Total Cost of Ownership (TCO) Model for Liquid Cooled Data Centers

Author: Dr Andy E Young, Asperitas; Eduard Roytman, Microsoft Collection: 2026-OCP-EMEA-Summit-Cooling-Environments-And-Liquid-Cooling

Summary

This presentation introduced OCP TCO tool v1.3, an open-source, macro-free Excel spreadsheet for modeling the total cost of ownership of data centers across any combination of air cooling, liquid cooling, and hybrid configurations. The tool has been in development for over three years, originally contributed by Wolf Brink of Promation, and is co-led by Dr Andy Young (Asperitas) and Eduard Roytman (Microsoft) within the OCP Cooling Environments workgroup. It is available free of charge from the OCP Marketplace. The tool's core value proposition is converting qualitative technology claims into defensible, auditable financial comparisons — addressing the problem that proprietary TCO tools lack transparency and vendor-neutral TCO comparisons are rare.

The tool models data centers hierarchically: data center → up to four data rooms → up to four cluster types per room → up to three rack types per cluster → up to three chassis types per rack. At every level the user can assign power, cooling, and efficiency overheads. It supports all major cooling topologies — perimeter CRAH, in-row CRAH, direct-to-chip (DTC) with L2A CDU (side-car), open/closed passive/active RDHX, DTC with L2L CDU, and full immersion — and can model geographic location's effect on cooling economization via climate data (hot/cold, dry/humid inputs rather than GPS coordinates). Financial outputs include full CAPEX breakdown, annual OPEX, Net Present Value (NPV) via discounted cash flow, depreciation, and a "break-even hosting rate" — the minimum metered energy unit cost at which the data center pays off its capital.

Version 1.3 (released the week following the summit) introduced several architectural improvements: CDU placement moved from rack to cluster level (FWS model) with manifolds in rack (TCS model); air cooling modeled at dataroom, cluster, and rack levels; improved top-down input flow from three user personas (IT Operations, Facilities Operations, Finance Operations); improved finance calculations including fully automated NPV without Goalseek; CDU placement now in cluster; and detailed overhead breakdown outputs. A worked Paris vs. Barcelona comparison demonstrated the tool's precision: same 50 MW setpoint, same DTC+RDHX configuration, but Barcelona's lower seasonal economization hours result in 20 kW higher mechanical losses and a €879,382/year higher break-even hosting rate (charging 0.352 €/kWh in Paris vs. 0.354 €/kWh in Barcelona at 0.10 €/kWh buy-in rate, metered at L3 rPDU outlets).