This article was originally published in the 2026 Energy Storage Report.
Interconnection queues stretching five to seven years are making grid capacity tight—fast. Add AI’s skyrocketing power demand and it’s no surprise data centre developers are scrambling for on-site power. Microgrids offer a pragmatic path forward: start in islanded mode, stay resilient through disturbances, and later connect to the grid without re-architecting the power system. But building a microgrid that can withstand AI-driven volatility while targeting five-nine (99.999%) availability—roughly five minutes of downtime a year—is where software determines outcomes.
As grid codes grow more complex and projects scale to multi-GWh storage, the stakes for stability are too high for slow, piecemeal controls. For years, GEMS software has been the choice of first-to-market movers because it delivers the speed, specificity, and site-wide coordination complex markets demand. These capabilities are no longer optional; they are the foundation of AI-era data centres that require uncompromising reliability and uptime.
At the heart of an AI data centre microgrid is software that moves seamlessly between grid-forming and grid-following modes. In practice, this means enabling blackstart capabilities to bring critical loads online during a blackout, maintaining strict frequency and voltage control to protect asset health, and orchestrating engines, batteries, renewables, UPSs, and balance-of-plant under one unified system. At millisecond timescales, every extra connection adds latency and risk. That’s why complex or multi-GWh-scale sites depend on a single control platform capable of deterministic, site-wide response.
AI training workloads are uniquely volatile, making rapid, large step changes in their load profiles that do not align with thermal generator operating patterns. Pairing engines, battery storage, and intelligent controls solves this mismatch: energy storage absorbs load spikes in milliseconds, stabilising the engine-side load and maintaining site power quality. Meeting these fast-response requirements depends on seamless end-to-end control: coordinated assets, precise measurements at the point of interconnection, and reserve management in the BESS enabled by accurate state-of-charge measurement.
Intelligent full-site command software doesn’t just keep engines and UPSs from tripping or overheating; it optimises the microgrid for efficiency, cost, and sustainability. GEMS Dispatch balances renewables, engines, and BESS to reduce fuel utilisation and maximise renewable output, all while preserving headroom for contingencies. Crucially, it keeps the site grid-friendly. With the right controls, data centre loads can maintain low-voltage ride through, provide fast frequency response, and participate in ancillary-services markets—without sacrificing on-site power quality.
When speed-to-power and near-100% availability are the end game, microgrid control is the answer. Data centres need power that is stable, predictable, and immediately available. GEMS-enabled microgrids deliver exactly that: absorbing AI volatility, protecting critical assets, and orchestrating every asset to provide flexible, always-on power with confidence.
Once every six weeks, you will get the top picks – the latest and the greatest pieces – from this Insights channel by email.