On October 7, the Indiana Utility Regulatory Commission did something it had never done before: it approved a data-center power plan under a law written specifically for data centers. The vote was unanimous. The project is AES Indiana’s plan to serve a 390-megawatt Google data center in Monrovia, in Morgan County southwest of Indianapolis, with regulated solar, battery storage, and transmission lines. The order puts the total project cost at $737.3 million.
That sentence contains the whole argument now playing out across Indiana. The commission found that the plan satisfies a 2025 state law, House Enrolled Act 1007, which requires data centers and other large electricity users to cover 80 percent of the costs reasonably assigned to them, and to protect existing and future customers. AES Indiana filed the case on April 22, 2026, as Cause No. 46394, and it is the first data-center filing under the new law. Commissioner Joby Jerrells, summarizing the commission’s view, said AES had testified that Google would pay for at least 100 percent of the costs under the requirement. “Not only does this agreement satisfy the statutory requirement, it exceeds it substantially,” Jerrells said.
And yet the order also puts $73.8 million of network upgrades into AES’s transmission rate base, to be recovered in future retail rates. Google pays $32.3 million directly for connection facilities. The rest of the cost split is where the template gets tested, because the two numbers tell different stories depending on who is reading them. AES Indiana says none of the costs are shifted to existing customers, and that Google’s payments into the utility system will more than make up for the rate-based share. The Citizens Action Coalition, the consumer group that intervened in the case, reads the same order and sees $74 million of transmission upgrades identified specifically because the Google campus is being added to the grid, now sitting in a rate base that ordinary customers fund.
The anatomy of the Monrovia deal
The project itself is a regulated build, and that is the first thing that distinguishes it. AES Indiana is not merely delivering power to Google; it is building the supply. The plan covers a solar facility, two battery energy storage systems, the electric lines that connect them, and an on-site substation, all designed mainly to serve large electricity customers, with the Google campus as the anchor. AES describes the wider investment at about $1.3 billion, against the $737.3 million total project cost stated in the order — the difference between the filed project and the broader investments AES associates with serving the campus.
The contract with Google runs at least 15 years, according to AES, and it includes terms meant to keep other customers from paying for the upgrades if Google scales back or leaves. AES’s modeling estimates about $770 million in savings for other customers over the term of the project, on the logic that spreading fixed costs across a much larger load base brings the cost per customer down. Google’s head of data centers in the Americas, Will Conkling, has said the company is paying for 100 percent of its electricity and any new infrastructure required to serve it. AES Indiana President Brandi Davis-Handy called the approval “an important milestone.”
The opposition, such as it was, focused on one line item. Ben Inskeep, program director at the Citizens Action Coalition, named the concern plainly: “What we’re concerned about is a $74 million set of transmission upgrades that AES Indiana identified that it had to make specifically because this Google data center is being added to the grid.” Brian A. Wright, an electric utility analyst with the Indiana Office of Utility Consumer Counselor, testified that the transmission upgrades would not be necessary without the data center’s energy needs — though the consumer counselor’s office did not oppose the plan. Inskeep’s argument was that the data center, not the customer, should foot the upgrade costs: “We already have high bills here in Indianapolis. We don’t need more rate increases caused by data centers to be added on top.”
AES’s answer came from its own witness, Bickey Rimal of Concentric Energy Advisors, who testified that assigning all of the transmission upgrades to Google would not be reasonable, because the utility had never done that with a customer before and because the upgrades benefit the whole system. The commission agreed. That reasoning is the crux of the Monrovia order: network upgrades triggered by a single customer can still be system upgrades, shared by the system. The 80 percent test, as applied here, is not a test of whether every dollar traceable to Google is paid by Google. It is a test of whether the overall arrangement leaves other customers protected — and the commission found that it does, on AES’s evidence.
There is also a transparency wrinkle that will follow this template into future cases. AES asked to keep portions of the temporary, electric, and construction service agreements and some testimony confidential, and the commission agreed, on the ground that the material includes trade secrets exempt from public disclosure under state law. The Citizens Action Coalition objected. AES spokeswoman Claire Rice later said the customer-protection terms and customer-benefit details were not redacted — but dollar amounts, rate schedules, and timelines in the agreements were. A template built on confidential exhibits is one other parties will struggle to audit.
Experiment two: the wholesale contract
Monrovia is not Indiana’s first answer to the hyperscaler question. It is the third, and the contrast with the other two is what makes Indiana the most instructive state in the country on large-load cost allocation right now.
The oldest of the three is Duke Energy Indiana’s arrangement for Meta’s data center in Jeffersonville, in Clark County. There, Meta contracted through a new limited liability company, Blocke, in a deal built around renewable power purchase agreements, with Duke permitted to buy electricity from the regional wholesale market on Blocke’s behalf when the contracted renewables do not deliver enough energy. The expected load and rates were redacted as confidential. The structure keeps the hyperscaler’s energy procurement largely at arm’s length from Duke’s regulated generation fleet: Meta assembles its own supply, the market fills the gaps, and the utility’s role is closer to wires and procurement agent than to builder.
