We're not here to sell you a battery. We're here to help with whatever spike is driving you to this page. A DC fast charger on a service the utility won't upgrade this year? A generator that takes 15 to 30 seconds to reach full power while your data center rides through? A crusher, chiller bank or GPU rack whose load swings faster than the supply can follow? A fire marshal asking harder questions about your battery room? Tell us which one it is, and we'll tell you honestly whether a non-flammable VSB battery buffer is part of the answer, not assume it is before we've heard your specifics.
In the rack: AI power demand is outrunning legacy UPS sizing assumptions. Traditional data-center UPS sizing was built around a steady IT load with a slow, predictable ramp. AI training and inference workloads break that assumption: Uptime Institute describes AI workloads producing power-demand spikes and drops "every second or two," with load doubling within milliseconds in worst cases. GPU racks now draw 30-100 kW versus 8-15 kW for a conventional server rack, and simultaneous spikes across multiple servers can exceed the rated capacity of row-level UPS modules sized under legacy assumptions.
In the battery room: five separate lithium-ion battery-room fires at data centers and telecom-adjacent facilities since 2022 (SK C&C/Kakao in South Korea, 2022; Maxnod in France, 2023; a Digital Realty facility in Hillsboro, Oregon, 2025; the NIRS facility in Daejeon, South Korea, 2025, with 858 TB of government data permanently lost; and a Microsoft facility in Goodyear, Arizona, reported 2026) share the same failure mode: a flammable liquid electrolyte in thermal runaway.
At the grid: the incident record at utility scale is cumulative, not isolated. In January 2025, a fire at Vistra's Moss Landing facility, at the time the world's largest BESS, evacuated roughly 1,200-1,500 nearby residents, closed a stretch of Highway 1, and led to a lawsuit, an EPA enforcement settlement, and new California legislation (SB 283) requiring fire-authority engagement and a pre-energization inspection for BESS developers. South Korea recorded roughly 28 documented BESS fires between 2017 and 2019 that froze that country's storage market for over a year. NFPA 855's entire setback and separation regime, and UL 9540A's large-scale fire test, exist because lithium-ion can propagate a thermal-runaway fire from one unit to the next.
Sources: Uptime Institute Journal · DatacenterDynamics · NFPA Journal, Feb. 2026 · Energy-Storage.News · US EPA · Solar Power World (SB 283) · Energy-Storage.News (South Korea)
Most storage is bought like a bucket: how much energy fits in the box. A battery buffer works like a conveyor belt: how fast power moves through it. That difference decides how much battery you actually have to buy.
Under sustained load, lithium's limit is heat. It rises with the square of the current, so packs are held to a continuous rating, commonly about 2C, and a current 20 ft LFP block pairs 4 MW of inverter with 8.1 MWh of cells. To cover a large spike you buy parallel cells so each one stays cool, and most of that capacity sits idle. VSB cell-modules are rated far above lithium's C-rate, so it's sized to the spike itself. The shorter the spike, the bigger the difference.
Capacity to buy. Lithium held to 2C continuous; VSB sized to the event's energy or a 20C planning ceiling (a BF planning assumption, not a manufacturer rating), whichever is larger. Before depth-of-discharge and end-of-life margins. Illustrative, not a site quote. LFP block: ess-news, May 2025.
It removes a documented fire-risk category: in the rack, the room, or the yard. VSB's solid electrolyte doesn't have the thermal-runaway failure mode lithium-ion does, independent of how well any individual facility's suppression or compartmentalization performs. Every setback distance in NFPA 855 and every large-scale fire test in UL 9540A exists because lithium-ion can catch fire and propagate to the next unit. VSB can't propagate that fire because there's no liquid electrolyte to fail in the first place. That's a structural difference in siting and insurance exposure, not a suppression or monitoring improvement layered on top of the same risk.
It removes the dedicated cooling plant, not the facility's own IT-load or process cooling. Lithium-ion batteries typically need a controlled 20-25°C environment (or active HVAC/liquid cooling at BESS scale) to hit their rated life. VSB's −40°C to +100°C operating range needs no active cooling at all, which removes that mechanical subsystem, its parasitic load, and its failure mode specifically from the battery room or storage site. This is a distinct, smaller-scope claim from "solving data-center cooling": the compute floor's chiller plant and PUE remain a separate problem VSB does not address. Response time for fast grid services (frequency regulation, sub-second arbitrage) is a fair, current lithium advantage VSB has not yet benchmarked at BESS scale. That gap is real, not glossed over.
It removes replacement and augmentation events, not just extends one battery's life. Lithium-ion UPS batteries typically run 8-10 years in service, meaning a facility with a 20-year design life goes through a full battery-room replacement roughly twice over that window. Utility and C&I storage contracts are typically written against a capacity-fade curve, with a mid-life augmentation cycle (new racks or cells added around year 8-10) budgeted into the financial model as standard practice. At 15,000+ cycles (a Vanadis projection) and a 20+ year design life, VSB is sized to run the same 20-year window on the original installation: fewer scheduling, decommissioning, and requalification events across the life of an asset whose whole value proposition is uptime.
