT Equipment & Fit-Out

  • Server racks, cables, patch panels
  • Network switches, routers, security appliances
  • Power distribution units (PDUs)
  • Monitoring and management systems
  • All customer-owned IT equipment

Operations & Management

  • 24/7 facility operations (if needed; some operators staff themselves)
  • Remote hands support
  • Facility management and maintenance
  • Security monitoring and incident response (unless outsourced)

Who Uses Powered Shells

Neoclouds (CoreWeave, Lambda Labs, Crusoe Energy)

Neoclouds are dedicated GPU and accelerator providers. They need custom infrastructure to optimize for their specific workload (GPU density, liquid cooling, high-speed interconnect) and carrier architecture (direct connections to multiple carriers, custom routing).

Turnkey colocation is too restrictive (can’t deploy custom cooling, can’t manage carriers independently). Ground-up construction is too slow. Powered shells let neoclouds deploy a 10MW GPU cluster in 6 months rather than 18+ months.

Hyperscalers Expanding into New Markets

Google, Amazon, and Microsoft are expanding data center presence into secondary markets (tier 2 cities like Salt Lake City, Nashville, Kansas City). They already have operational expertise, vendor relationships, and internal engineering teams. They just need the building and power.

A powered shell deployment in a new market is faster and lower-risk than acquiring an existing turnkey facility. If the market underperforms, they can repurpose the building or sell it to another operator.

Large Enterprises with Custom Requirements

Enterprise IT teams sometimes have specific infrastructure requirements (dedicated cooling for high-security networks, custom air-gapping, specific power architecture for failover). Turnkey facilities can’t accommodate these. Powered shells let enterprises build exactly what they need.

GPU Cluster Operators & AI Training Specialists

Companies deploying large GPU clusters for model training (research labs, AI startups, enterprise AI centers) often need custom cooling and power distribution that turnkey facilities can’t support. Powered shells enable these deployments.


Powered Shell vs. Turnkey Colocation vs. Build-to-Suit

DimensionPowered ShellTurnkey ColocationBuild-to-Suit
Cost per kW/month$80-120$130-200$150-250 (construction cost amortized)
Deployment timeline3-6 months2-8 weeks12-24 months
CustomizationHigh (cooling, networking)Limited (standard rack/cage)Complete (design from scratch)
Fit-out responsibilityTenantOperatorTenant or operator (negotiated)
Operational complexityHigh (tenant staffs own ops)Low (operator manages)High (custom systems)
Power densityUp to 500+ kW per facilityLimited (typically 5-25 kW per rack)No limit (design dependent)
Best forNeoclouds, hyperscalers, large AI operatorsSmall to mid-market deploymentsCustom enterprise requirements

Build vs. Lease Economics: When Powered Shell Makes Sense

Powered Shell Lease Economics

Monthly cost: $80-120 per kW of deployment capacity Annual cost (10MW deployment): $9.6M – $14.4M 3-year lease: $28.8M – $43.2M

Pros:

  • No capital expenditure for building
  • No real estate risk (if market changes, you can leave at lease end)
  • Operator handles building maintenance, utilities, real estate taxes
  • Faster to deploy than ground-up construction

Cons:

  • Recurring lease cost (no equity buildup)
  • Limited customization options (building structure is fixed)
  • Lease termination penalties for early exit

Ground-Up Construction Economics

Construction cost: $150-250 per square foot Total for 10MW facility: $300,000-500,000 (assuming 2,000-3,000 sq ft) Buildout timeline: 12-24 months Ongoing operational cost: $1.2M-2.4M annually (utilities, maintenance, staffing) 5-year TCO: $6M-12M operational + construction

Pros:

  • Long-term asset ownership (potential appreciation)
  • Complete customization
  • No lease restrictions or termination penalties

Cons:

  • High upfront capital expenditure
  • High execution risk (construction delays, cost overruns)
  • Long deployment timeline (12-24 months)
  • Residual real estate risk (what if you need to shut down or relocate?)
  • Operational burden (you manage everything)

Break-Even Analysis

For a 10MW deployment:

  • Powered shell: 3-year cost = $28.8M-$43.2M
  • Ground-up construction: 3-year cost = $6M-$12M (operational) + $300K-$500K (construction)

Ground-up construction is cheaper over 3 years IF construction completes on time. But if deployment is delayed, powered shell becomes cheaper (delay costs money and delays revenue).

