Selecting a Clean Agent Fire Suppression System for US Data Centers

Clean agent selection in the United States is not a product choice. It is a set of constraints that resolve into a product choice, and the constraints arrive from four directions that do not coordinate with each other: the fire protection standards that define how much agent and how fast, the occupational exposure limits that define whether a person can be in the room when it goes off, the environmental rules that define what you are allowed to buy and recharge after 2026, and the physics of the equipment you are protecting. Get the order of operations wrong and you end up with a system that passes inspection and fails the room.

This is a selection guide, not a product catalogue. It works through the US regulatory stack, the agent families and their measured parameters, the design quantities that follow from them, and the two constraints that most often invalidate an otherwise correct selection — enclosure tightness and acoustic energy. It ends with the battery room, which is now the reason most clean agent selections have to be revisited.

The US standard stack, and which document owns which decision

Four documents govern a data center clean agent system in the US, plus one insurer data sheet that is not a standard and is treated like one. Selecting an agent before knowing which of these applies to the room is the most common order-of-operations error.

Table 1 — The US clean agent stack. Selection decisions cross all five documents; no single one is sufficient.

One boundary is worth stating explicitly, because it removes CO2 from consideration in almost every data center application. Carbon dioxide is not a clean agent under NFPA 2001; it is covered by NFPA 12, and it is prohibited for use in spaces that are normally occupied [1][7]. A CO2 system in a white space that operators walk through is a life-safety finding, not a design preference. The other halogenated legacy agent, Halon 1301, has not been produced in the US since 1994 and is only available as reclaimed stock [7].

What NFPA 75 requires before agent selection even starts

NFPA 75 sets the conditions the clean agent system has to satisfy, and two of its provisions change the answer more than any agent property. The first is detection sensitivity: large IT areas are expected to use very early warning detection — aspirating smoke detection (ASD), commonly referred to by the VESDA brand — rather than relying on spot smoke detectors alone, because a clean agent system's advantage is that it acts on a fire measured in kilowatts, not megawatts [3][8]. If detection is slow, the agent selection is being asked to compensate for a detection problem it cannot fix.

The second provision is the one that quietly invalidates designs. NFPA 75 requires that the enclosure of IT equipment — including containment systems — must not prevent the clean agent from reaching the equipment, and where contained aisles would stop the agent from achieving design concentration throughout the protected volume, the suppression system has to be modified to match [3]. A hot-aisle or cold-aisle containment system is a physical obstruction to a total-flooding agent. Aisle doors, end-of-row panels, and overhead baffles turn one room into a set of connected compartments with different filling behaviour, and the containment that was added for thermal efficiency becomes a fire protection constraint. The same halls that added containment for density reasons are usually the ones running hybrid air-liquid heat rejection, where the containment layout is already a design variable.

NFPA 75 also drives the detection placement that the suppression design inherits: below the raised floor, above the ceiling in the return plenum, and at the rack inlet, because those are the three volumes where a cable or a power connection can smoulder out of sight of the room detector [3]. Each of those volumes is also a separate flooding volume for agent quantity purposes, which is why subfloor and ceiling plenum suppression is a quantity question as much as a detection question.

The agents, and the parameters that actually decide between them

US data centers select from four agent families. Each has a representative product, and each family behaves differently in all four constraint directions. The following table is the comparison that matters; the numbers are the standard's own design values, not vendor claims about them.

Table 2 — Agent comparison on the parameters that decide a data center selection. Design concentrations and NOAEL/LOAEL from NFPA 2001; unit-volume quantities are the standard's Annex A values at Class A design concentration.

Reading the table correctly

Three of these rows are frequently misread, and each misreading produces a wrong selection.

Design concentration is not the minimum extinguishing concentration. The design concentration is the extinguishing concentration multiplied by a safety factor, and NFPA 2001 sets that factor at 1.2 for Class A surface fires where detection and actuation are automatic, 1.3 for Class B flammable liquid fires, and the result may not fall below the heptane minimum [1][2]. For Class A hazards, the design concentration must also be at least the greater of that calculated value and 95 percent of the NOAEL when the space is normally occupied [1]. The last clause is the one that binds in practice: a room with flammable liquids stored in it is a Class B hazard with a 1.3 factor, and the resulting concentration can push straight through the exposure ceiling.

