AI-ERA DATA CENTERS IN SOUTHEAST ASIA – PART 4 OF 6

Battery Strategy as an Asset and Infrastructure Decision

James Soh

The occupancy permit is delayed. The Authority Having Jurisdiction (the AHJ, the body with legal authority to inspect, approve, and certify that a building is safe to occupy) has flagged issues during inspection that should have been caught earlier. The battery capacity was changed during the project. The drawings and specifications were updated. But the documentation submitted to the AHJ was not. The fixes are not technically complex. But each one takes time. And a rescheduled AHJ inspection does not happen the following week.

The shift from in-rack to sidecar Battery Backup Units (BBU) is already visible across current and next-generation NVIDIA platforms. In the GB200 and GB300 NVL72, the BBU is integrated into the rack itself. In the Vera Rubin NVL72, rated at approximately 130kW, in-rack BBU trays remain viable.

However, in the Rubin Ultra, where power density reaches 600kW to 1MW per rack, physics forces the BBU into a dedicated sidecar cabinet adjacent to the compute rack. That sidecar is not a minor infrastructure detail. It is an energy storage installation inside the data hall, at a scale and density that most regional fire authorities have not previously evaluated. Every Rubin Ultra deployment brings its own zoning, fuel load, and suppression questions directly into the hall. The AHJ reviewing that configuration is not reviewing a centralised battery room. They are reviewing something new, at a density that existing approvals may not yet cover.

That scenario is frustrating and costly. It is also the manageable version.

The version that keeps project directors awake is the data hall design that was changed to accommodate sidecar battery racks, and where the battery chemistry selected sits in a high-risk classification rather than a moderate-risk one. That is no longer a documentation exercise. An AHJ reviewing a high-risk chemistry in a configuration they have not previously approved, in a jurisdiction where ESS guidance is still evolving, can decline to certify. Not delay. Decline. The capex is committed. The client timeline is running. And the decision that produced this outcome was made in an engineering workstream, at a level where its regulatory and commercial consequences were never visible.

Battery strategy is a governance problem. The technical complexity is real and it requires senior executive or even board-level attention given that batteries are now required to be in close proximity with the GPU servers.

———

The decision that ended up causing major headache

Battery systems used to belong in MEP specifications. They surfaced at the capex line and when something went wrong. In AI-era data centers they sit at the centre of the investment decision, because AI clients read your battery architecture the same way they read your rack density and your cooling topology. Chemistry, placement, zoning, and in-hall flexibility determine which halls can host which workloads, on what timelines, and under what regulatory and insurance conditions.

A build-to-suit project lets you co-design around a specific client. A speculative build locks in battery and ESS decisions long before you know which workloads will land in which halls. In both cases the choices made at concept design, often delegated below board level, determine whether the facility delivers on the investment thesis or becomes an asset that spends its life being worked around.

Boards and senior leaders do not need to become battery engineers. They need to know enough to ask the right questions and to recognise when a battery decision has escalated beyond its workstream. The AHJ does not distinguish between a decision made at board level and one made in an engineering meeting. The consequences land at the same address.

———

What battery strategy actually covers

At the chemistry level, most sites still rely on Valve-Regulated Lead-Acid (VRLA) as the workhorse of centralised UPS rooms. Lithium Iron Phosphate (LFP) is increasingly the choice for new deployments, with better energy density, longer cycle life, and a more manageable thermal-runaway profile, though it brings fire-behaviour characteristics that authorities and insurers are still calibrating to. As explained in Data Center Primer, UPS batteries exist to bridge outages, not provide long-duration storage; most designs target 5–15 minutes of ride-through, enough to bring generators online, with significant implications for room sizing and placement.

At the architecture level, three patterns are now in play across Southeast Asia's AI campuses:

• Centralised UPS and battery rooms, which remain the baseline, familiar to authorities and insurers, straightforward to govern.

• In-hall UPS and cabinetised ESS, which move storage closer to the AI load but add zoning, fuel-load, and suppression complexity.

• In-rack or sidecar systems, which typically arrive with the tenant, sometimes ahead of the operator’s readiness to govern them safely.

The gap between what tenants bring and what operators are ready to govern is where much of the current risk sits, and where AHJ surprises are most likely to originate.

———

Preventive: the decisions that cannot be undone

The most consequential battery decisions are made before a single battery is installed, at concept design, when structural grids, floor loadings, fire zones and room allocations are still fluid. Which halls are permitted to host in-hall ESS and at what energy density. What separation distances and fire ratings apply. Which chemistries are in scope and whether that position is defensible to local authorities and insurers. What the worst credible event is in each zone.

These are not engineering details. They define the facility’s risk envelope and the approval conversations the organisation will be having for the next decade. In Data Center Primer the Owner’s Project Requirements process forces these questions into the open at around the 30-percent design stage, treating battery room allocation and zoning as investment-level decisions rather than late-stage MEP detailing. A chemistry or layout change made after that stage, without updating the AHJ submission, is how manageable problems become permit delays.