For a 1MW or smaller deployment, powered shell is almost always cheaper than ground-up construction.


When Powered Shell Makes Sense; When It Doesn’t

Powered Shell Is the Right Choice If:

Your deployment is 1MW or larger. Below 1MW, turnkey colocation is simpler and cheaper. ✓ You need custom cooling. If your workload is 50+ kW per rack (GPU clusters, AI training), you need liquid or advanced cooling that turnkey can’t provide. ✓ You need to deploy in 6 months or less. Powered shell is 3-6x faster than ground-up construction. ✓ You have in-house infrastructure expertise. You’re staffing your own ops team, managing your own cooling/networking systems. If you don’t have this expertise, turnkey colocation is better. ✓ You want to minimize capital expenditure. Powered shell lease is cheaper than buying land and constructing from scratch. ✓ Your workload requires high power density. If you need 500+ kW per facility, turnkey won’t accommodate you. Powered shell will. ✓ You need geographic redundancy fast. Deploying in multiple regions? Powered shell is faster and lower-risk than ground-up builds in each region.

Powered Shell Is NOT the Right Choice If:

Your deployment is under 500 kW. Turnkey colocation is simpler, cheaper, and has lower operational overhead. ✗ You don’t have an in-house infrastructure team. Powered shells require tenant operations. If you don’t have the expertise, turnkey colocation (where the operator manages everything) is better. ✗ You need immediate deployment (under 3 months). Turnkey colocation is faster. Powered shell requires 3-6 months for fit-out. ✗ You want to minimize operational complexity. Turnkey is simpler. You rent space; they manage the rest. Powered shell, you’re managing most of the infrastructure. ✗ Your workload is standard compute (CPUs, not GPUs). Standard servers work fine in turnkey colocation. Powered shell’s advantages don’t apply.


The Powered Shell Market in 2026

Demand

Powered shell demand is driven by AI infrastructure build-out. CoreWeave, Lambda Labs, and Crusoe have all announced powered shell deployments in major markets. Hyperscalers are expanding powered shell availability as part of their capacity growth.

Supply

Supply is constrained. Most existing data center operators specialize in turnkey colocation. Powered shell is a newer model and requires different operational expertise.

Operators entering the powered shell market in 2026:

  • Hyperscalers (Google, Amazon, Microsoft)
  • Traditional operators (Equinix, CyrusOne, Digital Bridge) pivoting to powered shell
  • New operators focused specifically on powered shell for AI (e.g., dedicated GPU hosting platforms)

Pricing

Powered shell pricing ranges widely by market, facility quality, and tenant requirements:

  • Premium markets (San Francisco, Northern Virginia): $120-180/kW/month
  • Secondary markets (Salt Lake City, Dallas, Columbus): $80-120/kW/month
  • Emerging markets (Nashville, Kansas City): $70-100/kW/month

Pricing is expected to decline as supply increases and the market matures.


The Powered Shell Fit-Out Process: What to Expect

Phase 1: Lease & Design (Month 1-2)

  • Negotiate lease terms (rate per kW, term length, expansion options)
  • Complete facility design (cooling architecture, power distribution, networking layout)
  • Obtain permits and utility sign-offs
  • Order long-lead equipment (cooling systems, power distribution, network infrastructure)

Phase 2: Infrastructure Deployment (Month 2-4)

  • Install power distribution (electrical work, subfeeders, metering)
  • Deploy cooling infrastructure (liquid cooling loops, in-row units, or PAC systems)
  • Install cable plant (network fiber and copper)
  • Connect to carriers (fiber hand-offs from carriers to network equipment)

Phase 3: Testing & Commissioning (Month 4-6)