The NOAEL is a design ceiling, not a safety margin to trade against. For halocarbon agents in normally occupied spaces, NFPA 2001 caps the design concentration at the NOAEL and limits human exposure to five minutes [1][2][7]. At 7.0 percent design against a 9.0 percent NOAEL, HFC-227ea sits with roughly 2 percentage points of headroom — comfortable. At 4.7 percent against 10 percent, FK-5-1-12 sits with 5.3 points — more than twice the margin. That is the real reason FK-5-1-12 is preferred in occupied control rooms and network operations centres, and it is a standards consequence, not a marketing one.

Inert gases are not "safer because they are natural." They displace oxygen to extinguish, which means the exposure limit is an oxygen limit. NFPA 2001 permits up to 43 percent concentration — roughly 12 percent oxygen — for five minutes, 43 to 52 percent for three minutes, and concentrations above 52 percent only in normally unoccupied spaces with a 30-second maximum exposure window [1][2][7]. An IG system designed to a 42 percent concentration in a room with people in it has essentially no margin left for the day a fire is larger than the design fire.

The arithmetic: what the selection costs in cylinders, pipe, and vent area

Agent selection determines three physical quantities that drive most of the installed cost. Working the same 800 m2, 4.2 m high white space — 3,360 m3 — through both halocarbon and inert options makes the divergence concrete.

Quantity arithmetic. Unit-volume values are NFPA 2001 Annex A figures at Class A design concentration; bank footprint assumes a standard 180 L cylinder on a two-tier rack.

The mass difference between the two halocarbons looks small — 11 percent — and it is. The floor-area difference is negligible. What is not small is everything downstream of the agent volume: inert gases require pipe sized for a 60-second or 120-second discharge of a large gas volume, which means larger bore, more nozzles, and a distribution network that has to be flow-calculated with the system manufacturer's listed software [1]. The halocarbons discharge in ten seconds or less, through smaller pipe, with fewer nozzles.

Enclosure integrity: the constraint that fails more systems than agent choice

A clean agent system is designed against a room that holds agent, and almost no data center room holds agent the way the calculation assumes. Enclosure integrity is where the design meets the building, and it is the most common source of a system that was correct on paper and ineffective in service.

NFPA 2001 requires the enclosure to hold at least 85 percent of the minimum design concentration at the highest level of the protected contents for ten minutes, or for as long as is necessary for trained personnel to respond [1][2]. In an unmanned data hall the authority having jurisdiction can and often does require longer; 20 to 30 minutes appears in insurer expectations for high-value spaces. The mechanism is a descending interface: agent leaks out of the lower portion of the room and air is drawn in above it, so the concentration the equipment sees is the concentration above the interface, and it falls as the interface descends [12].

The practical measurement is the door fan test, which NFPA 2001 describes in Annex C. A calibrated fan is mounted in a door, a pressure differential of about 10 Pa is established, the leak flow is measured, and the result is converted into an equivalent leakage area (ELA) — the size of the single hole that would produce the same flow at that pressure [1][12]. The ELA then drives the predicted hold time.

Hold-time sensitivity. The model is the standard descending-interface form; the leakage inventory is the practical order in which remediation pays off.

The remediation order matters because the first two items on that list are usually responsible for more than half the measured ELA. Cable penetrations are the standing problem: every change to the cable plan opens a hole, and firestop that is rated for fire does not necessarily seal for agent retention. A room that passed its door fan test at commissioning can fail it after two years of adds and moves without anyone having touched the suppression system. Subfloor leakage is also the reason a raised floor changes the enclosure calculation rather than just the cable route — the plenum is a second protected volume with its own leakage path, which is why subfloor suppression is sized separately. Enclosures that are being brought up to standard alongside a liquid cooling retrofit should be measured once, after the coolant loop penetrations are in, not twice.

