6 min read

Sponsored

Updated: April 2026

Data centres face triple pressure: rising energy costs, AI workloads that double rack densities, and EU regulations requiring Scope 1/2 reporting from 2026 onwards. The biggest lever is cooling – it consumes up to 40 percent of total power. Those who want to save 50 percent of cooling energy without tearing out existing infrastructure need a software-first approach. NEXAIRA.Systems from ebm-papst delivers exactly that.

Key takeaways

• 30 to 40 percent of data centre power goes into cooling – the largest operational cost driver after the IT load itself (IEA, 2024).
• AI workloads drive rack densities: By 2030, 40 percent of all data centre workloads will run on GPU clusters with 40+ kW per rack (Dell’Oro Group, 2025).
• Retrofit instead of rebuild: NEXAIRA.Systems optimises existing cooling systems via digital twin and AI control – without hardware replacement.
• Result: up to 50 percent less cooling energy – validated in live operation with a German data centre operator (900 MWh and 240 tonnes of CO2 saved annually).
• Data stays on-site: On-premise architecture, no cloud lock-in for sensitive operational data.

What is NEXAIRA.Systems? NEXAIRA.Systems is an AI-powered platform from ebm-papst for optimising cooling systems in data centres. The system creates a digital twin of the entire cooling infrastructure, continuously calculates the most efficient operating parameters in real-time, and reduces energy consumption by up to 50 percent – without hardware replacement and without interfering with IT operations.

Why hardware alone is no longer enough

More efficient fans and modern refrigerants deliver single-digit percentage improvements. That is not enough when cooling loads explode at the same time. According to IEA forecasts, global energy consumption for data centre cooling will rise from 136 TWh (2024) to over 250 TWh by 2028. New hardware solves the problem locally. But a data centre consists of hundreds of components – chillers, cooling towers, pumps, fans, heat exchangers – all of which must be coordinated with each other.

The bottleneck is not in individual components, but in the orchestration of the entire system. This is where software comes in: system-wide optimisation in real-time, across all cooling circuits. The same principle drives the debate on sustainable AI data centres – the question is not whether you cool, but how intelligently.

An example illustrates the scale: The liquid cooling market is growing from 6.65 billion US dollars (2025) to 29 billion US dollars by 2033, according to Grand View Research. AI workloads alone are driving the share of GPU-based compute loads from 15 percent to 40 percent by 2030, according to Dell’Oro Group. The EU Energy Efficiency Directive (EED) requires all data centres over 500 kW to submit annual energy reports from 2025 onwards. Those who do not optimise their cooling system pay twice: higher power costs and higher compliance effort.

How the digital twin controls the cooling system

The principle: A digital twin maps the complete cooling system in the grey space – chillers, pumps, cooling towers, fans, pipes. The Cooling Supply Optimizer calculates the most energy-efficient operating points for every single component based on real-time sensor data, weather forecasts, and load predictions.

Instead of fixed setpoints, the system adapts parameters dynamically: when outside temperature drops, it reduces chiller load and harnesses free cooling. When a GPU cluster spins up, it anticipates the cooling demand and regulates proactively – before the white space heats up. Control is fully automatic via the existing building management system (BMS).

The critical point for IT managers: the system does not interfere with the white space. Server cooling remains exactly the same. NEXAIRA.Systems optimises only the delivery of cold – more efficiently, not differently. A comparable logic underpins modern cooling concepts that combine physical hardware with digital monitoring.

Retrofit without downtime

Every data centre operator knows the dilemma: modernisation normally requires downtime. NEXAIRA.Systems avoids this because it runs on top of existing infrastructure. Sensors, BMS connectivity, digital twin calibration – all during live operation. No chiller replacement, no pipeline retrofitting.

Connectivity is deliberately broad: MODBUS-RTU, WiFi, Ethernet, Bluetooth, NFC and cloud-to-cloud. This allows even older legacy systems that were not designed for IoT to be integrated. A typical deployment includes sensors at all critical cooling components, a local edge controller for data capture, and NEXAIRA software on a dedicated server in the grey space. Overall installation takes between two and six weeks depending on system scale.

For those who want to modernise hardware later, AxiBlade.Perform fans and NEXAIRA.Systems (860 and 950 mm diameter) provide components that natively communicate with NEXAIRA. But this is not mandatory – the software works independently of fan manufacturer.

What predictive maintenance means in practice

Intelligent cooling is not just about saving energy. NEXAIRA.Systems continuously monitors vibration patterns, filter conditions, and heat exchanger fouling. The system detects anomalies before they become failures – and in a data centre, this is not a comfort feature but an availability issue.

A clogged filter that reduces the air supply of a chiller module by 15 percent does not trigger an immediate alarm. But it forces neighbouring modules to compensate, increases their wear and drives up energy consumption. NEXAIRA detects such cases, reports them, and suggests the optimal maintenance window – aligned with the operational calendar.

For security-conscious IT teams, this matters: cooling failure is the second most common cause of unplanned data centre downtime after power outages. Those who address this preventively protect not only hardware but also SLAs.

The greatest efficiency reserve lies not in individual components, but in their interplay. A cooling system that adapts its setpoints to load and environment in real-time saves more than any hardware upgrade.
ebm-papst, NEXAIRA.Systems product documentation

ROI: What CFOs care about

50 percent less cooling energy sounds like a marketing claim. Validation in live operation shows it is realistic. For a mid-sized data centre with 5 MW IT load and a 35 percent cooling share, 50 percent savings means roughly 8,750 MWh annually. At an industrial power price of 15 cents per kWh, that is over 1.3 million euros per year.

Implementation requires no investment in new refrigeration machines. Costs lie in software deployment, sensors, and BMS integration. Payback typically runs under 18 months.

For ESG reporting, NEXAIRA.Systems delivers usable data via a dashboard: energy consumption per cooling circuit, CO2 savings per month, efficiency trends over time. This simplifies CSRD compliance and makes sustainability progress demonstrable to investors and regulators.

Specifically: the dashboard aggregates consumption data from all cooling circuits, automatically calculates Scope 1/2 emissions, and exports results in standard formats for the sustainability report. For facility managers, reporting effort drops from days to hours. For CFOs, the cooling plant becomes a documented efficiency project with measurable ROI instead of a cost driver.

Learn more about NEXAIRA.Systems and intelligent cooling for data centres here.

Frequently Asked Questions

Does NEXAIRA.Systems work with cooling systems from other manufacturers?

Yes. The system communicates via standard protocols (MODBUS-RTU, Ethernet, WiFi) and can be integrated with any BMS. The AxiBlade.Perform fans from ebm-papst offer native integration, but are not required.

Do operational data need to go to the cloud?

No. NEXAIRA.Systems operates entirely on-premise. All operational and sensor data stay in the local network. Optional cloud connectivity for remote monitoring is available, but not required.

How long does implementation take?

Integration occurs during live operation without downtime. Sensor installation, BMS connectivity, and digital twin calibration take a few weeks depending on system complexity. AI optimisation begins immediately after the calibration phase.

What happens if NEXAIRA fails?

The existing BMS takes over control with the last known setpoints. NEXAIRA.Systems is an optimisation overlay, not a replacement for base-level control. Cooling continues in all cases.

What savings are realistic?

The 50 percent figure refers to cooling energy, not total power consumption. In practice, one German data centre operator achieved 900 MWh and 240 tonnes of CO2 savings annually. Actual results depend on system size, age, and prior optimisation level.

Image source: Pexels / Brett Sayles (px:4508751)