Evaluation Criteria
Tier Certification and Availability
Verify Uptime Institute Tier certification — specifically, whether the facility has Design, Constructed Facility, and Operational Sustainability certifications. Design certification alone does not guarantee operational performance. Ask for the facility's actual availability history over the past 3–5 years.
Power Density Capability
The most commonly overlooked criterion. Verify the maximum kW per cabinet the facility can support, and whether liquid cooling is available for AI-dense deployments. Many older facilities are limited to 5–10 kW per cabinet — insufficient for modern workloads.
Power Redundancy
Verify the power redundancy architecture: N+1 or 2N UPS, number of utility feeds, generator capacity and redundancy. Ask specifically: what happens to your cabinet if a UPS module fails? If a generator fails to start?
Connectivity
Number of carriers, diverse fiber entry points, available bandwidth, and cross-connect options. For latency-sensitive applications, verify the facility's network latency to your key locations.
Physical Security
Verify security controls: perimeter security, building access controls, data center floor access (mantrap, biometric), CCTV coverage, and security monitoring. Ask about security incident history and how incidents are disclosed to customers.
Geographic Location
Evaluate natural disaster risk (flood, seismic, tornado, hurricane), distance from your primary operations, and regulatory jurisdiction. For disaster recovery sites, ensure sufficient geographic separation from primary sites.
Technical Due Diligence
Facility Tour
A physical tour of the facility is essential. Observe: cleanliness and organization (indicator of operational discipline), cable management quality, visible equipment age and condition, hot/cold aisle containment implementation, and staff professionalism. A poorly maintained facility is a red flag regardless of certifications.
Documentation Review
Request and review: single-line electrical diagrams, mechanical system drawings, Tier certification documents, SOC 2 Type II audit report, and recent maintenance records for critical systems. Reluctance to provide documentation is a warning sign.
Reference Customers
Request 3–5 reference customers with similar requirements. Ask specifically: Have you experienced any outages? How did the provider communicate during incidents? How responsive is the support team? Would you renew your contract?
Capacity Assessment
Verify that the facility has sufficient available capacity for your current and projected requirements. Ask about the facility's current utilization rate and expansion plans. A facility at 90% capacity has limited ability to accommodate your growth.
Contract Terms
SLAs and Remedies
Verify SLA commitments for power availability, cooling, and network connectivity. Understand the remedy structure — most SLAs provide service credits, not actual damages. Verify that the SLA covers the specific failure scenarios you are concerned about.
Term and Termination Rights
Data center contracts typically run 1–5 years. Understand early termination penalties and the conditions under which you can exit without penalty (provider breach, facility closure, acquisition). Negotiate termination rights for material SLA failures.
Expansion Options
Negotiate rights of first refusal for adjacent space and pre-agreed pricing for expansion. Without expansion rights, you may be forced to relocate when you outgrow your initial footprint.
Price Escalation
Understand how pricing can change during the contract term. Common escalation mechanisms: CPI-linked increases, power cost pass-through, and renegotiation at renewal. Cap annual escalation in the contract.
Change of Control
Negotiate change-of-control provisions that give you termination rights if the provider is acquired. Data center consolidations following acquisitions can result in facility closures or service degradation.
Provider Types
Wholesale Colocation
Large footprints (typically 250+ kW) leased as dedicated space. Customer owns all IT equipment and manages their own infrastructure within the leased space. Lowest cost per kW; requires significant operational capability.
Retail Colocation
Individual cabinets or small suites in a shared facility. Provider manages facility infrastructure; customer manages IT equipment. Higher cost per kW than wholesale but lower minimum commitment. Appropriate for organizations without large-scale data center requirements.
Managed Hosting
Provider owns and manages both facility and IT infrastructure. Customer pays for compute, storage, and network capacity. Highest cost but lowest operational burden. Appropriate for organizations without IT infrastructure expertise.
Build vs. Buy Decision
The build-vs-buy decision for data center infrastructure depends on scale, control requirements, and organizational capability:
- Build (own facility): Maximum control, lowest long-term cost at scale (10+ MW), required for highest security requirements. Requires significant capital, construction expertise, and operational capability.
- Colocation: Enterprise-grade facilities without capital cost of construction. Appropriate for most enterprise deployments. Less control than owned facility.
- Cloud: No facility investment; pay-as-you-go. Highest cost at scale; appropriate for variable workloads and organizations without infrastructure expertise.
Most enterprises use a hybrid approach: colocation for primary infrastructure, cloud for burst capacity and development workloads.
Common Procurement Mistakes
- Selecting on price alone: The cheapest facility is rarely the best value when total cost of ownership and risk are considered
- Ignoring power density limits: Discovering that the facility cannot support your density requirements after signing a contract is expensive
- Not verifying operational track record: Design certifications do not guarantee operational performance
- Insufficient contract review: SLA remedies, termination rights, and expansion options are negotiable — engage legal counsel
- Single-site strategy: A single data center location creates geographic concentration risk; plan for disaster recovery from the start
- Underestimating growth: Negotiate expansion rights before you need them; available capacity disappears quickly in high-demand markets
RFP Process
A structured RFP process ensures consistent evaluation across providers:
- Requirements definition: Document power, cooling, space, connectivity, security, and compliance requirements
- Market survey: Identify qualified providers in target markets
- RFI (Request for Information): Gather basic capability information to qualify providers
- RFP issuance: Detailed requirements and pricing request to qualified providers
- Proposal evaluation: Score proposals against defined criteria
- Site visits: Physical tours of shortlisted facilities
- Reference checks: Conversations with existing customers
- Contract negotiation: Negotiate terms with preferred provider(s)