None of this makes VSB automatically the right fit for your site; that depends on specifics we don't have yet. VSB is TRL 6-7: Vanadis Energy demonstrated a working system with Burgenland Energie in Austria (Dec 2025) and is preparing refinery pilots in Saudi Arabia. It is not yet certified (UL 1973, UL 9540A and IEC 62619 are planned), bankable, or field-referenced at a US site. This is a first-pass positioning page, not a sized deployment or case study. Tell us what's actually going on at your site and we'll listen before we pitch anything back.
The pressure point looks different depending on where you sit: a greenfield AI campus racing to maximize density, a microgrid asset that has to survive outdoors with no staff nearby, or a charging site whose utility service can't carry the peak. All three converge on the same question: how much floor space, service capacity and mechanical overhead does the battery itself demand?
Every square foot a battery room and its HVAC/suppression gear occupies is a square foot not filled with revenue-generating racks. Removing that mechanical footprint is a design-freedom question as much as a safety one.
Outdoor, all-weather, unstaffed duty means a cooling system that runs on the battery's own stored energy is a parasitic load on top of everything else. No active cooling required means more of what's stored actually reaches the load during an outage.
A DC fast charger can pull several times what a constrained service delivers. A buffer refills slowly from the grid between sessions and covers the peak, so the site can shrink or avoid the service upgrade. VSB cycles all day without a cooling plant, sized to the charger rather than to a thermal margin.
$/kWh rewards buying more capacity, which is exactly what an energy battery has to do to survive a spike. For a buffer, compare cost per kW delivered and cost per cycle over the rated life, and count what the installation needs around the battery.
Upfront: a non-flammable, no-active-cooling chemistry is a candidate to remove dedicated HVAC/chiller plant, chemical suppression, and the blast walls or thermal barriers a hazard-contained battery room requires, plus a shorter, less contested permitting path with the local fire authority, which is frequently the long pole in bringing new capacity online.
Ongoing: a battery that doesn't need active cooling isn't spending part of its own stored energy keeping itself cool, and a chemistry that removes the fire-risk category also removes the suppression and cooling-loop inspection contracts built around it. A 20+ year design life is sized to run past the point where lithium-ion UPS batteries, and utility/C&I augmentation cycles, are typically due.
These are directional framing points from general industry commentary on BESS/data-center siting economics, not an audited BF cost model. The real number for your site depends on your current battery room, your utility rate structure, and your local fire authority. Tell us your specifics and we'll work the actual math with you.
Ask every vendor, us included, for these under the same conditions: rate, depth of discharge, temperature and end-of-life threshold. Here's how the chemistries stack up.
| What matters at the site | Lithium-ion (NMC / LiPo) | LFP (LiFePO₄) | VSB (Vanadium Solid-State) |
|---|---|---|---|
| What sets the size | Continuous C-rate limit; bought well above the spike's energy | Continuous C-rate limit (blocks commonly ~0.5C); same oversizing | The spike itself: power and duration, inside a conservative 20C planning ceiling |
| Fire / thermal runaway | Propagating; needs suppression, vents, setbacks | Reduced but real: runaway above ~270 °C | Non-flammable chemistry; no thermal runaway or propagation (manufacturer claim, independent testing pending) |
| Cycle life (deep) | 500-2,000 | 3,000-6,000 | 15,000+ (Vanadis projection), over 90% capacity retained |
| Service life / augmentation | 5-8 years; frequent augmentation | 10-12 years; mid-life augmentation typical (yr 8-10) | 20+ years; not required by design |
| Fast-charge / ramp response | 0.5C-3C; mature grid-service benchmark | 0.5C-3C (EV-class cells); mature grid-service benchmark | Up to 60C at the cell-module; not yet benchmarked at BESS scale |
| Operating temperature | 0 to +45 °C (active cooling required) | −20 to +60 °C (still wants cooling) | −40 to +100 °C, no active cooling |
| Certification | Mature UL 9540 / UL 9540A path | Mature UL 9540 / UL 9540A path | Planned: UL 1973, UL 9540A, IEC 62619 |
| Freight / DG class | UN 3480 / 3481, Class 9; ≤30% SoC on aircraft | UN 3480 / 3481, Class 9; same DG burden | Classification pending with the manufacturer; no DG exemption claimed. |
| Recyclability | ~50% recoverable | ~50% recoverable | ≈85% of cell material recovered |
Specifications indicative; final ratings confirmed per project. VSB has not yet completed UL 9540A testing, NFPA 855 review or an independent-engineering bankability review for US sites.
Vanadis Energy builds 12 V cell-modules into 48 V packs of 1.2 kWh, each with its own BMS, and stacks the packs in cassettes inside an outdoor container. Vanadis rates the cell-modules well above what a container can use, so in a buffer the inverter and cabling set the container's power rating, not the cells. We plan at 20C, about 24 kW per pack.
On paper, yes. Fire code treats any battery as an energy storage system, and we'll permit it as one. The difference is what it's built and bought for: short, repeated power events rather than hours of stored energy. If your job is hours of backup, a lithium energy battery is probably the better buy, and we'll tell you so.
Because it's the wrong measure for a buffer. $/kWh rewards buying more capacity, which is exactly what an energy battery has to do to survive a spike. Compare cost per kW delivered and cost per cycle over the rated life, and count the chiller plant and suppression you may not need. Give us the spike and the service and we'll work both numbers.
Tell us the load, how long it lasts, how often it repeats and what the service or generator can carry. We'll size the buffer against it and show you the math, including where VSB isn't the right answer. You're not wrong to want a straight answer instead of a sales pitch.