POWER
By Arturo Di Filippi, global offering director for large power, Vertiv Advancing Uninterruptible Power Supply (UPS)
technology for AI-driven power demands H
igh-performance graphics processing unit (GPU) clusters that are central to AI operations, can swing from near-idle to full load capacity in milliseconds. This volatility strains every link in the power train, from energy storage units to grid connections. Engineers are now turning to innovative uninterruptible power supply (UPS) controls to address these challenges and provide longevity.
Protecting utility and on-site power- generating equipment
A key development is the Input Power Smoothing (IPS) feature for stabilising input power, which helps to shield upstream components from AI’s dynamic demands. By transforming the UPS into a dynamic buffer, it absorbs peak loads and releases energy as needed, resulting in a consistent draw from utilities or onsite generators. Fluctuations that could otherwise ripple through the system are contained, preventing stress on transformers, switchgear and distribution networks. This capability is increasingly important given the extreme variability of AI loads. IPS enables systems to handle rapid swings without propagating instability upstream, minimising thermal and electrical strain on infrastructure while improving generator performance by avoiding sudden ramps. Compliance with utility standards is
At its core, an intelligent algorithm monitors to operator-set limits, using stored energy strategically without unnecessary cycling. This allows the UPS to act as a stabilising interface between highly dynamic AI workloads and comparatively rigid power infrastructure.
Protecting energy storage designed to protect battery systems during abrupt shifts: Battery Shield. Established UPS designs often tap into batteries for quick energy bursts, leading to repeated shallow discharges that accelerate equipment wear. This new approach handles sudden dynamic loads by modulating the DC link voltage, storing just enough energy to bridge rapid transitions without unnecessary cycling.
10
The system can manage full-range power jumps from zero to 100 per cent load, without engaging the battery system. The advantages are multifaceted - critical backup power stays intact for real blackouts, extending battery durability and cutting down on costly replacements. Overall system reliability improves, and engineers gain more focused data centres.
Managing the state of charge Central to this functionality is state-of-charge (SOC) oversight. The system operates within smoothing duties with emergency readiness. During normal grid operation, it limits cycles to preserve longevity, and when generators kick in, it expands smoothing to prioritise input steadiness. When paired with batteries designed for highdynamic applications, this control strategy is even more effective, enabling faster response and more precise energy management. This adaptive strategy means that batteries aren’t depleted prematurely, enabling them to maintain their primary backup function even under prolonged volatility.
These innovations have undergone rigorous lab scrutiny using specialised AI load emulators. Scenarios included diverse power steps, cycle frequencies, battery setups and consistent - input currents remained steady amid wild output swings, validating the dual protection for batteries and external systems. This underscores a paradigm shift, where UPS units evolve from passive guardians to proactive managers in AI ecosystems.
APRIL 2026 | ELECTRONICS FOR ENGINEERS
The future of AI power protection Looking ahead, as AI applications intensify, mastering these dynamic patterns will help to of demand changes imposes new stresses across the power ecosystem. UPS systems, positioned at the heart of this interface, must adapt beyond their conventional clean- power delivery role.
surges can erode battery health through cumulative micro-cycles, clashing with their outage-protection mandate. Internal handling of transients via capacitors and energy storage for when its really needed. Conversely, in cases where volatility threatens upstream stability intentional battery involvement smooths the path. Battery selection is crucial. AI favours high-rate chemistries that handle swift charge-discharge without lag, enabling sustained buffering. Capacity sizing extends beyond outage autonomy to encompass smoothing duration, introducing fresh design metrics. Simulations reveal that ample storage prolongs stable operation under power train architecture.
Ultimately, AI’s expansion demands agile power strategies rooted in actual behaviours, not outdated norms. By integrating smart controls, high-performance batteries and tailored sizing, engineers can prepare to-deploy solutions empower data centres evolution.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50
