Data Center FAQ

A data center is a facility that houses computer systems, servers, networking equipment, and storage systems. These facilities provide the physical infrastructure—including power, cooling, and security—needed to operate IT equipment reliably. Data centers range from small server rooms to massive hyperscale campuses spanning millions of square feet.

Colocation (colo) is a service where businesses rent space, power, and cooling in a third-party data center to house their own servers. Instead of building their own facility, companies pay monthly fees for rack space, power (measured in kW), and connectivity. This provides enterprise-grade infrastructure without capital expense.

Hyperscale data centers are massive facilities typically exceeding 100MW of power capacity, built by cloud providers (AWS, Azure, Google). They're characterized by enormous scale (500,000+ SF), custom infrastructure, advanced automation, and strategic locations near cheap power. Hyperscalers drive most new construction globally.

PUE (Power Usage Effectiveness) measures data center energy efficiency by dividing total facility power by IT equipment power. A PUE of 1.0 would be perfect efficiency. Average PUE is around 1.58; best-in-class facilities achieve 1.1-1.2. Lower PUE means lower operating costs and better sustainability.

Data centers use various cooling methods: CRAC/CRAH units circulate chilled air, hot/cold aisle containment separates airflows, free cooling uses outside air, liquid cooling brings coolant to chips for AI workloads, and evaporative cooling uses water in dry climates. Modern facilities combine methods for optimal efficiency.

The Uptime Institute Tier system rates reliability: Tier 1 is basic (99.671% uptime), Tier 2 adds redundant components (99.741%), Tier 3 is concurrently maintainable (99.982%), Tier 4 is fault-tolerant with 2N redundancy (99.995%). Most enterprise colocation is Tier 3 or higher.

N+1 redundancy means having one more component than needed. If 4 generators (N) are required, N+1 means 5 so one can fail without impact. 2N means fully duplicating everything. N+1 is standard for most components; 2N is used for critical systems in high-reliability facilities.

Colocation pricing varies by market and requirements. Retail colocation runs $150-300 per kW/month in primary markets. Wholesale (1MW+) runs $80-150 per kW/month. Key factors: location (NoVA cheapest, Silicon Valley most expensive), power density, contract length, and connectivity needs.

Retail serves smaller deployments (racks to cages) with shared infrastructure and flexibility but higher per-kW pricing. Wholesale serves large deployments (1MW+) with dedicated space at lower rates but longer commitments (5-15 years). Enterprises typically start retail and scale to wholesale.

Key factors: Power availability and cost (60-70% of opex), fiber connectivity and latency, natural disaster risk, water for cooling, tax incentives, labor market, land costs, and permitting. The best sites balance low power costs with strong connectivity and minimal risk.

Northern Virginia became #1 due to: proximity to internet exchanges and federal agencies, favorable Dominion Energy rates, business-friendly taxes, strong fiber, and first-mover advantage. The market has 3+ GW of capacity and continues growing despite power constraints.

Fastest-growing US markets: Phoenix (abundant power, low risk), Dallas (ERCOT, low costs), Atlanta (Southeast hub), Columbus (Midwest connectivity). Internationally: Singapore, Tokyo, Frankfurt, and emerging markets in Malaysia, Indonesia, and Middle East. AI demand is driving unprecedented growth.