What Is Colocation?
Colocation (often abbreviated “colo”) is the practice of housing your servers, networking equipment, and storage hardware in a third-party data center facility. Instead of building and maintaining your own data center, a capital-intensive proposition that can cost $10M+ for even a modest facility, you rent space, power, and cooling from a colocation provider.
Enterprise buyers choose colocation for several reasons: predictable monthly costs, access to carrier-neutral connectivity, physical security, and the ability to scale capacity without building new facilities.
Key Evaluation Criteria
When comparing colocation facilities, focus on these critical factors:
Tier Rating
The Uptime Institute’s tier classification system rates data centers from Tier 1 (basic) to Tier 4 (fault tolerant):
- Tier 1: Single path for power and cooling, 99.671% uptime (28.8 hours downtime/year)
- Tier 2: Redundant components, 99.741% uptime (22 hours downtime/year)
- Tier 3: Concurrently maintainable, 99.982% uptime (1.6 hours downtime/year)
- Tier 4: Fault tolerant, 99.995% uptime (0.4 hours downtime/year)
For mission-critical workloads, Tier 3 is the minimum recommended standard. Tier 4 is preferred for financial services, healthcare, and government applications.
Power Capacity
Power is measured in kilowatts (kW) per rack or cabinet. Standard enterprise deployments use 6-8 kW per rack, while high-density and AI workloads may require 20-50+ kW per rack.
STAT 1.38 Average PUE across all facilities in the GoDataCenters directory
Key power questions to ask:
- What is the maximum kW per rack available?
- Is metered or circuit-based power billing used?
- What redundancy level is provided (N, N+1, 2N)?
- Is on-site generation (diesel/gas) available for grid outages?
PUE (Power Usage Effectiveness)
PUE measures how efficiently a data center uses energy. A PUE of 1.0 would mean 100% of energy goes to computing equipment. In practice:
- PUE 1.2-1.4: Good efficiency (modern facilities)
- PUE 1.4-1.6: Average
- PUE 1.6+: Below average, often older facilities
Lower PUE translates directly to lower operating costs, as less energy is wasted on cooling and overhead.
AI and High-Density Considerations
If you’re planning to deploy GPU workloads for AI training, inference, or high-performance computing, evaluate these additional factors:
- Liquid cooling support: Direct liquid cooling (DLC) or rear-door heat exchangers for 30+ kW/rack densities
- Power density: Ensure the facility can deliver the kW per rack your GPU clusters require
- Network fabric: Low-latency interconnects between racks and rows for distributed training
- GPU hosting experience: Ask the provider about their track record with GPU deployments
Certifications and Compliance
Depending on your industry, you may need facilities with specific certifications:
- SOC 2 Type II: Standard for service organizations, covers security, availability, and confidentiality
- ISO 27001: International information security management standard
- HIPAA: Required for healthcare data (look for HIPAA-compliant facilities with BAA availability)
- PCI DSS: Required for payment card data processing
- FedRAMP: Required for federal government workloads
How to Request and Compare Quotes
The colocation procurement process typically follows these steps:
- Define requirements: Power (kW), space (cabinets/cages), connectivity, compliance needs
- Shortlist facilities: Use a directory like GoDataCenters to filter by location, tier, and capabilities
- Request quotes: Submit RFQs to 3-5 providers for competitive pricing
- Evaluate proposals: Compare on total cost of ownership, not just rack rate
- Site visit: Tour your top 2-3 facilities before signing
- Negotiate terms: Contract length, power escalation clauses, SLA terms
Ready to start your search? Browse facilities or request a custom quote to get personalized recommendations from our advisory team.