Two leakage areas, not one

The current version of NFPA 2001 recognizes that the enclosure has to be measured twice, for two different purposes [12]. The hold-time calculation uses the total leakage area. The peak-pressure evaluation uses the below-ceiling leakage area (BCLA), which is roughly the half of the total that lets agent out near the top and thereby relieves the discharge pressure. A room with a large ELA concentrated at floor level holds agent badly; a room with the same ELA concentrated at ceiling level relieves pressure well but may still retain agent acceptably. Treating the two as one number produces either an unnecessary pressure relief vent or a structural failure.

Peak pressure: the failure mode that damages the building

A clean agent discharge is a mass of gas entering a fixed volume in a controlled time, and the enclosure sees the result. Inert gas discharges present as a single large positive pressure spike, because the agent is stored as compressed gas and arrives at ambient temperature. Halocarbon discharges are more complex: the agent vaporizes and cools, producing an initial negative pressure excursion followed by a positive spike, and both can be larger than the design team expects [12][13].

Measured values from controlled discharges in a single enclosure, at each agent's minimum design concentration with comparable leakage areas, show the split clearly [13]:

Table 3 — Representative peak pressures during discharge, same enclosure, comparable leakage areas. If the enclosure's structural load limit is below these values, venting or structural reinforcement is required.

The negative pressure row is the one that gets missed, because "over-pressure" is the intuitive hazard. FK-5-1-12 produced the largest negative excursion measured in that series and the smallest positive one, which means a room built for an inert system's positive spike is not automatically adequate for a fluoroketone's suction phase. Suspended ceilings, lightweight partition walls, and doors that open outward are the elements that fail in the negative direction.

This is also why the industry's pressure relief sizing is framed as a solution space rather than a single number. A research programme run for the NFPA 2001 committee and published through NIST established that the older inert-gas peak-pressure formulae under-predicted actual pressures, that halocarbons can generate pressures as high as inert agents under some conditions, and that humidity influences the halocarbon result [14]. The design construct that came out of it is the vent-to-volume ratio: a horizontal axis of enclosure leakiness in cm2 of leakage area per m3 of volume, with a peak-pressure curve setting a lower bound and a retention-time curve setting an upper bound. The acceptable designs live in the wedge between them.

Solution-space arithmetic from the NIST research programme behind the current NFPA 2001 peak-pressure method. The practical reading: retention time and vent area are a single coupled decision.

Acoustic energy: the constraint that damages the equipment

Agent selection has a failure mode that has nothing to do with fire. An inert gas discharge moves a large volume of gas through nozzles in under two minutes, and the resulting sound field can destroy hard disk drives. This is not theoretical, it is measured, and it has taken a bank offline.

The evidence chain runs from laboratory measurement to a real outage. Siemens Corporate Technology established that hard disk performance can be affected from 110 dB, with disruption typically starting at 120 dB and the worst sensitivity in the 500 Hz to 8 kHz band; conventional nozzles were measured at roughly 130 dB, and some orifice-flow nozzle designs exceed 145 dB of sound power [11][15][16]. Independent work presented by Tyco reports the same range, with orifice-flow inert systems above 145 dB and regulated-flow systems around 138 dB [16]. In 2017 an ING Bank data centre in Bucharest was taken offline for ten hours by an inert gas discharge test: the sound field vibrated the drive chassis, the vibration reached the read-write heads, and the drives came off track. The bank's own monitoring recorded levels above 130 dB, the top of its instrumentation range [17].

Acoustic attenuation arithmetic. The 6 dB per doubling rule means nozzle layout, not just nozzle choice, determines what a drive experiences.

Three consequences follow for selection. First, for any data hall with spinning disk storage, the choice is between a halocarbon, which discharges a small volume quickly and does not produce an HDD-damaging sound field, and an inert gas with acoustic nozzles and a layout that keeps nozzles off direct lines to racks [6][11][16]. Second, where inert gas is required and acoustic nozzles are used, a room acoustic calculation is part of the design package — the attenuation from nozzle to drive is a room property, not a nozzle property [16]. Third, all-flash halls have far less exposure to this failure mode, and the acoustic requirement can be relaxed accordingly, which is one of the few places where a storage technology decision changes a fire protection selection.

