Bergen Engines Supplies 500MW for AI Data Centers

AI data center expansion is increasing pressure on electrical infrastructure, particularly where high-density compute clusters create rapid, unpredictable load changes that conventional grid connections may struggle to accommodate. In this context, Bergen Engines secured an order from Liberty Energy for more than 500 MW of on-site power generation capacity for large-scale U.S. AI data center projects.

Distributed power architecture for AI compute loads
The agreement centers on Liberty Energy’s effort to build dedicated power infrastructure for AI-era data centers through its Liberty Power Innovations (LPI) division, which focuses on distributed power deployment. The installations are designed to operate either fully off-grid in island mode or in parallel with the grid, depending on site requirements.

This reflects a broader infrastructure shift in the data center sector, where developers are increasingly evaluating on-site generation to avoid transmission bottlenecks, shorten deployment timelines, and improve operational resilience. AI workloads differ from conventional enterprise data center demand because compute clusters can generate abrupt power fluctuations at sub-second timescales, creating challenges for generation stability and power quality.

Engine configuration and installed capacity
Under the contract, Bergen Engines will supply 45 B36:45V20 AG1 gas generator sets, each rated at 11.2 MWe, for a combined installed capacity exceeding 500 MW.

The generator systems use medium-speed reciprocating gas engines, a configuration commonly selected for large-scale continuous-duty industrial power generation due to efficiency and durability under baseload operation. In this deployment, the engines will provide primary generation capacity capable of sustaining independent operation without grid support.

Liberty’s broader infrastructure package also incorporates its Forte modular power generation architecture and Tempo power quality system.

Managing fast AI load transients without oversizing generation
The power architecture combines Bergen Engines’ generation systems with Marelli Motori alternators and Piller Power Systems’ SHIELDX™ Dynamic Power Stabilization technology.

A key engineering challenge in AI infrastructure is handling rapid load step changes without forcing operators to oversize generation assets. According to the technical specifications provided, SHIELDX™ uses a flywheel-based stabilization architecture that absorbs sub-second power fluctuations of up to 30%.

This buffering function reduces transient stress on generator assets while maintaining conditioned power delivery to sensitive IT equipment. By decoupling abrupt load swings from the generation system, the architecture is intended to improve plant efficiency and reduce fuel consumption, emissions, and physical footprint compared with systems sized purely around transient peak demand.

Piller stated the stabilization system is designed for operating lifecycles exceeding 20 years.

Power infrastructure beyond grid dependency
The deployment illustrates how the digital infrastructure sector is increasingly adopting dedicated power systems rather than relying exclusively on utility interconnection timelines.

For hyperscale and AI computing environments, power quality has become as important as capacity. Voltage instability or transient disturbances can affect sensitive computing hardware,

Edited by Aishwarya Mambet, Induportals Editor, with AI assistance.

www.bergenengines.com