Ohio's 350-MW Data Center Microgrid: Who Pays When the Grid-Independent Deal Goes Wrong?

A major project developer in New Albany, Ohio has tapped French environmental services company Veolia to operate and maintain a 350-megawatt microgrid designed to supply 100 percent of an AI data center campus's electricity without connection to the regional grid^[1]. The project integrates on-site gas engines, linear generators, and a 430-megawatt-hour battery energy storage system^[3], and Veolia frames it as a win for both the data center operator and surrounding communities: faster power access without waiting years for grid interconnection^[4], and reduced pressure on already strained transmission networks.

That framing is only half the story. What the announcements do not disclose is whether Ohio's utility regulator, the Public Utilities Commission of Ohio (PUCO), has extracted any protective terms in return for this private energy infrastructure escaping the normal interconnection and cost-allocation process. The silence is the mechanism worth interrogating.

Start with the contract's status. Is Veolia's operating agreement filed with PUCO as a special contract or a long-term service agreement? If filed, what is redacted from the public docket regarding capacity ratchets, minimum-take obligations, collateral, and term relative to the asset life of the microgrid? If not filed, why does a 350-MW facility generating its own power escape utility commission oversight entirely? The developer is undisclosed^[1], which means no one outside the negotiation knows whether the data center owner has agreed to take all the power produced, or whether Veolia's operation costs and risks are front-loaded into developer contracts with no clawback if the load underperforms. If the developer defaults or the AI workload shrinks, does Veolia bear the loss, or does the grid-independent framing become a cover for off-balance-sheet stranded assets that never surface in a rate case?

The deeper issue is what this deal says about grid availability in Ohio. The announcement touts it as a response to "lengthy interconnection delays"^[4], but does not ask: why does Ohio's interconnection queue have a seven-plus-year backlog? Duke Energy Ohio and American Electric Power (AEP) control roughly 70 percent of the state's load and both have been expanding their interconnection queues with data-center "studies" and "feasibility" requests that inflate expected load and justify capex in rate cases years before those loads sign firm contracts. A 350-MW off-grid project looks like a relief valve for those same utilities' inability or unwillingness to process grid connections efficiently. Veolia's microgrid is not reducing demand on constrained networks; it is providing an escape hatch for developers unable to navigate the utilities' own queue backlogs. The real relief for ratepayers would come from PUCO requiring the utilities to clear their interconnection queues within a set timeline or face penalties, not by outsourcing the problem to private microgrids.

The battery storage component warrants scrutiny. The 430-megawatt-hour system is sized to buffer the data center's intermittency, but Veolia's press materials do not specify the chemistry, round-trip efficiency, or whether the system will be allowed to provide grid services (peak shaving, frequency regulation) to AEP's grid when the data center is idle. If the battery is dedicated solely to the campus, it is stranded capacity from the grid's perspective; if it can sell grid services, the developer gets a revenue stream that offsets operational costs but is not disclosed in the contract terms. Ask PUCO: has the developer applied for grid-service revenue rights, and if so, should those revenues reduce the developer's cost of capital or be shared with ratepayers?

The clean alternative is straightforward and proven. Data-center loads in Ohio should be required to meet three conditions: (1) Bring Your Own Capacity (BYOC) for at least 50 percent of contracted MW, deployed within three years of signing a utility interconnection agreement; (2) a curtailable/flexible-load tariff that allows the data center to access cheaper grid electricity at a discount in exchange for agreeing to shed load during system peaks (1, 3 percent of annual hours); and (3) a long-term demand ratchet (85, 90 percent of contracted transmission and generation capacity for 15+ years) so that if the load underperforms, AEP or Duke cannot socialize unused capacity costs into residential rates. These terms are standard in Virginia's Schedule GS-5, Oregon's Schedule 96, and Ohio's own special-contract framework for large industrials. PUCO should open a docket immediately to require all data-center interconnection requests filed after this announcement to propose one of these three paths, or file as conventional grid-based customers accepting full cost responsibility and socialized-cost risk. The microgrid deal sidesteps that choice. Veolia's contract should be filed with PUCO within 30 days, with all redactions tied to specific confidentiality requests that PUCO then rules on individually. Without that filing, the utility commission has ceded its jurisdiction to a private operating company and its undisclosed developer, and Ohio ratepayers have no seat at the table when the deal's assumptions collapse.