Every fragmented decision in business eventually gets fixed by agreeing on a single number. Sales agreed on pipeline value. Marketing agreed on qualified leads. Engineering agreed on uptime. Once a function has a number that everyone uses the same way, the arguments shift from which metric matters? to how do we move this one? That is a much more productive argument.
Server IT efficiency does not yet have that number in widespread use. ITEEsv is the candidate. It exists. It is formal. It is in the same family of metrics every data centre operator already uses. And it answers a question every stakeholder in the room is already asking, just in different words.
What ITEEsv actually is
ITEEsv stands for IT Equipment Energy Efficiency for servers. It is the formal ISO/IEC 30134-4 metric, in the same standards family as PUE (Power Usage Effectiveness) and WUE (Water Usage Effectiveness). It is referenced in the European Code of Conduct for Data Centre Energy Efficiency and is one of the metrics required for reporting under the EU Energy Efficiency Directive (EED). The difference between ITEEsv and the more familiar PUE and WUE is that PUE and WUE describe the building, how efficiently it delivers power and uses water to the IT load, while ITEEsv describes what that IT load actually does with the energy once it arrives.
That distinction is the whole point. PUE is a building metric. ITEEsv is a server metric. Improving one tells you almost nothing about the other.
How ITEEsv is calculated
ITEEsv is the ratio of total maximum performance to total maximum power consumption across the servers in a data centre, expressed as the sum of each server's peak performance divided by the sum of each server's peak power draw:
ITEEsv = Σ SMPEi / Σ SMPOi
Where SMPE is the peak performance of each server, taken from an approved benchmark, and SMPO is the peak power consumption of that same server in kW. Sum across the estate (or the server group of interest), then divide. A higher result means more useful work per Watt at peak. Improving the average over time means replacing or reconfiguring servers so that the new SMPE/SMPO ratio is better than the old.
The SMPE unit depends on which benchmark is used. Three are approved:
- SERT: produces ITEEsv(SERT), reported in SSJ_ops (Server-side JavaScript Operations)
- SPECpower_ssj2008: produces ITEEsv(SPECpower_ssj2008)
- LINPACK: produces ITEEsv(LINPACK)
Three rules sit on top of the formula. A single benchmark must be applied across all servers in any given calculation; benchmarks cannot be mixed. Servers where SMPE and SMPO cannot be accurately determined are excluded from the calculation rather than estimated. And when comparing the ITEEsv of a data centre over time, the same benchmark must be used at every measurement point; ITEEsv(SERT) in 2024 cannot honestly be compared with ITEEsv(LINPACK) in 2025.
Our reports use ITEEsv(SERT). Each report lists the value in SSJ_ops per Watt, the benchmark used, and, for partial-estate calculations, the server group or subsystem the figure applies to so the result is interpretable in context.
Why a common number matters
In a typical refresh decision, four stakeholders care about the answer for four different reasons:
- The engineer wants to know whether the new kit can run the workload reliably and leave headroom for the next one.
- Procurement wants to know whether the configuration is the right one given the budget envelope and the supplier landscape.
- Finance wants to know what the lifetime cost looks like and how it compares against the cost of keeping the existing estate running.
- Sustainability wants to know what the change does to the organisation's reported footprint.
Each of those stakeholders currently speaks a different metric to describe the same decision. The engineer talks about utilisation and performance per watt. Procurement talks about unit price. Finance talks about TCO. Sustainability talks about Scope 2 and embodied carbon. Nobody is translating between them in any rigorous way, and the buying group is, predictably, fragmented.
It is the one figure that means something to every stakeholder at once, which is exactly what a fragmented buying group needs.
What ITEEsv encompasses
Done properly, ITEEsv encompasses energy efficiency, performance per watt, capacity headroom and lifecycle cost. The engineer sees performance per watt and headroom in the same number. Finance sees lifecycle cost. Sustainability sees energy efficiency. Procurement sees a defensible single metric they can use to compare configurations on a like-for-like basis.
That doesn't dissolve the underlying tradeoffs. It exposes them. The right ITEEsv for an exchange running latency-critical workloads is not the right ITEEsv for a research compute estate. Both are knowable. Both can be moved.
Why this is Interact's number to define
ITEEsv is ownable as a category by Interact because Interact built the dataset that makes it measurable at scale. It is one thing for the metric to exist on paper, in a standards document. It is another thing for it to be measurable in practice, against a real estate, against real workloads, against the candidate replacement configurations a buyer is actually weighing.
That second thing requires the homework. It requires the kind of dataset that takes nine years to build and cannot be reverse-engineered from spec sheets. That is the thing Interact has spent the last decade quietly assembling, and it is what makes ITEEsv real rather than aspirational.
If you want to understand where your current estate sits on ITEEsv, and what the next refresh would do to it, that is the conversation a Verify engagement starts. Your Position on this site is the simplest place to begin.