A colocation RFQ should include power requirements (kW per rack and total load), cooling specifications, network and connectivity needs, physical space dimensions, security and compliance mandates, SLA uptime guarantees, pricing structure, timeline, and AI-specific capabilities like GPU density and liquid cooling support. This structured approach ensures facilities can accurately quote and helps you compare apples-to-apples across providers.
Introduction
Choosing a colocation provider is one of the most critical infrastructure decisions your organization will make. Unlike cloud services where vendors handle the physical details, colocation puts you in the driver’s seat. You specify exactly what you need, and providers respond with pricing and capabilities that match your requirements.
But here’s where many IT directors and infrastructure teams stumble: they write vague or incomplete RFQs, leading to mismatched quotes, hidden costs, and vendor negotiations that drag on for months. A poorly structured RFQ can result in oversized monthly bills, insufficient power density for future AI workloads, or connectivity that doesn’t scale with your ambitions.
A well-written Request for Quote (RFQ) does more than get you pricing. It establishes a clear baseline for comparison, forces you to articulate infrastructure needs you might not have fully thought through, and gives vendors a precise specification to quote against. The difference between a three-week evaluation and a three-month procurement nightmare often comes down to RFQ quality.
This guide walks you through the exact sections every colocation RFQ should include, common pitfalls to avoid, and how to structure your evaluation to make vendor comparisons fast and fair.
The Difference Between RFI, RFQ, and RFP: Why It Matters for Colocation
RFI (Request for Information): Your first conversation with potential vendors. You’re gathering general information about capabilities, service areas, and high-level offerings. RFIs are looser, exploratory, and don’t require a formal response.
RFQ (Request for Quote): A formal request for pricing on a specific, detailed set of requirements. You’re saying, “Here’s exactly what we need. Give us your price.” A good RFQ results in comparable quotes from multiple providers.
RFP (Request for Proposal): Goes deeper than pricing. You’re asking vendors not just for quotes but for how they’d solve your problem: their approach, timeline, staffing, SLAs, value-add services. For colocation, most organizations use an RFQ; an RFP is necessary only if you’re bundling colocation with managed services or infrastructure migration.
For a colocation RFQ, your job is to make specifications so clear that every vendor quotes the same thing. That’s how you actually compare apples to apples.
Step 1: Define Your Power Requirements (The Foundation of Every Colocation RFQ)
Power is the single most critical specification in a colocation RFQ.
Per-Rack Power Density: Specify how much power each rack will consume in kilowatts (kW). Your RFQ should state: “We require X kW per rack, with capability to scale to Y kW per rack within 12 months.”
Total Facility Draw: Specify total power consumption across all racks. If deploying 20 racks at 10 kW each, that’s 200 kW total.
Redundancy and UPS Requirements: State whether you need N+1 power redundancy. Specify UPS requirements: how long should the facility keep you online if the grid fails?
AI-Readiness Angle for 2026: A rack with 8x NVIDIA H100s can draw 20+ kW and require specialized power distribution and cooling. Your RFQ should ask: “What is your maximum supported power density per rack? Do you have facilities already running 20kW+ per rack?”
Step 2: Cooling and Environmental Specifications
Cooling Capacity: Ask for cooling capacity in kW per row, per aisle, and total facility. For high-density deployments (10+ kW per rack), ask about hot-aisle containment, in-row cooling, or liquid cooling solutions.
Cooling Type: Specify air cooling (standard) or liquid cooling for dense AI workloads. Ask whether the facility supports rear-door heat exchangers, in-rack coolers, or direct-to-chip liquid loops.
Temperature and Humidity Standards: Standard specs are 15–35°C with 20–80% relative humidity. Ask what their SLA is for environmental conditions.
AI Cooling Considerations: A single GPU rack can produce 20+ kW of heat. Your RFQ should ask: “Do you have liquid-cooled racks available? Have you deployed GPU-intensive workloads? What is your typical inlet temperature for dense compute?”
Step 3: Network and Connectivity Requirements
Primary Network Connection: Specify port speed and redundancy (e.g., “Two 100 Gbps ports on separate circuit vendors”).
Carrier Diversity: Require at least two independent carriers with different fiber routes.
Cross-Connect Costs: Ask for typical costs per carrier and per port type. This is a hidden cost many buyers miss.
Network Latency: Specify acceptable latency to primary offices or cloud regions.
AI Workload Bandwidth: Ask about bulk transfer capabilities for massive AI training datasets.
Step 4: Physical Space and Equipment Specifications
Rack Configuration: Specify number of racks and type. Include whether full or partial-rack arrangements are acceptable.
Equipment Dimensions: Include a bill of materials: server models, network switches, custom enclosures, total weight per rack.
Delivery and Setup: Specify deployment timeline, setup requirements, and whether expedited onboarding is needed.
Access and Maintenance: Specify 24/7 access needs, badge procedures, and visitor policies.
Step 5: Security and Compliance Requirements
Physical Security: Multi-factor access, surveillance, guards, visitor logs.
Certifications: SOC 2 Type II, ISO 27001, HIPAA, PCI DSS, FedRAMP as applicable.
