Data centers have spent the past two years racing to put lithium-ion batteries inside the building — not as a UPS afterthought, but as grid infrastructure. Batteries now jump interconnection queues, smooth out diesel-free backup power, and even sell frequency response back to the grid. What got less attention is that fire marshals, code bodies, and insurers have spent the same two years rewriting the rules those batteries now have to live under. For anyone underwriting a lease or a build in 2026, the battery room is no longer a mechanical-engineering footnote. It’s a line item that can move your schedule, your premium, and your site plan.
Why the batteries moved inside the fence line
The shift happened for practical reasons. On-site battery energy storage systems (BESS) let developers accept a smaller, faster grid interconnection and use stored charge to cover peak draw, a strategy TechInfraHub covered in detail as utilities’ interconnection queues stretched past 2,000 GW nationally. BESS also sidesteps the multi-year lead times now hitting diesel generators and large power transformers, and in several deregulated markets it earns revenue by discharging into the grid during demand spikes. The result: BESS is no longer a niche add-on at the edge of the site plan — multi-megawatt-hour installations are increasingly standard at hyperscale campuses, often sited directly adjacent to, or inside, the white space building envelope.
That proximity is the problem. A diesel yard sits outside the building and burns down to a known, bounded footprint. A lithium-ion battery room sits inside or against occupied space, and when a cell goes into thermal runaway, there is still no reliable way to extinguish it — crews largely contain and cool rather than put it out. The industry built a decade of backup-power doctrine around diesel. It is now retrofitting that doctrine, in real time, around chemistry that behaves very differently when it fails.
Installed BESS costs have also fallen enough — into a roughly $250 to $580 per kWh range as of 2026 — that adding megawatt-hours of storage is no longer a boutique decision reserved for flagship campuses. It shows up in mid-market colocation RFPs now, often bundled into the power story alongside on-site generation rather than called out as its own system. That’s exactly where fire-code and insurance diligence tends to get skipped: a buyer evaluating total power cost per kW rarely has a line item for “verify the sprinkler design was tested against this specific chemistry,” even though that’s now a material cost and schedule driver in its own right.
Moss Landing was the wake-up call
The incident that reset the conversation was the January 2025 fire at the Moss Landing battery storage facility in California, which forced the evacuation of roughly 1,200 nearby residents for about 24 hours. Moss Landing wasn’t a data center, but it was large-scale grid BESS, and it became the reference event that fire officials, insurers, and code committees cite when justifying tighter rules for any big lithium-ion installation — including the ones now going into data center campuses. Industry tracking cited by battery-safety researchers puts global utility-scale BESS failures at roughly 87 since 2012, or about one a month, even as the failure rate per gigawatt deployed has fallen more than 98% over six years as chemistry and management systems improve. The base rate is low and falling. The consequence of a single miss, next to a building full of racks and people, is what regulators are now pricing for.
NFPA 855’s 2026 rewrite raises the bar
NFPA 855, the US standard governing stationary energy storage installations, picked up a meaningful update for 2026. The new edition extends coverage beyond lithium-ion to iron-air, sodium-ion, and zinc-based chemistries now entering the market, and it adds mandatory hazard analysis — documented fire, explosion, and toxic-gas risk assessment before a facility is approved — plus annual emergency operations plan reviews, annual refresher training for on-site personnel, and formal coordination with local fire responders on response protocols. It aligns testing expectations with UL 9540A, the large-scale fire test standard insurers and authorities increasingly treat as the baseline rather than a best practice. For a buyer, the practical effect is that a design frozen against the 2023 edition of NFPA 855 may no longer satisfy the authority having jurisdiction by the time the facility reaches commissioning — and retrofitting hazard analysis and sprinkler validation after the building is framed is far more expensive than designing to the new edition from the start.
Need Expert Guidance?
Talk to a Data Center Expert
21+ years of hands-on experience in data center design, operations & infrastructure. Book a quick discovery call to discuss your project.
📞 Book a Discovery CallJurisdictions are moving in the same direction on their own timelines, and not always in sync with each other, which is its own diligence problem for multi-site portfolios. Fire and Rescue NSW issued a position statement in 2026 requiring 240/240/240-minute fire resistance for load-bearing battery room construction, with no reduction in that level, and asking that sprinkler design be validated against large-scale fire test data rather than assumptions carried over from other occupancies — a notably stricter bar than some developers had been designing to. The same statement reclassifies structures housing diesel generators and fuel tanks from a non-habitable category to an industrial one, pulling backup power infrastructure that used to be an afterthought into the same stringent review as the battery room itself. NSW’s regulator separately expects data centers to reach 12% of grid demand by 2030, which is part of why it is tightening the rules now rather than after the next incident. US states are moving too: California’s SB 283 and SB 38 push BESS projects toward the highest available safety standard, and Texas is advancing state-level permitting with mandatory emergency response plans attached. New York and Massachusetts have both seen local moratorium proposals specifically targeting battery storage, separate from the broader data center siting fights already playing out in those states.
Insurers are no longer treating this as a rounding error
The code changes are catching up to what underwriters already concluded. An estimated 28% of lithium-ion battery fires trace back to manufacturing defects rather than installation or operating error, which means the risk travels with the cell supplier, not just the integrator — a distinction insurers now ask about directly in underwriting questionnaires. Risk modelers put the direct asset loss from a fire at a typical 100 MWh facility at $25 million or more, before counting business interruption, regulatory fines, or remediation, and the dedicated BESS fire-protection and suppression market is forecast to grow from roughly $4.2 billion in 2025 to $12.8 billion by 2034 — a 13.2% compound annual growth rate that only makes sense if insurers and operators expect to keep spending on mitigation, not just hope the base rate stays low. That compounds the exposure TechInfraHub flagged in its broader look at data center insurance capacity: premium growth is already lagging capex growth sector-wide, and battery rooms are one of the specific line items pushing claims severity up even as frequency stays rare.
What this means before you sign
None of this makes on-site BESS a bad bet — the economics of queue-jumping and diesel avoidance are real, and failure rates per gigawatt are improving, not worsening. But the fire-code and insurance picture is now an underwriting item, not a footnote, and buyers should treat it that way in diligence. Ask which edition of NFPA 855 the jurisdiction has actually adopted, since enforcement lags the standard by a year or more in many places. Ask whether the sprinkler and suppression design was validated against UL 9540A large-scale test data specific to the cell chemistry and rack configuration being installed, not a generic listing. Ask the insurer, not just the EPC, what separation distances and construction ratings they’re underwriting to — the EPA’s recommended 330-foot isolation zone for commercial BESS is a starting point, not a guarantee of the number your carrier will actually require. And get the cell-level defect and traceability data from the battery supplier in writing before you assume manufacturing risk sits entirely with your integrator.
The facilities getting financed cleanly in 2026 are the ones where the fire-protection engineering, the insurance underwriting, and the code compliance all point to the same design — not three separate conversations that only meet for the first time at a claims review. Start that conversation with the Data Center Buyer’s Toolkit before your next site walk.
Sources
Written by
Raajeev Ratra
Data Center Infrastructure Expert | 21+ Years in DC Design, Operations & Project Management
Raajeev is a seasoned data center professional with hands-on experience in hyperscale facilities, colocation design, power & cooling infrastructure, and global DC operations. He shares practical insights to help engineers and IT leaders build better infrastructure.