Unison Energy's Game Plan for AI Data Center Power
Discover how Unison Energy is tackling power challenges in AI data centers and reshaping the future of energy solutions!
The grid wasn't built for this, and companies building AI infrastructure are finally accepting that reality.
Interconnection queues across the U.S. now stretch five to seven years in many regions. For a data center developer trying to bring a hyperscale AI campus online in 24 months, that math simply doesn't work. Something has to give — and increasingly, what's giving is the assumption that utility power will be there when you need it.
That's the gap Unison Energy is stepping into. The Connecticut-based Energy-as-a-Service developer, backed by Tiger Infrastructure Partners, builds, owns, and operates behind-the-meter generation systems — primarily natural gas combined heat and power (CHP) and microgrids — for large industrial and commercial energy users across North America. In January 2026, the company appointed Mariko McDonagh Meier as CEO, a veteran with roughly two decades of experience spanning utilities, distributed energy, and grid services. The timing is deliberate. Unison isn't pivoting to meet this moment — it's been building toward it.
Why the Power Problem Is Worse Than the Headlines Suggest
Everyone knows AI data centers are power-hungry. A single modern GPU cluster can consume 20–50 MW. A large AI campus easily clears 100–200 MW. The numbers are dramatic enough on their own.
But the deeper issue isn't consumption — it's timing. Grid operators and utilities are dealing with a queue of interconnection requests that collectively represent hundreds of gigawatts of new load. The process for approving, studying, and eventually connecting that load is measured in years, not months. Developers aren't just waiting for power — they're waiting in line behind thousands of other projects, with no guarantee they will reach the front.
This creates a structural problem that no amount of capital can easily solve. You can pre-order GPUs. You can secure land. You can sign hyperscale tenants. But if the substation upgrade that feeds your campus is three years behind schedule, your data center is a very expensive parking lot.
The consequences cascade quickly. Project timelines slip. Financing gets complicated when revenue timelines are uncertain. Tenants start looking at alternative sites. And in a market where AI compute demand is compressing deployment timelines, a two-year delay is essentially a competitive death sentence.
What Unison's Model Actually Does
Unison's core proposition is straightforward: don't wait for the grid. Build generation on-site, behind the meter, and deliver it as a service so the developer doesn't have to own and operate a power plant.
The CHP systems Unison deploys aren't just backup generators sitting idle in a parking lot. Combined heat and power captures the thermal energy produced during electricity generation and puts it to work — for cooling, heating, or other process loads — achieving total system efficiencies that can reach 70–80%, compared to roughly 33–40% for conventional grid power delivery. For a data center burning through power 24/7, that efficiency differential isn't marginal — it compounds into significant operating cost advantages over a multi-year contract.
The Energy-as-a-Service structure matters too. Instead of a developer spending $50–100 million of its own capital to build and staff a private power plant, Unison owns the asset and sells the output under a long-term contract. The developer gets dispatchable, on-site power without the balance sheet hit or the operational complexity. It's the same logic that drove data centers to adopt colocation and cloud — don't own infrastructure that's not your core competency.
Recent deployments illustrate the model at work. Unison's CHP project with General Mills in Missouri is exactly the kind of anchor reference that translates to AI data center conversations: a large, continuous industrial load with real efficiency and reliability requirements. The leap from food processing to data center isn't as wide as it sounds — both need dense, reliable, cost-predictable power delivered around the clock.
Interconnection Delays Are Redrawing the Rules
Here's the non-obvious angle most coverage misses: behind-the-meter on-site generation isn't just a workaround for interconnection delays — it's increasingly becoming the *primary* power strategy, not a backup hedge.
Developers who locked in grid-connected sites two or three years ago are watching interconnection timelines extend further as they wait. Meanwhile, developers who structured projects around on-site generation from the start are moving faster, with more predictable cost structures and fewer regulatory dependencies. The project that doesn't need a new substation doesn't wait for a new substation.
This is reshaping site selection in ways that will compound over the next decade. Locations that historically lost deals because of weak grid infrastructure are now viable candidates if they have access to natural gas supply and sufficient land for on-site generation. The traditional hierarchy — transmission capacity first, everything else second — is being renegotiated in real time.
It also changes the risk profile of development financing. Lenders financing a grid-dependent project have to underwrite interconnection risk: the possibility that the utility timeline slips, costs balloon, or capacity gets curtailed. An on-site generation project with a contracted fuel supply and an EaaS operator managing the system presents a fundamentally different risk picture — one that sophisticated infrastructure investors are beginning to price accordingly.
What Developers Should Actually Be Doing Now
For developers navigating this environment, the strategic calculus has shifted. A few observations worth taking seriously:
Don't treat on-site generation as a contingency plan. Developers who bake grid power in as the primary strategy and on-site as a backup are still operating on 2019 assumptions. Given current interconnection timelines, those roles may need to flip on many sites.
Gas-based microgrids and CHP systems require their own lead times — equipment procurement, permitting, fuel infrastructure — so the planning horizon needs to move up, not shrink. McDonagh Meier's appointment specifically signals that Unison is positioning for complex, large-scale deployments where EaaS relationships need to be structured early in the development process, not bolted on at the end.
Energy management at the campus level is also becoming a distinct competency. AI workloads are not uniform — training runs create massive demand spikes, while inference workloads are more predictable. On-site generation systems need to be designed with that variability in mind, and operators need contractual flexibility to manage it. The EaaS model, when structured correctly, can accommodate load variability in ways that static utility contracts cannot.
Where This Goes From Here
Natural gas CHP is today's answer, but it's not the permanent one. The same behind-the-meter infrastructure logic applies to hydrogen fuel cells, small modular reactors, and next-generation storage systems as those technologies mature and reach cost parity. The developers and EaaS operators who build the operational and contractual frameworks for on-site generation now are positioning themselves to swap in cleaner fuel sources as they become viable — without rebuilding the entire model.
The data center industry spent the last decade optimizing PUE and negotiating renewable energy credits. The next decade will be defined by who controls dispatchable power — power that's there when you need it, at the scale you need it, regardless of what's happening on the grid.
Unison is betting that "the grid will handle it" stops being a viable development strategy. Given what's in the interconnection queue right now, that bet looks increasingly well-timed.
Explore more about Energy-as-a-Service and how it can transform your data center operations.