Backup power systems have long been essential for data centre operators striving to meet uptime targets. As demand for AI grows, grid infrastructure ages, and outages become more likely, the reliability of backup systems is increasingly seen as a strategic priority rather than an operational necessity. Industry research published by the Uptime Institute – “Five data centre predictions for 2026” – suggests that power failures will be more frequent, more complex, and more significant in 2026 and beyond. With that in mind, the old model of “test the generator once a year and hope for the best” is no longer good enough.
Grid instability is a growing risk
According to Uptime Institute’s report, “Uptime Intelligence expects resiliency to re-emerge as a major area of focus and concern for the industry.”
Ageing transmission infrastructure, electrification, intermittent generation, and the explosive growth of AI computing are putting increasing strain on power grids. In September 2025, the US Department of Energy (DoE) warned that grid power outages could increase one hundredfold by 2030. The US is not alone. In April 2025, Spain and Portugal were hit by one of the largest blackouts recorded in Europe in decades. In that same year, hundreds of smaller blackouts were caused by a myriad of failures and extreme weather events, suggesting that Europe’s electricity system is also at risk.
With sources pointing to more frequent disturbances, voltage fluctuations, and long-term outages, data centres that once expected rare grid failures must now plan for more regular outages.
Most data centres overcome this risk by installing backup generators. However, with frequent grid fluctuations and outages putting greater stress on critical power systems, ensuring the reliability of the generator is critical. This changes the role of testing. As grid risk rises, load testing becomes a frontline defence, not a compliance checkbox.
From “maintenance asset” to strategic resilience tool
While large outages are becoming less likely relative to industry growth, Uptime Intelligence data shows that, when failures do occur, their consequences are more severe. This shift necessitates a new approach to generator testing.
While using a load bank to test generators has always been critical in terms of maintenance – testing alternator housing, windings, bearings, controls and cooling systems, and avoiding wet stacking – the extra pressures of grid instability and the rise in high-density infrastructure mean that operators must go beyond maintenance.
Where many operators may previously have relied on testing once or twice a year, testing should now include scenario-based testing, including black starts, ATS transfers and N+1 failures. When backup systems consist of several generators, testing is also a vital step in validating synchronised or parallel systems, including verifying that they start automatically, synchronise, and support failover mechanisms. Tests should also assess how generators would perform in the event of a partial power failure.
Proper loadbank testing can provide data that can directly influence board level decisions around resilience, investment, and risk planning. When asked, “How confident are we in our backup power?” Load bank results become the only credible answer.
On-site power growth
As grid capacity tightens, and fears of systemic failure mount, more data centres are investing in prime power systems, gas turbines, hybrid generation and shared power infrastructure across campuses. While becoming independent of the grid and adding redundancy may reduce reliance on large backup systems, new generators and power systems also require careful validation.
Full system testing should be undertaken during commissioning, to confirm that the system – whether diesel generators, gas turbines or other power sources – performs as expected under real operating conditions. This includes validating load sharing and synchronisation in multi-generator or hybrid generation systems, as well as checking system settings and components to ensure safe transitions and fault response. Equally, thermal and cooling systems should be tested to confirm they can handle operations without overheating. Additionally, fuel systems must perform flawlessly to deliver clean and polished fuel for hours upon end, from large fuel storage systems across a range of ambient temperature throughout the year.
None of that can be done meaningfully without load banks. In a world of multi-megawatt data centres and off-grid power generation, skipping load testing is effectively commissioning in the dark.
Compliance moves centre stage
As data centres create more energy on-site, prime power, built-in redundancy and battery storage mean that some facilities are actively supplying the grid through demand response arrangements. This move from data centres as passive consumers to “grid interactive” resources are likely to bring a new level of compliance. Many grid-interactive sites are being asked to meet new grid connection rules, including performing and documenting commissioning tests, as well as regular testing, to ensure safe grid-parallel operations and reduce risks.
As compliance requirements tighten, load bank testing is likely to become a regulatory enabler, demonstrating proven reliability.
The need for full load
Of course, while new pressures abound, we can’t overlook the age-old issue of wet stacking in backup generators. Wet stacking usually occurs when engines operate significantly below their rated output level, which is common in backup scenarios. This excessive no-or-low-load runtime causes a buildup of unburned fuel deposits around the combustion chamber, injector nozzles, piston rings, turbocharger, and exhaust. This, combined with condensed water, results in carbon particles and moisture gathering around the exhaust system. Potential outcomes include reduced power, increased emissions, and higher operational costs.
Regularly running the generator at full load during load bank testing can resolve these issues. If the generator is used below the rated output level on a regular basis, pairing it with a load bank more permanently will prevent “wet stacking” from happening. Without load banks, generator failures often appear for the first time during a real outage, when it’s already too late.
The bottom line: Load banks are still mission-critical
As power resilience remains a real risk for data centres, load bank testing remains the backbone of a credible uptime strategy. As well as their ongoing role in testing whether backup power is “outage ready” and preventing equipment damage, load banks are increasingly important in supporting tiered resilience strategies and de-risking on-site power plants.
In a world of unstable grids, AI-power consumption, and high uptime targets, backup power that hasn’t been tested using a load bank isn’t really backup power at all.
