Executive Summary

• Commercial electricity demand is expected to overtake residential demand for the first time on record in 2026.
• As engine-driven microgrids become more widely deployed, operators must think beyond simply installing enough generating capacity.
• With this is mind, in this article, Oxford Flow’s Executive Vice President Shane McDaniel asks ‘where is all the power coming from?’

 

AI is accelerating investment in digital infrastructure, but its impact extends beyond electricity demand. As workloads fluctuate, they are also changing the engineering requirements placed on the fuel gas systems supporting engine-driven generation.

In June, Chevron and Microsoft announced plans to develop a dedicated 2.67 GW natural gas-powered facility to supply a Microsoft data centre in West Texas, underlining the scale of investment now flowing into new approaches to powering AI infrastructure. 

More recently, the US Energy Information Administration (EIA) forecast that US electricity demand will continue to reach record highs through 2027, driven largely by AI data centres and wider electrification.

Commercial electricity demand is even expected to overtake residential demand for the first time on record in 2026. Taken together, these developments point to a broader shift in how operators are thinking about reliable, scalable power.

For many, that means looking beyond traditional grid connections. Engine-driven microgrids and other forms of distributed generation are increasingly being deployed to provide flexible, resilient power where grid capacity is constrained or reliability requirements are particularly high.

Rather than relying on a single source of electricity, operators are building more diverse power architectures capable of responding to changing demand while maintaining the levels of uptime modern data centres require.

Securing additional generating capacity is one challenge, but ensuring those assets can respond reliably to fluctuating AI demand is another. 

Engine-driven microgrids move centre stage

Unlike large turbine-based power plants, gas-fired reciprocating engines can ramp up quickly as computing workloads fluctuate. That flexibility also creates frequent engine load transitions: as engines start up, slow down or change output, fuel demand changes with them. Gas pressure regulation systems must therefore respond to big swings in flow rate and inlet pressure while maintaining a stable and consistent pressure control point.

This trend extends well beyond individual projects. Last year, the US Department of Energy announced an $8 million investment in microgrid innovation to help strengthen the resilience of critical infrastructure. More recently, Energy Institute reported that the United States now uses more electricity to power data centres than any other country, illustrating the sheer scale of AI-driven demand. 

Whether the discussion centres on grid capacity, planning, resilience or dedicated generation, as engine-driven microgrids become more widely deployed, operators must think beyond simply installing enough generating capacity. They also need confidence that those assets will continue performing reliably as power demand changes.

Reliable power depends on more than the engine

When these AI workloads fluctuate throughout the day, the demands placed on engine-driven microgrids change with them. When computing demand increases, engines respond by generating more power.

As power demand changes, gas demand changes too. That creates operating conditions that differ significantly from many traditional industrial applications, where fuel demand is comparatively stable.

For engine-driven power systems to operate reliably, fuel pressure must remain stable at the engine inlet. Even relatively small variations can influence combustion behavior and generation performance.

In many cases, performance issues that appear to originate with the engine can actually stem from instability within the gas pressure regulation system supporting it. Reliable fuel delivery therefore becomes fundamental to reliable power generation.

Maintaining stable fuel pressure depends on more than steady-state accuracy. Transient response, pressure droop, repeatability and lock-up stability are all key engineering considerations as fuel demand changes.

Together, these characteristics allow generation assets to respond immediately to changing loads while supporting efficient combustion and consistent engine performance.

Engineering for continuous operation

Data centres differ from most industrial facilities because planned shutdowns are rarely an option. Generation systems are expected to operate reliably for extended periods while responding continuously to dynamic demand. Every element supporting those engines therefore needs to perform predictably, even as operating conditions change.

For pressure regulation systems, that means delivering rapid and stable pressure response while maintaining reliable performance over long operating periods.

Simplified regulator designs with fewer moving components also reduce potential failure points and help minimise maintenance requirements across distributed generation assets. Consistent performance under changing operating conditions also helps support a more reliable, lower-intervention operating model.

In worker-monitor arrangements, a compact architecture allows the regulators to be installed close together, reducing the dead pipe between them. For engine OEMs and skid manufacturers, this can reduce skid footprint, simplify piping, save weight across the overall package and lower engineering effort.

These may seem to be relatively small engineering considerations, but they become increasingly significant as AI infrastructure grows in scale.