People: time delay, abort, and the 2026 exposure rule

NFPA 2001 separates the question of whether an agent can extinguish a fire from the question of whether a person can be in the room when it does. The exposure limits are tied to concentration and duration, and the current edition tightens them.

For halocarbon agents in normally occupied spaces the design concentration may not exceed the NOAEL and exposure is limited to five minutes [1][2][7]. For inert gases the limits are 43 percent for five minutes, 43 to 52 percent for three minutes, and above 52 percent only in unoccupied spaces with a 30-second window [1][2][7]. Systems in occupied spaces require a time delay — typically 30 or 60 seconds — with both audible and visual notification running through the countdown, and an abort switch that is a dead-man type requiring constant pressure, located near the means of egress so a person can hold it while confirming a false alarm and then leave before releasing [7].

An amendment issued in 2026 adds a requirement that changes how high-concentration designs are approved. TIA 25-1, effective 5 May 2026, revises NFPA 2001 section 4.3.4 to require an egress time study where design concentrations carry human exposure times shorter than five minutes: such concentrations "shall not be permitted in normally occupied spaces unless an egress time study is performed relative to the reduced exposure times and approved" [18]. The annex the TIA points to frames the study as a comparison between available safe egress time (ASET) and required safe egress time (RSET), using the methods in the SFPE Handbook, or alternatively an actual timed egress simulation in the protected space [18].

The 2026 supply reality: AIM Act, HFC, and PFAS

Agent selection now has a procurement half that did not exist a decade ago. Two separate regulatory tracks affect the two halocarbon families from opposite directions, and the resulting situation is asymmetric in a way that a selection decision has to account for.

HFC-227ea under the AIM Act

The American Innovation and Manufacturing Act directs the EPA to phase down HFC production and consumption by 85 percent from historic baseline levels by 2036, on the Kigali schedule. The cap steps down: 60 percent of baseline from 2024 through 2028, 30 percent from 2029 through 2033, and 15 percent from 2036 [9]. Fire suppression is a covered sector, and the 2024 Emissions Reduction and Reclamation rule — the first rule under subsection (h) — added management requirements that begin to bite in 2026 [19][20].

Table 4 — AIM Act fire suppression compliance dates under the Emissions Reduction and Reclamation rule. Mission-critical military uses and onboard aerospace applications are excluded.

The practical effect on selection is a maintenance-path question, not a supply-cutoff question. Virgin HFC-227ea becomes progressively scarcer and more expensive as the allowance cap steps down, and from 2030 a new system must be charged with reclaimed agent. A facility selecting an HFC system today is committing to a service life in which every recharge is a reclaimed-supply transaction with an associated recordkeeping obligation. That is workable — the reclaimed supply chain exists and the industry has done this before with halon — but it belongs in the total cost comparison, because it is a recurring operational obligation rather than a one-time procurement.

Charge arithmetic. GWP values are the 100-year figures used in the AIM Act and the SNAP programme; the loss assumption is illustrative of service-life losses, not a regulatory figure.

FK-5-1-12 under the PFAS track

The fluoroketone is in a different position, and the distinction is chemical rather than commercial. FK-5-1-12 is not an HFC and is therefore not subject to the AIM Act phase-down; it remains acceptable under the EPA's SNAP programme, and it remains in production with multiple UL-listed and FM-approved system manufacturers [10]. Its atmospheric lifetime is measured in days and its 100-year GWP is 1, so the climate argument that drives the HFC phase-down does not apply to it in the same way.

What changed is the supply landscape. 3M, the original manufacturer of the Novec 1230 fluid, announced its exit from PFAS manufacturing and ceased production of the fluid at the end of 2024 as part of a broader exit completed in 2025 [10]. The significance is easy to overstate and easy to understate. Overstating it: FK-5-1-12 is not withdrawn from the market, other producers continue to make the chemical, and the systems built on it remain listed and supported. Understating it: the brand-name fluid supply is no longer controlled by the company that commercialized it, and the pricing and availability of the fluid becomes a function of a smaller producer base.