  • Test power systems under load
  • Test cooling under simulated operating load
  • Test network connectivity and performance
  • Run security and compliance audits
  • Commission facility and begin operations

Critical Timeline Items

  • Power utility connection: 6-12 weeks (plan early)
  • Carrier fiber provisioning: 6-12 weeks (plan early)
  • Equipment procurement: 8-16 weeks for specialized cooling and power systems
  • Building permits: 2-8 weeks depending on jurisdiction
  • Construction/installation: 4-8 weeks depending on complexity

Key lesson: Order long-lead items (utilities, carriers, specialized equipment) immediately upon lease signing. Delays here cascade into overall timeline delays.


FAQ: Powered Shell Data Centers

Q: What’s the difference between a powered shell and a “white box” data center? A: Powered shell and white box are often used interchangeably. Both refer to a facility where the operator provides building and power, and the tenant provides cooling, networking, and IT equipment. Some vendors use “white box” to imply more stripped-down (less building infrastructure included). For comparison purposes, treat them as the same.

Q: Who typically manages facility operations in a powered shell? A: It depends on the agreement. Some tenants staff their own 24/7 operations team (especially neoclouds and hyperscalers). Others hire a third-party facility management company. The operator usually provides utilities, basic building maintenance, and security perimeter. Cooling, networking, and IT operations are tenant responsibility.

Q: How long does it take to fit out a powered shell? A: 3-6 months for a standard deployment. Complex deployments (e.g., custom liquid cooling, multi-carrier redundancy) can take 6-9 months. Timeline is driven by long-lead items (power utility connection, carrier fiber, specialized cooling equipment).

Q: What happens if I outgrow my powered shell deployment? A: You have several options: (1) Expand within the same facility if space is available, (2) Lease additional powered shell space in the same operator’s portfolio, (3) Build a second facility in a different market, (4) Migrate some workloads to turnkey colocation in other locations for redundancy.

Q: How much does it cost to fit out a powered shell? A: Highly variable depending on requirements. Basic fit-out (standard cooling, standard networking): $200,000-$500,000 for a small facility. Large deployments with custom cooling and multi-carrier networking: $2M-$10M+. Get quotes from integrators familiar with powered shell deployments.

Q: What if I need to relocate or terminate the lease early? A: Most powered shell leases have early termination clauses with penalties (typically 6-12 months of remaining lease payments, or a percentage of remaining term). Negotiate favorable early termination terms during lease negotiations. Some operators offer “migration services” to move equipment to a different facility within their portfolio at reduced cost.

Q: Is powered shell suitable for smaller deployments (under 1MW)? A: It’s technically possible but rarely cost-effective. Turnkey colocation is cheaper and simpler for deployments under 500kW. Powered shell economics favor larger deployments (1MW+).

Q: How do I find powered shell operators in my region? A: Start by contacting hyperscalers (Google, Amazon, Microsoft) in your region to ask about powered shell availability. Contact traditional operators (Equinix, CyrusOne) to ask about powered shell offerings. Search for “neocoloud partners” or “powered shell data center” in your target market.


Conclusion: Powered Shell as the Infrastructure Model of Choice for AI

Powered shell data centers represent a fundamental shift in how AI infrastructure is deployed. They’re neither turnkey (too restrictive) nor ground-up builds (too slow). They’re the Goldilocks option: fast enough, customizable enough, and economical for large-scale AI operators.

If you’re deploying 1MW+ of AI infrastructure in 2026 or beyond, powered shell should be your default model. It will reduce your time-to-deployment from 18+ months to 6 months, lower your capital expenditure compared to ground-up builds, and give you the customization you need for extreme AI workloads.

Want to explore powered shell options for your AI infrastructure? Visit GoDataCenters.com/get-a-quote to compare powered shell availability across 50+ markets, get facility recommendations matched to your specific power and cooling requirements, and connect with operators offering powered shell leases at competitive rates.

Sourcing capacity?

One requirement, matched against 4,562 facilities and 1,010 providers worldwide.

Get a quote