Airsys Strengthens Global Leadership Team with New Executive Appointments
November 9, 2026
Airsys moves into heat reuse as ThermAI turns GPU servers into home heating
November 12, 2026AI is pushing data centers into a capacity crunch, and many operators are still planning with metrics that can’t show where their provisioned power actually goes. That’s the central finding of a new knowledge brief from Aberdeen Strategy & Research, a division of Spiceworks Ziff Davis.
The brief, written by Jim Rapoza, vice president and principal analyst at Aberdeen, examines how Power Compute Effectiveness (PCE), an open metric published by the Open Compute Project, gives organizations a design-level view of how much provisioned power is allocated to compute.
What problem does Aberdeen identify?
According to Aberdeen’s research, increasing costs of resources and power is the top data center challenge, cited by 35% of enterprise and service provider respondents. Power usage and optimization also ranks among the top three challenges to successful AI implementations, cited by 39% of organizations building AI infrastructure.
The brief explains why: as rack densities rise, a growing share of provisioned power goes to cooling, electrical distribution losses and auxiliary loads. That leaves stranded capacity that can’t support IT equipment.
What is Power Compute Effectiveness (PCE)?
PCE is defined as provisioned power for compute divided by total provisioned data center power. Aberdeen highlights three distinctions:
- It’s evaluated at the design level. PCE is calculated from single-line diagrams, equipment submittals and commissioning documentation, not live operating conditions.
- It measures effectiveness, not efficiency. PCE shows how provisioned power is allocated. It does not measure runtime energy consumption or compute performance.
- It makes stranded capacity visible. By isolating power committed to overhead before it reaches the server, PCE helps reveal where capacity is being trapped.
Airsys developed PCE, and the Open Compute Project accepted and published the PCE white paper.
How is PCE different from PUE?
Power Usage Effectiveness (PUE) measures operational energy efficiency. PCE measures how provisioned power is structurally allocated. Aberdeen’s conclusion is that operators can no longer rely solely on operational energy-efficiency metrics such as PUE; they also need visibility into how power is provisioned from the ground up.
What are top-performing organizations doing?
Organizations scoring in the top 30% of Aberdeen’s data center success metrics are 66% more likely than peers to be modernizing data center power capabilities and three times as likely to be modernizing cooling. Among organizations that modernized cooling, the most-cited benefits were improved compute performance (55%), reduced data center power consumption (42%) and a better ability to meet peak demand (41%).
The brief also outlines a multi-zone thermal approach that matches cooling to heat grades, helping teams shift more provisioned power toward compute within existing grid commitments. As Aberdeen notes, putting that headroom to use may still require investment in IT hardware, cooling and power distribution.
Frequently asked questions
What does the Aberdeen brief cover?
How PCE helps data center operators see how provisioned power is allocated to compute and why that matters as AI drives capacity bottlenecks.
What is PCE?
Power Compute Effectiveness: provisioned power for compute ÷ total provisioned data center power.
Is PCE an efficiency metric?
No. PCE measures allocation. PUE measures operational energy efficiency.
What is return on invested power (ROIP)?
A financial outcome derived from PCE, reflecting the return when reallocated power supports revenue-generating IT, relative to the infrastructure investment required.
Who developed PCE?
Airsys developed PCE, and the Open Compute Project accepted and published the PCE white paper.
