Ask a data centre design team which standards govern the building and you will get a confident answer. The EN 50600 series. Uptime Institute Tier certification. Perhaps a hyperscaler’s own construction specification layered on top. These are serious frameworks, well understood by everyone in the room, and they shape decisions from the incoming supply to the raised floor.
Ask the same team which standards govern the atmosphere inside the battery room, and the answer tends to get quieter.
That gap is not carelessness. It is structural. EN 50600 puts safety explicitly outside its scope. The Tier Standard concerns itself with availability and maintainability, not with what happens to a person who opens a plant room door. Neither document is deficient, they simply were never written to answer the question. The difficulty is that the obligation does not disappear just because the design standard is silent on it.
The duty sits somewhere else entirely
Gas safety for a data centre is not held in one place. It is distributed across refrigeration law, electrical standards, combustion and ventilation guidance, and occupational health regulation. Each of those regimes has its own vocabulary, its own thresholds and, critically, its own owner on a project team. Nobody’s job title says “atmospheric risk”, so the pieces get picked up by whoever is nearest.
Batteries and energy storage are the clearest example of how quickly this moves. In the United States, NFPA 855 sets the frame, and its 2026 edition has raised the bar considerably: passive venting on its own is no longer accepted as the primary means of explosion control. The expectation now is active prevention in line with NFPA 69 – detection interlocked to exhaust, holding concentration below a quarter of the lower flammable limit, with backup power feeding the explosion-control system. NFPA 76, NFPA 111, NFPA 70 and the International Fire Code all touch battery room ventilation and hazardous location classification, while IEEE and ASHRAE Std 1635 supply the ventilation calculation, and UL 9540 and UL 9540A provide the test evidence behind spacing and mitigation claims.
In the UK and Europe the same duty lands in BS EN IEC 62485-2 and EN 50272-2, where accepted design practice aims to keep hydrogen below 1% by volume – a quarter of the level at which the atmosphere becomes explosive.
Cooling brings a different set of documents again. EN 378 is the principal European safety standard for refrigerating systems, covering charge limits by occupancy type, machinery room requirements, and where fixed detection and alarms are required. ISO 5149 does the equivalent job internationally and ASHRAE 15 in North America, with ASHRAE 34 classifying refrigerants by toxicity and flammability, including the A2L blends now appearing in new plant. Above defined thresholds, the F-Gas regime in the EU and its retained UK equivalent require automatic leak detection, with the detector itself subject to annual checking and to performance requirements set out in EN 14624.
Combustion and occupational exposure are a third strand. EH40 sets UK workplace exposure limits – 20 ppm over eight hours for carbon monoxide, 0.5 ppm for nitrogen dioxide, with short-term limits above those. OSHA 1910.1000 performs the same function in the United States, and COSHH requires exposure to be prevented or adequately controlled. Generator exhaust dispersion and separation from air intakes is arguably the most consequential requirement of all, and it sits with a ventilation standard such as ASHRAE 62.1 rather than with any gas standard, which is precisely why it has been overlooked, and why people have been hospitalised as a result.
Where the wrong document gets specified
Detector performance standards are a recurring gap. They differ by gas, and picking the wrong one is easy: EN 45544 covers toxic gases, EN 50104 oxygen, EN 14624 refrigerants and ISO 26142 hydrogen. A detector certified to one is not automatically appropriate for another.
Functional safety is the second common miss. As soon as detection initiates ventilation, isolation or shutdown, it has stopped being a monitoring system and become a safety function, which brings IEC 61508 and IEC 61511 and a SIL requirement into scope. NFPA 69 makes the connection explicit – higher permitted concentrations are conditional on interlocks demonstrated through a SIL 2 rated system.
Two sources of obligation that never appear on the drawing
Insurers routinely ask for more than regulators do. FM Global guidance for battery energy storage looks for emergency exhaust to activate at 10% LEL hydrogen – a threshold that is meaningless unless gas detection is installed to see it. Property standards address refrigerant machinery rooms in comparable terms.
Hyperscaler and colocation requirements often exceed local code as well, and they arrive contractually rather than as advice. Increasingly they are specific about detection and ventilation in both battery and mechanical spaces.
What good practice looks like
Taken together, these frameworks ask organisations to do four things: assess hazardous atmospheres and classify zones where flammable gases may be present; use certified equipment appropriate to the hazardous area and the region; interlock detection to action so ventilation, alarm and shutdown respond automatically; and verify system reliability, including functional safety performance, with audit-ready documentation to show it.
None of that is a one-off exercise. Every change of battery chemistry, refrigerant blend, cooling method or suppression agent alters the hazard profile and resets the assessment. Compliance is a moving target, and the organisations that treat it as a floor rather than a finish line are the ones that stay operating.
If you would like to review how your current gas detection arrangements map against these obligations, contact a Crowcon specialist to start the conversation.
INDUSTRY INSIGHTS
Subscribe to get the
latest insights
in your inbox
Read about Crowcon‘s Privacy and Cookie policy here. If you change your mind, you can unsubscribe at any time




