Back to BlogAnalysis · EU Compliance & DCIM

    Closing the 20% EED Gap Schneider Can't Reach

    Schneider says EcoStruxure DCIM covers 80% of EU EED reporting. The missing 20% is IT equipment efficiency — exactly what BMC telemetry actually sees.

    May 2026 12 min readSensaka Research

    In January 2026, Schneider Electric's sustainability lead Alison Matte said something to Data Center Dynamics that European data center operators should be reading carefully. The company's flagship EcoStruxure IT DCIM platform — the most widely deployed data center infrastructure management software in Europe — covers roughly 80% of what the EU Energy Efficiency Directive requires.

    The other 20% is IT equipment efficiency and server utilization. And by Schneider's own account, they aren't building toward it. The reasoning given was straightforward: they don't have access to what's happening inside a server, they can't give accurate predictions on it, and they aren't going to try.

    This is not a roadmap gap. It is a structural one. For European operators staring down annual EED submissions, it raises the obvious question: where does the missing 20% come from?

    What the EED actually demands

    The recast Energy Efficiency Directive commenced in May 2024 and applies to every data center in the bloc above 500 kW IT load. Each year, operators must submit a defined set of metrics into a European database:

    • Floor area and installed IT power
    • Data volumes processed
    • Total energy consumption
    • Power Usage Effectiveness (PUE)
    • Temperature setpoints
    • Waste heat utilization
    • Water Usage Effectiveness (WUE)
    • Renewable energy share
    • IT equipment efficiency and server utilization

    On the facility side — power, cooling, water, environmental — a mature DCIM like EcoStruxure handles most of it. PUE is calculable from facility metering. WUE is calculable from cooling instrumentation. Temperature setpoints come from the BMS layer. That is the 80%.

    The 20% is everything that depends on what is happening inside the servers: how efficiently IT equipment converts electricity into useful work, how utilized the compute is, how power draw correlates with actual workload demand. None of that data lives at the facility layer, and none of it can be reliably estimated from there.

    Why facility-only DCIM hits a structural ceiling

    Schneider's framing in the DCD interview was honest. They don't have visibility into the server, so they aren't going to make claims they can't substantiate. That's the right answer for a regulator-graded compliance tool. It is also a permanent architectural statement, not a temporary one.

    Every facility DCIM platform on the European market — Schneider EcoStruxure, Nlyte, Sunbird dcTrack, Vertiv Environet — is built around the same data sources: smart PDUs, branch circuit metering, CRAC/CRAH instrumentation, environmental sensors, and BMS integration. They are excellent at the white space. They are structurally blind to the rack contents.

    The IT efficiency portion of EED reporting requires data the facility layer simply does not have:

    • Per-server power draw, not per-circuit or per-PDU
    • Server-level utilization across CPU, memory, and GPU
    • The relationship between power consumed and computational work performed
    • Server Computing Efficiency benchmarking, per the EN 50600 family of standards
    • Workload-attributed energy consumption for ESG reporting at the application level

    Operators can estimate some of this by dividing PDU readings across the rack. Estimation is not what regulators or auditors actually want — and the next iterations of the EED will tighten this requirement, not loosen it.

    Where the 20% actually lives: the BMC

    Every modern server ships with a Baseboard Management Controller — a dedicated processor on the motherboard that operates independently of the operating system, communicates over a separate management network, and reports on the physical state of the machine in real time.

    The BMC is where IT equipment efficiency data actually exists:

    Per-node power draw

    Not derived from PDU arithmetic, but reported directly by the server's onboard power telemetry, typically once per second, per power supply.

    Component-level energy attribution

    Modern Redfish implementations expose CPU package power, GPU TDP utilization, memory subsystem power, and fan power as discrete streams. Paired with OS-layer utilization, the result is a defensible energy-per-workload number.

    PSU efficiency curves

    Power supply efficiency varies non-linearly with load. The BMC reports both input and output power, which means efficiency is observable, not assumed.

    Idle versus active accounting

    The IT efficiency story EED ultimately wants is one about idle waste. Servers idling at 50% of nameplate draw is the largest hidden cost in most enterprise data centers, and a regulatory red flag where benchmarking exists. The BMC sees it. PDUs do not.

    None of this is theoretical. The data is already being generated by every iDRAC, iLO, iBMC, and XClarity controller in the European fleet. It is simply not being collected — because the tools that monitor the facility do not speak BMC, and the tools that speak BMC were not built for compliance reporting.

    What the missing 20% looks like in practice

    Consider a 5 MW European colocation site running mixed Dell and HPE hardware across 1,200 servers. The operator's EcoStruxure deployment cleanly produces:

    • Annual PUE of 1.42, from facility metering
    • WUE via Schneider's automated reporting
    • Floor area, installed power, temperature setpoints, all clean
    • Waste heat utilization captured via the cooling integration

    The auditor asks the next question: what was the IT equipment energy efficiency? How much of the 5 MW IT load actually performed useful work?

    The facility-layer answer is some version of "we don't have that at server granularity." Estimation by rack division is offered. The auditor records it as estimated.

    The BMC-layer answer, ingesting from the same 1,200 servers via Redfish, produces:

    • Per-server average and peak power for the reporting year
    • Idle-state energy attribution, typically 15–25% of total IT load
    • PSU efficiency distribution across the fleet
    • Power-to-utilization curves by server class
    • Workload-attributed energy when paired with the orchestration layer

    That is the 20%. It is not exotic. It is a different data plane.

    The complement, not the replacement

    This is not a case for ripping out Schneider, Nlyte, or Sunbird. Facility DCIM platforms are doing the 80% well, and the cost of replacing a working deployment is rarely justified by the EED gap alone.

    The cleaner architecture is layered. Keep the facility DCIM doing what it does. Add an IT-equipment intelligence layer that ingests directly from the BMCs. Pipe the IT efficiency outputs into the same compliance reporting workflow. The auditor sees one consolidated report. The operator runs two complementary tools.

    This is the position Sensaka DCOS Hardware Sentry is built for. We sit at the BMC layer — Redfish, iDRAC, iLO, iBMC, XClarity, IPMI — across mixed-vendor European fleets. The output includes EED-shaped IT efficiency exports designed to slot into existing facility-DCIM compliance bundles. On-premises deployment for EU data sovereignty. Transparent per-node pricing.

    Three steps before your next EED submission

    If you are operating above 500 kW in the EU, three practical steps before your next reporting window:

    Step 01

    Audit your current submission for estimation flags

    Anything calculated by dividing facility metering across racks is, by regulatory standard, an estimate. Identify those line items now, before the auditor does.

    Step 02

    Ask your existing DCIM vendor what their IT efficiency coverage actually is

    If the answer matches Schneider's public position — facility-strong, IT-equipment-blind — you have a known gap to close, on a regulatory clock.

    Step 03

    Pilot BMC ingestion on a single rack

    No vendor commitment required. Sensaka can connect to one rack of mixed hardware and produce the missing 20% as a sample report. Take that report into your next compliance review and let it stand on its own.

    The EED is not going away. The next revisions of the directive will tighten IT-efficiency requirements, not relax them. The structural gap Schneider has acknowledged today becomes a compliance failure tomorrow.

    Schneider gave the European market the most useful piece of competitive intelligence of 2026. The question is what operators do with it.

    Sensaka DCOS Hardware Sentry is the BMC and out-of-band intelligence layer for European data centers. To see the missing 20% on your hardware, contact us or request an online trial.

    See the missing 20% on your hardware

    Pilot BMC ingestion on a single rack. EED-shaped IT efficiency exports, on-premises, EU data sovereignty.

    References: power usage effectiveness and Schneider Electric.