On the regulatory side, the state-level picture is active but has so far concentrated on PFAS in firefighting foam rather than on gaseous clean agents. In the 2026 legislative sessions, roughly 23 states carried forward about 280 PFAS-related bills and 17 states introduced close to 100 new ones, with the enacted and pending measures targeting AFFF — Florida's HB 1019 phasing out PFAS foam through 2029, Missouri's HB 1925 barring PFAS foam sales from 1 January 2028, and Florida's CS/SB 1230 prohibiting AFFF use for non-emergency training from 1 July 2026 with a sales ban from 2027 and a possession prohibition from 2028 [21]. As of early 2026 there is no federal regulation specifically governing the disposal of PFAS-based gaseous extinguishing agents [21]. A designer specifying FK-5-1-12 is not currently exposed to a gaseous-agent PFAS ban, but should track the state legislation rather than assume the foam-focused drafting will stay foam-focused.

The battery room is a different problem, and it is now the reason selections get reopened

Lithium-ion battery energy storage and lithium-ion UPS rooms have changed clean agent selection more than any other development in the last five years, and the change is not a matter of picking a better agent. It is a matter of accepting that clean agents are not a complete answer for the hazard.

The physics is the constraint. A lithium-ion cell in thermal runaway generates its own oxidizer as the cathode decomposes, so it can sustain combustion without external oxygen, and the reaction produces its own heat. A clean agent that works by displacing oxygen or by interrupting the combustion chain can knock down the visible flame, but it cannot cool the cell, and it cannot stop propagation from one cell to the next [22][23]. A clean agent system in a battery room protects the room; it does not stop the cell.

NFPA 855 is the governing standard for stationary energy storage installation, and it treats lithium-ion installations as a distinct hazard class rather than as a variation of an electrical room. The thresholds matter for selection: installations at or above 20 kWh of lithium-ion capacity fall under the standard's scope, compared with 70 kWh for lead-acid, and a lithium-ion installation above 600 kWh inside an occupied building is directed into a dedicated fire-rated room [23]. Separation distances default to 3 ft and can be reduced with unit-level test data [23].

Two technical points are worth getting exactly right, because they are widely stated incorrectly. First, UL 9540A is a test method, not a product certification; it characterizes how a specific battery system behaves in thermal runaway — whether it propagates, and what gases and energy it releases — and the output feeds the hazard analysis. The product-level listing is UL 9540. A specification that asks for "UL 9540A certified batteries" is asking for something that does not exist [23]. Second, the 2026 edition of NFPA 855 requires a hazard mitigation analysis for essentially all lithium-ion installations, and it no longer accepts a NFPA 68 deflagration venting design as the primary explosion protection strategy; the emphasis moves to preventing the accumulation of a flammable mixture through NFPA 69 active controls, combined with early gas detection [23]. The same project-level thinking — a hazard analysis that decides the architecture before the equipment is specified — is what a DCIM scope is supposed to enforce on the electrical side.

What the battery room selection actually looks like

The strategy that the standards now describe is detect-then-manage rather than flood-and-extinguish. Off-gas detection for hydrogen, carbon monoxide, and volatile organic compounds at thresholds around 25 percent of the lower flammability limit gives an early warning measured in minutes to hours before a visible fire; the response sequence is ventilation, isolation of the affected unit, and cooling, with clean agent available for the flame but not relied on for the hazard [22][23]. Because the gases produced in thermal runaway include flammable species, the detection system that matters for safety in a battery room is a gas detection system, and the suppression system is downstream of it.

Battery room sequence. The clean agent decision sits fifth, not first — which is the practical difference between a battery room and a white space.

A selection procedure that produces one answer

Working the constraints in the right order collapses the choice quickly in most cases. The sequence below follows the dependencies rather than the agent list.

