Data Center Economics Advisory

Infrastructure decisions, modeled to the dollar.

Acquisition price is one line item. RFG models the full 3–5 year total cost of ownership of enterprise infrastructure — compute, power, space, licensing, and operations — across on-prem, cloud, and AI factory deployments.

Illustrative model · 1 MW enterprise workload
TCA ONLY
$8.5M
ON-PREM TCO
$15.2M
CLOUD TCO
$21.9M
HARDWARE / TCA POWER & SPACE OPS & LICENSING CLOUD CONSUMPTION

FIGURES ILLUSTRATIVE. ENGAGEMENTS MODEL YOUR WORKLOADS, RATES, AND CONTRACTS.

The Core Discipline

TCA tells you what you paid.
TCO tells you what it costs.

Most infrastructure decisions are won or lost in the 3–5 years after the purchase order. RFG models both sides so the trade-offs are explicit before capital is committed.

TCA

Total Cost of Acquisition

The day-one number: what it takes to put the infrastructure on the floor.

  • Server, storage, and network hardware
  • Racks, PDUs, and cabling
  • Deployment, integration, and migration services
  • Initial software and support contracts

TCO · 3–5 YR

Total Cost of Ownership

The number that hits the P&L every quarter for the life of the platform.

  • Power, cooling, and PUE-adjusted energy spend
  • Data center space, leases, and colocation fees
  • Software licensing tied to cores, sockets, and capacity
  • Maintenance, support renewals, and operations staffing
  • Refresh timing, residual value, and exit costs

What We Model

Six practice areas. One method.

Every engagement produces a defensible, assumption-transparent economic model your CFO can interrogate line by line.

Workload Economics

On-Prem vs. Cloud for Enterprise Workloads

Steady-state enterprise workloads price very differently on owned infrastructure than on consumption billing. We model both paths over 3–5 years — including egress, growth, and contract terms — to show where each genuinely wins.

AI Infrastructure

AI Factories: On-Prem vs. Cloud vs. Neocloud

GPU economics are unforgiving. We compare owned AI factories against hyperscale cloud and neocloud GPU providers on cost per token, utilization sensitivity, power, and contract risk.

Compute Platforms

CPU Compute: TCA & TCO Across All OEMs

Server-by-server modeling of on-prem CPU deployments across every major compute OEM, on both Intel and AMD platforms — configuration, pricing, performance per watt, and per-core licensing impact.

AI Accelerators

AMD vs. NVIDIA: Performance and Cost

Accelerator selection by the numbers: training and inference performance, memory capacity and bandwidth, ecosystem maturity, availability, and fully loaded cost per unit of work.

Modernization

Consolidation & Modernization Savings

Aging fleets burn money in place. We quantify the cost, space, power, and licensing recovered by consolidating onto modern platforms — and the payback period on the refresh.

Sustainability

Energy & Carbon Impact

Modernization is a sustainability lever. We translate consolidation into kWh avoided, carbon reduced, and reporting-ready figures that stand up alongside the financial case.

From the Library

Latest research

View all publications →

Put a number on your next infrastructure decision.

Engagements typically begin with a focused workload or fleet assessment and deliver a board-ready economic model in weeks, not quarters.