Facility Hardening: Backup generators, redundant utility feeds, fire suppression systems.
AI-Specific Security: Ask about IP protection policies for proprietary AI models.
Step 6: SLA and Support Requirements
Uptime Guarantee: Specify target (99.9% = 8.7 hours downtime/year; 99.99% = 52 minutes/year).
Credits and Penalties: Ask what credits you receive if SLA is missed.
Maintenance Windows: Specify acceptable windows and notification requirements.
Technical Support: Clarify hands-on support inclusion, hourly rates, and escalation procedures.
Step 7: Pricing Structure and Terms
Monthly Recurring Charges: Break down by rack space, power, network, floor space.
Non-Recurring Charges: Setup fees, cross-connect installation, cabling labor.
Contract Length and Pricing Lock: Specify preferred term and ask about annual increase caps.
Termination Terms: Ask about early exit fees.
Step 8: AI-Readiness and Future-Proof Specifications for 2026
GPU Support: Ask about deployed GPU workloads, NVIDIA H100/A100 experience, liquid cooling support, power delivery for GPU-dense racks.
Power Density Above 20kW/Rack: Explicitly ask maximum supported density and whether 20kW+ infrastructure exists today.
Emerging Cooling Technologies: Liquid-to-air rear-door exchangers, in-rack liquid coolers, direct-to-chip cooling.
AI-Readiness Scoring: Platforms like GoDataCenters pre-score facilities on AI-readiness across power density, cooling, network, and GPU experience.
Common Mistakes in Colocation RFQs
- Under-specifying power density: “high power” is meaningless; specify actual kW per rack.
- Ignoring future growth: State current and projected requirements.
- Not breaking out pricing: Insist on itemized costs.
- Vague network requirements: Specify port speeds, carrier diversity, redundancy.
- Failing to account for AI-readiness: Don’t RFQ traditional facilities without GPU questions.
- Omitting compliance details: HIPAA/SOC 2 gaps discovered post-signature are costly.
- Setting unrealistic timelines: Ask about lead times upfront.
How to Evaluate RFQ Responses: A Scoring Framework
Weighted Scoring Matrix:
- Power capacity and reliability: 25 points
- Cooling and environmental controls: 20 points
- Network and connectivity: 20 points
- Security and compliance: 15 points
- SLA and support: 10 points
- Price and contract terms: 10 points
Score each vendor 0–10 per criterion, multiply by weight. Use price as tiebreaker between similarly-scored vendors. Always do reference checks before signing.
Streamlining RFQs with GoDataCenters
GoDataCenters pre-qualifies colocation facilities across 50+ U.S. markets with AI-Readiness Scores. Instead of issuing RFQs to dozens of unknowns:
- Filter by power, cooling, and network requirements
- See pre-scored AI-readiness
- Compare across markets on a standardized basis
- Shortlist 3–5 qualified facilities before issuing formal RFQs
This marketplace approach can cut RFQ timelines from 3 months to 4–6 weeks. Browse AI-ready facilities or get a quote to start your shortlist.
FAQ: Colocation RFQ Questions Answered
What’s the difference between an RFQ and an RFP for colocation?
An RFQ requests pricing for specific, defined colocation services (rack space, power, cooling, network). An RFP asks vendors to propose how they’d solve your broader problem, including service approach, staffing, and implementation timeline. Use an RFQ for standard colocation; use an RFP if bundling with managed services or infrastructure design.
How many facilities should I include in my RFQ process?
Aim for 3–5 qualified vendors. Fewer than 3 and you don’t have enough competition to negotiate effectively. More than 5 and evaluation becomes unwieldy. Use a marketplace like GoDataCenters to pre-filter, then RFQ the final 3–5.
Should I include my equipment bill of materials in the RFQ?
Yes. The more specific you are about server models, switch types, and total weight per rack, the more accurate the facility’s quote. Facilities can optimize racking layout, power distribution, and cooling when they know exactly what you’re deploying.
What’s a realistic timeline for RFQ-to-signature?
Single facility: 2–4 weeks. Multi-vendor evaluation (3–5 facilities): 6–10 weeks. This includes issuing RFQs (1 week), collecting responses (2–3 weeks), evaluating (1–2 weeks), negotiating (1–2 weeks), and legal review (1–2 weeks). AI-ready marketplaces can compress this by pre-qualifying vendors upfront.
What GPU-specific questions should I ask in my RFQ?
Ask: (1) Have you deployed GPU-intensive workloads, and at what density? (2) What is your maximum power per rack? (3) Do you support liquid cooling today? (4) What is your typical inlet temperature for dense compute? (5) Can you reference a customer running H100 or similar GPUs? These questions identify facilities with actual GPU experience.
Can I negotiate pricing after receiving RFQ responses?
Quoted prices are the opening offer. Facilities expect negotiation. Volume discounts, multi-year pricing locks, and waived setup fees are standard topics. Strongest position: shortlist 2–3 equally qualified vendors and negotiate competitively.
Ready to streamline your colocation search?
Browse AI-Ready Data Centers or Get a Quote to start your shortlist today.