When software executives and financial leaders contemplate cloud repatriation, their primary analytical focus is almost exclusively on compute hardware procurement: the capital expenditures associated with enterprise servers, NVMe storage arrays, and 100GbE network switches. However, in physical infrastructure engineering, hardware depreciation represents only one half of the total cost equation. The ongoing operational foundation of any bare-metal deployment is data center colocation: specifically electrical power density, cooling efficiency, and contract structure.

Unlike public cloud billing—where electricity, HVAC chiller plants, backup diesel generators, and uninterruptible power supply (UPS) systems are bundled into an opaque per-second virtual machine rate—colocation operators bill infrastructure owners based on physical rack units, circuit amperage, power draw, and Power Usage Effectiveness (PUE) multipliers. Understanding the nuances of wholesale power contracts, kW density scaling, and cooling economics is critical to executing a successful, profitable cloud repatriation strategy.

How Do Colocation Facilities Price Electrical Power and Rack Space?

Colocation facilities price data center infrastructure using three primary billing models: metered power based on actual kilowatt-hour (kWh) consumption multiplied by a facility PUE, unmetered flat-rate contracted kW allocations, or combined per-rack footprint fees bundled with designated circuit amperages.

The financial mechanics of enterprise colocation hinge on the fundamental constraint of modern data center engineering: power density, not physical floor space. In the early 2000s, data center cages were filled with lightweight, low-power single-socket servers drawing 2kW to 4kW per 42U rack cabinet. Today, a single 2U enterprise chassis packed with dual AMD EPYC 9654 processors and eight NVMe SSDs draws between 800W and 1,200W under sustained load.

Packing twenty such servers into a single standard 42U cabinet creates an electrical and thermal load of 16kW to 24kW. If an enterprise signs a contract for low-density space, the data center facility will exhaust its allocated cooling or circuit breaker capacity long before the rack is physically full, stranding valuable rack units. Structuring colocation agreements around high-density power delivery (typically 208V or 415V three-phase power) is the most effective method to compress colocation real estate overhead.

Decoding PUE: The True Multiplier of Operating Expenses

Power Usage Effectiveness (PUE) is the standard metric used by the data center industry to measure electrical efficiency. It is defined as the ratio of total energy entering the data center facility divided by the energy consumed strictly by the computing equipment:

PUE = Total Facility Energy / IT Equipment Energy

In a facility with a PUE of 2.0, for every 1 watt of electricity used by your server processors and storage drives, another 1 watt is consumed by data center chillers, CRAC (Computer Room Air Conditioning) units, lighting, and UPS battery transformation losses. Modern Tier III and Tier IV wholesale colocation facilities utilize direct evaporative cooling, cold-aisle containment, and waste-heat recapture to achieve PUE ratings between 1.15 and 1.28.

When negotiating a colocation master service agreement (MSA), the contract’s PUE calculation mechanism is paramount. In a metered power contract, the facility operator bills the tenant: (Tenant IT kWh) × (Contractual or Actual PUE) × (Wholesale kWh Tariff). If your infrastructure draws 50kW continuously across a year (438,000 kWh), the difference between a facility with a 1.20 PUE and an older legacy site with a 1.65 PUE is 197,100 additional kWh billed to your account. At an industrial rate of $0.12/kWh, that operational difference represents $23,652 in wasted annual capital.

To further protect infrastructure margins, engineering teams should pair high-density colocation footprints with optimized local NVMe caching and tuning strategies, such as Ceph BlueStore NVMe WAL & OSD Tuning, to maximize storage IOPS per watt of power consumed.

Financial Forensic Audit: Cloud Energy Margins vs. Owned Colocation

To deconstruct the immense margin that public cloud providers extract from enterprise compute, let us perform a reverse financial audit on a 100kW IT workload. In AWS or Google Cloud, sustaining a steady 100kW compute footprint equates to running approximately 250 high-end compute-optimized virtual machines around the clock.

At standard reserved or savings-plan rates, 250 enterprise cloud instances cost approximately $52,000 per month ($624,000 annually). When we examine the underlying physical energy economics, 100kW operating 24/7/365 consumes 876,000 kWh annually. In an enterprise colocation facility with a 1.20 PUE, total billed power equates to 1,051,200 kWh. At an average enterprise power rate of $0.11 per kWh, the actual electricity bill for that compute footprint is $115,632 per year ($9,636 per month).

Adding eight high-density colocation cabinets, dual redundant A+B 208V power circuits, and dual 10Gbps unmetered IP transit adds approximately $6,400 per month ($76,800/year). Amortizing the $220,000 server hardware capital expenditure over 36 months adds $6,111 per month. The total fully loaded cost to operate that 100kW enterprise workload in a private colocation cage is $22,147 per month ($265,764 annually)—compared to $624,000 on public cloud. The net savings exceeds $358,000 every single year.

Power Factor Optimization and Three-Phase Circuit Engineering

A critical yet widely overlooked component of data center colocation financial auditing is electrical Power Factor (PF) and phase balancing. In alternating current (AC) electrical systems, the apparent power delivered by the utility (measured in Volt-Amperes, or VA) is not always equal to the real power consumed by the computing hardware (measured in Watts, or W). The ratio between real power and apparent power is defined as the Power Factor:

Power Factor (PF) = Real Power (Watts) / Apparent Power (Volt-Amperes)

Modern enterprise server power supplies (such as Titanium-rated 80 PLUS server PSUs) incorporate active Power Factor Correction (PFC) circuits, typically maintaining a power factor between 0.95 and 0.99 under load. However, if an enterprise runs older hardware or unbalanced single-phase circuits across a three-phase power feed, the phase imbalance causes excessive neutral current and reactive power draw. Many enterprise colocation operators incorporate penalty clauses in their Master Service Agreements (MSAs), assessing monthly surcharges if a tenant’s aggregate power factor drops below 0.90.

To eliminate reactive power penalties and maximize electrical efficiency, infrastructure engineers should deploy intelligent rack-mount Power Distribution Units (PDUs) supporting 415V/240V three-phase wye distribution. By feeding 415V three-phase power directly to the cabinet PDU and stepping it down to 240V at the server power supply input, transmission currents are reduced by 42% compared to standard 208V feeds. Lower current equates to significantly reduced resistive thermal losses ($I^2R$) in power whips and busways, lowering internal cabinet heat generation and directly driving down facility cooling requirements.

Frequently Asked Questions About Colocation Power Costs