1. Classify the hazard, then check the occupancy. Class A only, or Class A plus a flammable liquid store? Normally occupied or not? These two questions set the required safety factor and the exposure ceiling, and the exposure ceiling is often the binding constraint.
2. Check the containment and the subfloor. Contained aisles, subfloor cable volume, and ceiling plenum each change the flooding volume and the filling behaviour. If containment obstructs the agent, resolve that before choosing an agent — it may change the required quantity enough to change the answer.
3. Measure the enclosure as it will actually be built. Door fan test with the below-ceiling leakage measured separately, giving two leakage areas for two calculations. A design built on an assumed ELA is a guess with a calculation attached.
4. Check the structures against the agent's pressure signature. Positive and negative, not just positive. Compare against the enclosure's load limit, and resolve the venting versus retention conflict inside the solution space.
5. Check the equipment against the agent's acoustic signature. Spinning disk storage in the room means either a halocarbon or an inert system with acoustic nozzles and a room acoustic calculation. All-flash relaxes this.
6. Check the occupancy against the concentration band. If the design concentration lands in the short-exposure band, the 2026 amendment requires an egress time study. Prefer a different agent over defending the study.
7. Check the service life against the supply rules. An HFC selection carries a reclaimed-supply and recordkeeping obligation from 2026 and a recycled-charge requirement from 2030. A fluoroketone selection carries a producer-base question and a state-legislation watch.
8. Only then select the product, and confirm the flow calculation is done with the system manufacturer's listed software, because the hydraulic behaviour is a listed-system property, not an agent property.

Two of those steps — the flooding volume and the enclosure leakage measurement — depend on the physical configuration of the room, which is also what determines the containment, the leak detection layout and the subfloor design. If you are specifying a white space or a battery room and need the equipment side of that scope priced alongside the suppression design, send us the project scope with the room dimensions, the containment layout and the agent family you are considering.

Frequently asked questions

Is FM-200 or Novec 1230 better for a US data center?

They are close enough on performance that the deciding factors are elsewhere. FK-5-1-12 has the larger NOAEL margin — 10 percent against a 4.7 to 6.0 percent design band, versus 9 percent against 6.7 to 8.7 percent for HFC-227ea — and a GWP of 1 against 3,220. HFC-227ea has a longer-established installed base and a more settled reclaimed supply chain. Where the room is normally occupied, the fluoroketone's margin is the tiebreaker. Where a large legacy HFC fleet already exists and technician familiarity matters, the HFC case is stronger. Neither is wrong; the exposure margin and the 2026 supply rules are what decide it.

Can we use inert gas in an occupied control room?

Yes, within the exposure limits, which are concentration and duration based: up to 43 percent for five minutes, 43 to 52 percent for three minutes, and above 52 percent only in normally unoccupied spaces with a 30-second maximum. A design that lands near the top of the permitted band leaves no margin for a fire larger than the design fire, and from May 2026 a design reaching into the short-exposure band requires an egress time study that the authority having jurisdiction has to approve. In a room that is genuinely occupied, the larger NOAEL margin of a halocarbon is usually the simpler answer.

How often does the door fan test have to be repeated?

At commissioning, and then on a recurring basis — the common industry interval is every three years, plus any time the enclosure changes materially. Cable penetrations, added CRAC units, new doors, and raised-floor work all change the leakage area, and a room can lose half its predicted hold time without anyone touching the suppression system. Where containment has been added or modified, the test should be repeated before the change is signed off, because the containment changes both the leakage and the filling behaviour.

Does aisle containment break the clean agent system?

It can. NFPA 75 requires that IT equipment enclosures, including containment, not prevent the agent from reaching the equipment, and where containment would prevent design concentration from being achieved throughout the volume, the suppression system must be modified. The practical mitigations are nozzles inside containment, pressure-relief provisions so agent can equalize through the aisle doors, or additional concentration margin — the last of which has to be checked against the exposure limits. The check belongs in design, before the containment is installed.

Why does an inert gas discharge destroy hard drives when a clean agent discharge does not?

Sound power, not chemistry. An inert system has to move a large gas volume in 60 to 120 seconds, and orifice-flow nozzles can exceed 145 dB of sound power; conventional nozzles have been measured around 130 dB, and hard disk degradation has been documented from 110 dB and disruption from 120 dB, concentrated in the 500 Hz to 8 kHz band. Halocarbon systems discharge a much smaller mass in under ten seconds and do not produce a damaging sound field. Mitigation for inert systems is acoustic nozzles, which reduce sound power by roughly 7 dB, plus nozzle placement that avoids direct lines to racks, plus a room acoustic calculation — and the requirement is written into insurer data sheets, not just the standard.

