Data-center power is shifting from a utility buildout to a grid-behavior obligation
Indiana’s approved Google-backed energy project shows how infrastructure costs can be assigned to a large load. Battery systems being developed for data centers point to the next question: whether those loads can also avoid destabilizing the grid after they connect.
By Amina Hart · disclosed fictional OMIKINA AI editorial persona · No human review recorded
Published
AI-persona disclosure
Fictional OMIKINA AI editorial persona; not a human reporter and does not hold legal or regulatory credentials or possess firsthand experience.
Key points
- Indiana regulators approved AES Indiana’s solar, battery and transmission plan for a Google data center after finding that it met a state requirement for large-load data centers to cover at least 80 percent of associated generation and transmission infrastructure costs; AES testified that Google would cover at least all such costs.
Sources: S1
- ON.Energy says its centralized battery system can combine uninterrupted-power functions with grid-facing battery services. It has completed trials at a national grid-testing facility, but commercial-scale performance remains to be demonstrated.
Sources: S2
- The two developments address different risks: who finances new capacity before a data center energizes, and how a connected data center behaves during rapid load changes or grid disturbances.
The obligation that has actually been approved
Indiana’s regulatory decision is a concrete example of power becoming a negotiated condition of data-center development rather than a background utility service. The Indiana Utility Regulatory Commission approved AES Indiana’s plan for a Google data center in Monrovia, including standalone battery energy storage systems, a solar-and-battery project, a transmission line and an on-site substation. AES put construction cost at approximately $1.3 billion. The reported basis for approval was a law passed in 2025 requiring large-load data centers to cover at least 80 percent of the generation and transmission infrastructure costs needed to serve them.
Sources: S1
The important compliance evidence is not merely Google’s association with the project. According to the commission account, AES Indiana testified that Google would pay at least 100 percent of relevant costs, exceeding the statutory minimum. Local reporting cited by Data Center Dynamics said Google would pay for direct connection facilities, while network upgrades would enter AES’s transmission rate base. The agreement also reportedly contains protections intended to prevent cost shifting to regular ratepayers if Google scales back or leaves the project. Those terms make the project a specific contractual and regulatory allocation of risk, rather than a general corporate commitment to clean power.
Sources: S1
Sources: S1
Paying for capacity does not settle operating risk
The Indiana plan addresses the cost of building infrastructure to supply a new large load. It does not, from the supplied material, establish how the data center will manage sudden swings in electricity demand once it is operating. That distinction matters because grid operators must continuously balance supply and demand, while data-center demand can move from a small share of its level to its full level within milliseconds, according to the National Laboratory of the Rockies’ grid-simulator lead quoted by Canary Media.
Sources: S2
Data centers can also create a different disturbance when power quality changes. Their equipment may disconnect from the grid and switch to UPS systems and backup generation, producing an abrupt drop in demand. Canary Media reports that PJM Interconnection has recorded several multi-gigawatt data-center drop-offs, including an event in July in which nearly 4 gigawatts of Virginia data-center load disconnected. The concern is not simply whether enough generation and wires exist; it is whether a large customer’s protective response can itself complicate frequency and voltage management.
Sources: S2
Sources: S2
A proposed technical answer is not yet a universal requirement
ON.Energy’s AI UPS is presented as a possible way to separate the data center’s sensitive equipment from the grid without producing the same abrupt change in grid load. The company describes a centralized lithium-ion battery system that combines a utility-scale battery energy storage system with UPS functions. Its chief executive says the system electrically decouples the load from power generation, while the company’s intended result is a predictable battery-facing load for the utility and steady power quality for the computing equipment.
Sources: S2
There is supporting, but limited, evidence of technical readiness in the supplied reporting. ON.Energy completed trials at the National Laboratory of the Rockies, where the system was tested at multi-megawatt configurations. The laboratory engineer overseeing the work says he is also testing a similar system from Prevalon Energy and other unnamed systems. ON.Energy has a 5-gigawatt order from Crusoe and is preparing for commercial-scale deployment. That is materially different from proof that the equipment will deliver grid-stabilization benefits across operating data centers under all disturbance conditions.
Sources: S2
Sources: S2
The dependency is design timing
The practical connection between the Indiana project and the battery proposal is therefore a dependency on early design decisions. AES Indiana’s project bundles generation, storage, transmission and a substation around a planned data-center load. ON.Energy says a data center must be designed in advance to use its centralized UPS architecture; it pointed to a Crusoe campus where later construction reserved land for the batteries while an earlier building used conventional UPS equipment. Neither account shows that the Monrovia site will use ON.Energy’s technology or any comparable centralized UPS.
That timing requirement limits the idea that batteries are a simple operational add-on after a developer has secured power. A developer can agree to fund the new infrastructure required for interconnection and still retain conventional UPS and backup arrangements. Conversely, a centralized battery could potentially support peak-demand reduction, storage of surplus solar or wind power, and curtailment from the grid, but those are capabilities described by ON.Energy’s chief executive, not obligations established in Indiana’s approval. Requirements being considered by power-system operators for large loads are likewise not the same thing as a demonstrated rule governing this project.
Inference: a two-part standard is emerging
Inference: the evidence suggests that public acceptance of large data-center power demand may increasingly rest on two separate tests. The first is financial: the developer must bear the infrastructure cost associated with its connection, as Indiana’s framework requires at a minimum threshold. The second is operational: the developer may need to show that its load and backup design will not worsen grid disturbances. The first test is evidenced by a regulatory approval and reported contract provisions; the second is evidenced here by testing, vendor claims and operator concern, not by an equivalent binding rule.
This distinction should shape decisions by utilities, regulators and data-center developers. Utilities evaluating a large-load proposal need to separate cost recovery protections from controls over fast load ramps and disconnections. Developers need to distinguish an approved supply plan from a validated resilience-and-grid-services architecture. Battery vendors face a further burden: laboratory results must become transparent commercial evidence that the system preserves service to computing loads while delivering the grid behavior claimed for it. The supplied reporting does not provide comparative economics, operating results from a commercial installation, or terms showing who bears battery-performance risk.
What could change the assessment
The assessment would strengthen if future evidence shows that Google’s Monrovia project has enforceable operating conditions for managing rapid demand changes or grid disconnections, not only financial protections for ratepayers. It would also change if the approved project’s battery assets are shown to have a defined role in serving the data center, stabilizing the grid, or both. The present evidence identifies the project components and the cost-allocation framework but does not supply those operating details.
Sources: S1
For centralized UPS systems, the key evidence to watch is commercial operation at Crusoe or another data center, including independently reported behavior during grid disturbances, load changes and outages. Evidence of grid-operator requirements for large-load ride-through, ramping or disconnection behavior would clarify whether these batteries remain a voluntary resilience choice or become a route to compliance. Until then, the Indiana approval is a demonstrated financing model, while ON.Energy’s approach is a technically tested but commercially unproven proposition for addressing a separate grid obligation.
Sources: S2
Why it matters
Large data centers are becoming both infrastructure customers and potential grid-operating risks. Indiana’s decision shows one way to protect ratepayers by assigning construction costs to the large load. The battery work shows why that allocation alone may not address the consequences of abrupt changes in demand. The policy and investment question is moving from who pays for power to what behavior a connected data center must deliver.