The Power Shift: How On-Site Energy is Redefining Data Center Growth

Data center operators want to focus on their core business, building and running efficient facilities, not managing power plants. But as energy demand grows, the power source has shifted from grid with intermittent backup diesel to Natural Gas Prime Movers as primary generation that must operate for years before grid connections are available.

Third-party ownership models for power generation assets are gaining wider acceptance. These structures, long used in energy markets, offer stable returns for investors while allowing data center developers to access resilient power without taking on the operational burden.

Utilities and gas suppliers increasingly require proof that projects are real—letters of credit or secured financing—before committing infrastructure. In many major U.S. markets, interconnection queues now stretch five to eight years, some even longer.

That’s driving a surge in on-site or “bridge” generation: temporary or modular power solutions that allow facilities to energize and operate years ahead of grid access or even forgo the grid all together.

If you’re running so-called bridge power for five years, that’s not bridge, it’s prime power. Many operators install behind-the-meter plants intending them as interim solutions, but they soon realize the economics justify keeping them permanently.

With utility prices rising, especially in constrained regions like PJM, on-site generation often becomes the lower-cost, long-term source.

We’re now seeing the data center asset and the power asset financed separately but integrated operationally.

When the power side is developed under an Energy Services Agreement (ESA) or EaaS structure, the responsibility shifts. Third-party specialists own, operate, and amortize generation, switchgear, and distribution assets over 10–15 years.

That structure essentially turns energy into a managed service for the data center.

Speed to market is everything. Our analyses show that earlier energization can be worth $0.10–$0.40 per kWh in added value to users. For AI and GPU hosting clients, activating even a year sooner can translate to billions in incremental revenue.

That’s why modular and on-site generation is no longer a niche, it’s a competitive advantage.

That is correct, there are some well qualified parties pursuing this approach. The reason it’s tricky is due primarily to the idle period associated with the repower. If a combustion turbine has been sitting for a long period and has not been well maintained, it could have damage that would incur substantial cost to overcome.

Additionally, modernizing the equipment to meet current codes and standards would take a significant engineering effort and expense. Most existing fossil plants never considered redundancy, so that cost also has to be factored into the equation. That said, existing facilities have permitting, rights of way, and infrastructure in place that can be leveraged to speed time to market and reduce other costs, so the entire asset mix needs to be evaluated on the whole.

In some instances, it can be a very viable approach. In others, it could be an integration nightmare. I guess it remains to be seen.

As of now, no fossil-fuel-based system can qualify for an Investment Tax Credit (ITC) unless construction began before January 1, 2025. Some equipment manufacturers anticipated this rule change and pre-purchased equipment for projects still under commercial negotiation. They are leveraging this stockpile of ITC-eligible assets throughout 2025, but those credits are expected to be fully utilized by January 1, 2026.

There is significant discussion in Washington about advancing a Production Tax Credit (PTC) under the new technology-neutral framework currently being reviewed. To qualify, generation assets may need to incorporate heat recovery in a Combined Heat and Power (CHP) configuration.

Absorption chillers, which convert waste heat into cooling, have not yet been widely adopted in the data center community, largely because operators are unfamiliar with the technology. But when applied at scale within CHP systems, the advantage is substantial.

Under current performance metrics, CHP with absorption cooling can reduce electric consumption by more than 1 kW per ton of cooling. For a typical 100 MW data center, this efficiency gain can free up an additional 20–30 MW for IT load rather than cooling load. That represents a major shift in overall consumption dynamics and could dramatically enhance the ROI of on-site generation, especially if new PTC rules are enacted.

Today’s hyperscale data centers often include hundreds of millions of dollars in generation assets.

You’re building not only compute capacity but utility-scale energy infrastructure, with all the associated permitting, maintenance, and tax implications. Operators need to think like energy companies: plan for redundancy, regulatory exposure, and long-term O&M costs.

We’re seeing the same modular thinking that reshaped data center design, now being applied to power.

Instead of a single 50 MW turbine, developers deploy clusters of 5–10 MW engines grouped into “power blocks.” This modularity allows phased build-outs—10 MW online quickly, then additional blocks as demand grows. It also improves redundancy and cuts time-to-revenue.

The old model, grid as primary, on-site as backup, is flipping.

In many markets, it’s now cheaper and faster to run on-site generation as your main source, using the grid for redundancy. Pair that with waste-heat recovery for chilled water or absorption cooling, and your power usage effectiveness (PUE) can fall below 1.0. It’s efficient, scalable, and resilient.

Energy, cooling, and digital infrastructure are converging into a single ecosystem. The future belongs to developers who can integrate scalable, modular, power-integrated platforms that evolve with regulation and technology.

Building flexible, behind-the-meter assets is how you de-risk investment, control cost, and stay ahead of the curve.