What is the difference between UL 9540A and UL 9540?

UL 9540A is a test method for evaluating thermal runaway fire propagation in battery energy storage systems; the test report describes how a specific system behaves and feeds the hazard mitigation analysis. UL 9540 is the product safety standard under which battery energy storage systems are listed. A specification asking for "UL 9540A certified" batteries is asking for a certification that does not exist — the correct language is a system listed to UL 9540 with a UL 9540A test report available. The distinction matters because the two documents support different decisions.

What changes for a clean agent system in 2026?

Three things. From 1 January 2026, no knowing venting of HFCs during installation, servicing, repair, or disposal, and servicing must use recycled HFCs. By 1 June 2026 — or within 30 days of hiring — fire suppression technicians must complete HFC handling training. From 5 May 2026, the NFPA 2001 amendment requires an egress time study where design concentrations imply human exposure times shorter than five minutes in normally occupied spaces. None of the three changes what an agent can extinguish; all three change what a facility has to document.

Can we still buy HFC-227ea in the US?

Yes. The AIM Act phases down production and consumption, not possession or use, and existing equipment does not have to be replaced. What changes is the supply path: the allowance cap steps to 30 percent of baseline for 2029 through 2033 and 15 percent from 2036, servicing from 2026 must use recycled agent, and from 2030 a new system's initial charge must also be recycled. Price and lead time are the mechanisms through which the phase-down reaches an individual facility, and a maintenance contract should be written with reclaimed supply and the recordkeeping obligation in mind.

