The expansion of data centres across Southeast Asia is often discussed in terms of digital ambition and economic growth. Less openly acknowledged is the pressure this growth places on energy systems and the quiet risks that come with it.
How Energy Sustainability is a Core Risk for Data Centres in the AI Era
Artificial intelligence is changing the nature of data centre demand. AI workloads require sustained, high-density power for extended periods, with minimal tolerance for interruptions, particularly in high-density AI training and inference environments where compute loads are continuous and power quality is critical.
This is not simply a matter of higher electricity consumption. It is a shift in how power must be delivered consistently, predictably, and under increasingly volatile conditions.
For many years, the industry has relied on a familiar model: grid power supported by diesel generators for emergencies. That approach is now being tested from multiple directions. Power demand is rising faster than grid reinforcement in several
regional markets, including parts of Southeast Asia where data centre capacity is expanding more quickly than transmission and generation upgrades. Renewable energy, while essential for decarbonisation, introduces variability that legacy systems were not designed to manage. At the same time, carbon constraints and operational scrutiny are making diesel-heavy backup strategies harder to justify over the medium to long term, as regulatory expectations, investor pressure, and operating costs continue to evolve.
In this environment, the limitations of conventional energy storage become clear. Short-duration batteries and uninterruptible power supplies are effective for brief disturbances. They are not designed to sustain operations through prolonged grid instability, renewable intermittency, or extreme weather events, which are occurring with increasing frequency and severity in many regions.
Duration is becoming the defining factor in energy resilience
This is why long-duration energy storage (formally referred to as Long Duration Energy Storage Systems, or LDESS) is moving rapidly from the margins into the core of data centre energy planning.
LDESS are designed to store and discharge energy over extended periods, often ten hours or more. The specific discharge duration depends on system design and application, and can vary across different long-duration storage approaches, reflecting the fact that LDESS is an umbrella term rather than a single technology. That duration matters. It allows data centres to absorb variability rather than react to it, to integrate renewable energy more deeply without compromising uptime, and to reduce dependence on diesel generation without increasing operational risk.
From my experience working with mission-critical power environments, one lesson stands out: reliability failures rarely occur at peak load. They occur when systems are asked to perform under stress for longer than they were designed to endure.
Energy resilience, in practice, is less about capacity than it is about duration.
Best practices
Treat energy storage as core infrastructure, not backup
The most resilient operators are no longer treating storage as an emergency add-on. They are planning for duration from the outset, modelling outage scenarios that last hours or days rather than minutes. They are integrating storage early into site design and capacity planning, instead of retrofitting solutions after vulnerabilities become apparent.
Build resilience through layered energy systems
Equally important is the move towards layered energy strategies. No single system does everything well. Grid supply, renewables, short-duration batteries, long-duration storage, and where necessary transitional backup generation each serve distinct roles. Together, they reduce single points of failure and create flexibility when conditions change.
Control and visibility are as important as capacity
Operational intelligence also matters. Storage delivers its full value only when paired with monitoring and control systems that can forecast demand, manage charge and discharge cycles, and respond dynamically to grid conditions. Hardware alone does not create resilience; how it is managed does.
Build trust through rigorous safety and regulatory discipline
Finally, long-duration storage needs to be approached with the same rigour as any other core infrastructure. Safety standards, regulatory engagement, and site-specific risk assessments are not procedural hurdles. They are part of building systems that operators can trust under pressure.
Resilience will define the next era of data centre performance
As AI continues to scale, energy continuity is becoming a strategic concern rather than a purely technical one, with direct implications for long-term operational resilience, risk management, and capital planning. Data centres will increasingly be judged not just on performance and security, but on their ability to operate reliably and responsibly as demand grows more complex.
Long-duration energy storage is not simply a decarbonisation tool. It is a resilience asset — one that helps ensure the digital infrastructure underpinning modern economies can perform as expected, even when conditions are far from ideal.
For data centres in the AI era, the question is no longer whether long-duration energy storage will be necessary. It is whether it is integrated deliberately and early or added later, under far less forgiving circumstances.