Three questions a board should ask before approving any ESS layout:

1. What is the worst credible event in each battery zone, and does our design contain it?

2. Have local authorities and insurers seen this layout, and do we have their preliminary position in writing?

3. Can our operations team realistically manage this configuration under normal conditions and under stress?

If those answers are not clear, the layout is not ready for approval.

———

Predictive and proactive: stewardship as governance

Once batteries are in service, the Battery Management System (BMS) becomes an asset-management instrument, not just a vendor monitor. Cell temperatures, voltage imbalance, state-of-health trends, and event history are the data that tell you whether a room remains an asset or has quietly become a liability.

Boards can ask one simple question: when did operations and asset management last sit together to review battery health data across the portfolio? If the answer is only when something went wrong, the organisation is managing reactively. A good battery health dashboard tells a CFO or investment committee, in minutes, which rooms are healthy, which are drifting, and what the replacement horizon looks like for each market and campus.

Proactive withdrawal, meaning removing batteries based on trends and defined thresholds rather than waiting for end-of-life or failure, is the clearest signal that an organisation is managing ESS as an asset. It is also how operators avoid the scenario where a marginal battery room becomes the reason an AI client’s hall cannot be recertified or expanded.

———

Commercial consequence and the shift test

Battery architecture is a commercial decision as much as a technical one. A facility built solely around centralised rooms cannot easily accommodate AI tenants who arrive expecting in-hall or sidecar ESS; the retrofit is expensive, slow, and requires a fresh AHJ conversation. A facility with selected halls pre-positioned for in-hall ESS can serve both tenant types at different price points with clearly defined risk parameters. Knowing which halls are AI-ESS capable and which are centralised-only, and reflecting that honestly in SLAs and pricing, is more credible to sophisticated clients than any generic AI-ready claim.

For operations teams, the governance test is direct: if the battery strategy cannot be explained in a ten-minute shift handover, it is too complex to govern safely from the boardroom either. The people who will respond to an ESS alarm at three in the morning are the same people who will determine whether a battery event stays contained or becomes a permit, insurance, and client conversation.

———

Regulatory and insurer reality check

Regulators and insurers are tightening expectations faster than many investment theses in the region have anticipated. The lithium-ion battery fire at a data center campus in Loyang, Singapore, which affected multiple battery and power rooms and disrupted major cloud services, focused regional fire-safety authorities on ESS placement and suppression in ways that are still working their way into updated codes. A battery-room fire in Blackwall, London in early 2026 drew the same conclusions in a different market. Updated requirements now include thicker compartment walls, tighter energy limits per room, and in some jurisdictions water-based suppression systems that raise structural and drainage costs significantly.

Authorities are stakeholders in battery decisions, not just gatekeepers of them. The governance gap that creates board surprises is the same gap that produces drawn-out AHJ approval conversations; developers who arrive at inspection with battery layouts that were changed after the original submission, or with chemistries the authority has not previously evaluated, create delays that fall on everyone. Early, transparent engagement with the AHJ, before the 30-percent design stage, is how operators avoid the rescheduled inspection scenario that opened this article.

By market, the picture is differentiated. Singapore and Johor face tightening ESS guidance, and cross-border projects must satisfy two regulatory regimes with different views on placement and protection. Thailand’s Eastern Economic Corridor authorities are encountering ESS configurations with limited prior exposure, making early engagement critical. Jakarta and Batam are seeing build-out outpace regulatory familiarity, while insurers increasingly expect standards that go beyond local code minimums. In Vietnam and the Philippines, early operators are effectively setting ESS precedent in markets where frameworks are still forming.

———

Five questions worth asking your teams this week

If battery strategy has not been on a board or senior leadership agenda recently, these are the questions that will tell you how much work remains:

Design and policy: “For each campus, which halls are permitted to host in-hall or cabinetised ESS, to what energy density, and what is the worst credible event in each battery zone, and has the AHJ seen that answer?”

Telemetry and health: “How often do we review battery health data at portfolio level, who is in the room, and what decisions changed in the last year because of what we saw?”

Withdrawal and replacement: “Do we have a proactive withdrawal policy, and when did we last remove batteries based on it rather than waiting for failure?”

Tenant mix and hall strategy: “Which halls can credibly support in-hall or sidecar ESS for AI tenants today, and how is that reflected in SLAs and pricing?”

Regulatory and insurer fit: “When did we last review our ESS configurations with local authorities and insurers, and did that review lead to any design or operational changes?”

If those questions cannot be answered clearly and consistently across your portfolio, the organisation has more work to do, regardless of how sophisticated the hardware appears.

———

Looking ahead to Part 5

Across Parts 1 through 4, fire engineering has been in the background of every discussion about chemistries, layouts and ESS decisions. In Part 5, it moves to the front. The Fire Marshal in the Boardroom will look at how power density, ESS chemistries and layouts translate into fire-engineering limits, how to work with regulators and insurers before those limits become surprises, and what that means for project scope, capacity decisions and the operations teams who have to live with the result.