References

1. National Fire Protection Association, "NFPA 2001: Standard on Clean Agent Fire Extinguishing Systems, 2025 edition" — design concentration and safety factors, NOAEL and LOAEL exposure limits, discharge times, hold time and enclosure integrity, cylinder and pipe network requirements. nfpa.org
2. Selvi Fire, "Clean Agent Fire Suppression and NFPA 2001" — 2025-edition design concentration by hazard class, safety factor values, agent-by-agent NOAEL and design concentrations, hold time at 85 percent of minimum design concentration, door fan test method, installation details. selvi.org
3. National Fire Protection Association, "NFPA 75: Standard for the Fire Protection of Information Technology Equipment" — IT equipment area protection, very early warning detection expectations, and the requirement that enclosures including containment not obstruct agent distribution to equipment. nfpa.org
4. National Fire Protection Association, "NFPA 76: Standard for the Fire Protection of Telecommunications Facilities" — detection and suppression expectations for carrier and telecom spaces. nfpa.org
5. National Fire Protection Association, "NFPA 72: National Fire Alarm and Signaling Code" — detection zoning, notification, monitoring, and release circuit requirements for suppression systems. nfpa.org
6. FM Global, "Data Sheet 5-32: Clean Agent Extinguishing Systems" — insurer requirements for clean agent systems in data centers, including hard disk drive protection and acoustic considerations. fmglobal.com
7. US Made Supply, "NFPA 2001: Clean Agent Fire Systems" — occupied space concentration limits and exposure durations by agent family, time delay and abort switch requirements, halon and CO2 status, inspection and maintenance intervals, portable clean agent extinguisher standards. usmadesupply.com
8. Sovereign Extinguishing Systems, "Impact of Noise on Computer Hard Drives" — Eurofeu test findings on sound pressure thresholds for drive failure, enterprise versus nearline drive sensitivity, and mitigation by nozzle spacing, sound-damped nozzles and drive enclosures. sovereignextinguishing.co.uk
9. US Environmental Protection Agency, "Phasedown of Hydrofluorocarbons: Management of Certain Hydrofluorocarbons and Substitutes under Subsection (h) of the American Innovation and Manufacturing Act — Final Rule Fact Sheet" (October 2025) — AIM Act phase-down schedule to 85 percent below baseline by 2036, reclamation standard, leak repair and automatic leak detection provisions, fire suppression initial installation requirement. epa.gov (PDF)
10. Rotaflow, "Paint Booth Fire Suppression" — FK-5-1-12 status as a halocarbon clean agent outside the AIM Act phase-down, SNAP acceptability, NOAEL and LOAEL framework, and periodic functional testing requirements. rotaflow.ca
11. Siemens Corporate Technology via Informationsdienst Wissenschaft, "'Quiet' Fire Extinguisher Protects Hard Disks" — hard disk damage from 110 dB and above, conventional nozzle measurement of approximately 130 dB, and the design targets of a low-noise nozzle below 100 dB. idw-online.de
12. Society of Fire Protection Engineers, "Clean Agent Enclosure Design" (Fire Protection Engineering, Issue 74) — descending interface model, door fan leakage measurement, peak pressure as a function of leakage-to-volume ratio, and the current NFPA 2001 requirement to measure two separate leakage areas for hold time and pressure evaluation. sfpe.org
13. 3M, "Preserving Room Integrity During Clean Agent Discharges" — measured peak positive and negative enclosure pressures by agent, including the negative pressure excursion characteristic of vaporizing halocarbons and the structural pressure load limits of typical construction. multimedia.3m.com (PDF)
14. National Institute of Standards and Technology, "Preventing Excessive Enclosure Pressures During Clean Agent Discharges" — vent-to-volume ratio solution space bounded by peak pressure and retention time, below-ceiling leakage area measurement, and the narrowing of the solution space for longer retention targets and for HFC-227ea. nist.gov (PDF)
15. DCN (Data Centre News), "Noise damage in the data centre" (February 2017) — inert gas discharge noise above 140 dB, acoustic shock effects on drive head alignment, disruption from 120 dB with some degradation below 110 dB, and constant discharge technology reducing relief flap sizing. view.joomag.com
16. Tyco, "The Impact of Sound on Computer Hard Disk Drives and Risk Mitigation Measures" (white paper) — nozzle sound power measurements for orifice flow and regulated flow inert systems, acoustic nozzle performance and coverage, and the requirement for room acoustic calculations to predict drive-level sound pressure. kipdf.com
17. Vice, "A Loud Sound Just Shut Down a Bank's Data Center for 10 Hours" — ING Bank Bucharest inert gas discharge test, measured levels above 130 dB, hard drive vibration and head misalignment, and the ten-hour recovery. vice.com
18. National Fire Protection Association, "Tentative Interim Amendment 25-1 to NFPA 2001, 2025 edition" — revision of section 4.3.4 requiring an egress time study for design concentrations associated with human exposure times shorter than five minutes in normally occupied spaces; issued 15 April 2026, effective 5 May 2026. docinfofiles.nfpa.org (PDF)
19. US Environmental Protection Agency, "American Innovation and Manufacturing Act: Fire Suppression Requirements for Hydrofluorocarbons" (February 2026) — compliance table for minimising HFC releases, recycled-agent servicing and repair, recycling prior to disposal, technician training by 1 June 2026, recordkeeping, and recycled-agent initial installation from 1 January 2030. epa.gov (PDF)
20. Trinity Consultants, "EPA's HFC Management Rule: Key Compliance Requirements" — consolidated compliance timeline for the Emissions Reduction and Reclamation rule, the 15-pound applicability threshold, and facility owner action list. trinityconsultants.com
21. Clean Energy/Environmental Legislative Tracker, "State PFAS legislation in the 2026 sessions" — carryover and newly introduced PFAS bills by state, enacted AFFF-phase-out measures including Florida and Missouri, and the absence of federal rules specific to PFAS-based gaseous extinguishing agents. Legislative session records
22. Energy Storage Safety Review, "Thermal runaway, off-gas detection and the McMicken Dam incident" — lithium-ion thermal runaway as a self-oxidising reaction, off-gas detection thresholds and response sequence, and the code activity that followed the 2019 Arizona explosion. energy storage technical review
23. National Fire Protection Association, "NFPA 855: Standard for the Installation of Stationary Energy Storage Systems" — applicability thresholds for lithium-ion and lead-acid installations, separation distances, dedicated room requirements above 600 kWh, the 2026 hazard mitigation analysis requirement, and the shift from NFPA 68 deflagration venting to NFPA 69 active prevention. nfpa.org