The consumer counselor’s office, in regulatory filings, agreed that the arrangement “does not create additional costs for other customers” and that it “ensures Blocke bears all responsibility for costs it causes and makes some contribution to fixed costs.” But the same fault line appeared here as in Monrovia. Inskeep noted that Duke received approval to shift $82 million of transmission upgrades required to connect the data center onto other ratepayers. The wholesale-contract model, in other words, did not eliminate the network-upgrade question; it moved it. The generation is the customer’s problem. The wires, it turns out, are still partly everybody’s.
That is a pattern worth noticing. In both the Duke and AES cases, the headline deal — the generation, the contract, the renewable procurement — was structured to protect other customers, and in both cases the transmission upgrades became the contested remainder. The reason is structural: a data center’s connection changes flows on the shared network in ways that cannot be neatly fenced off to one customer, and every utility witness in these cases has argued that shared network investment benefits shared network users. Consumer advocates counter that upgrades that would not exist without the data center should be paid for by the data center, whatever their incidental benefits. Indiana now has two commission decisions blessing the utility version of that argument, in different forms.
Experiment three: the affiliate
The third model is the most ambitious, and the largest by an order of magnitude. In northern Indiana, NIPSCO is serving Amazon through GenCo, a separately structured generation affiliate approved by the IURC in June 2026. The framework is designed to add up to 3 gigawatts of new capacity — two 1.3-gigawatt combined-cycle gas plants and a 400-megawatt, four-hour battery storage system — dedicated to Amazon’s data center campuses, with Amazon’s contracted capacity reaching about 2,800 megawatts by 2032 and initial power targeted for January 2027.
GenCo’s logic is ring-fencing by corporate form rather than by contract clause. By operating as a separate entity, GenCo isolates the cost of new growth to the data center customer: Amazon funds the generation and transmission infrastructure its demand requires, and existing NIPSCO customers are, in NIPSCO’s framing, protected from those costs even if the customer leaves before the contract ends. The companies estimate the arrangement will return roughly $1 billion to $1.25 billion to NIPSCO customers over 15 years in bill credits and savings, depending on whose estimate is cited — a range that itself illustrates how much of this debate runs on competing models rather than settled arithmetic.
The scale difference matters for the template question. Monrovia is 390 megawatts and a regulated build. Jeffersonville is a wholesale contract for a single campus. GenCo is a multi-gigawatt, gas-fired buildout wrapped in an affiliate structure, at roughly $7 billion of investment. If Monrovia’s cost split survives scrutiny, it becomes the template for the next regulated-build filing. If GenCo’s ring-fence holds through construction, rate cases, and any future dispute over stranded costs, it becomes the template for the next multi-gigawatt campus. And if either model cracks — if the rate-based upgrades in Monrovia become a political problem, or if GenCo’s separation proves porous — the other model gains.
The scrutiny that is still coming
None of the three models has yet survived a full cycle of scrutiny, and Monrovia’s approval arrived in the middle of a separate fight that could still touch it. The commission recently agreed to reconsider its June approval of a rate increase for AES Indiana, and that rehearing will consider evidence related to the Monrovia project and to the announced acquisition of AES’s parent company by BlackRock and other investors, with proceedings continuing into 2027. Customers keep paying the higher rates while the review runs. An approval order is not the end of a cost-allocation argument; it is the beginning of the record for the next one.
There is also the political backdrop, which in Indiana is unusually concrete. More than half of the state’s counties now restrict data centers in some form, through moratoria or bans, according to Indiana Public Media’s reporting on local measures. A cost-allocation model that reads as fair in a commission order can still fail in a county council meeting, and the siting pipeline for the next campus depends on both. The Monrovia approval does not settle the local question in Morgan County, where residents intervened in the case. It settles the regulatory question of whether this particular deal may proceed.
Why Indiana is the template fight
Step back and the shape of the thing is clear. The national debate about data centers and electricity bills has mostly been fought in slogans: hyperscalers should pay their own way, or hyperscalers lower everyone’s bills by spreading fixed costs. Indiana is where those slogans are being converted into enforceable structures, and the conversion is producing three different answers at the same time, in the same state, under the same commission.
The regulated-build answer says: let the utility build clean supply for the campus, hold the customer to a long contract with exit protections, and let a share of network upgrades sit in rate base because the network is shared. The wholesale-contract answer says: let the customer assemble its own supply, let the market fill the gaps, and keep the utility’s role narrow — but accept that the wires question survives anyway. The affiliate answer says: separate the whole enterprise into its own company, fund it entirely from the customer, and return the surplus to existing customers as credits.
Each model has a different theory of what a utility is for. The regulated build treats the utility as a builder. The wholesale contract treats it as a conduit. The affiliate treats it as a holding company with a firewall. Those are not just financing choices; they are different answers to the question of who bears the risk when a 390-megawatt customer, or a 3-gigawatt one, changes its mind, scales back, or simply arrives later than planned.
For developers and utilities watching from other states, the lesson of Monrovia is narrower than the headlines suggest. The IURC did not rule that hyperscalers pay 100 percent of everything. It ruled that a specific arrangement — a 15-year-plus contract, exit protections, direct payment for connection facilities, and a rate-based share of network upgrades justified as system benefits — satisfies a statute that demands 80 percent and customer protection. Future applicants will cite that ruling, and future intervenors will cite the $73.8 million. Both will be quoting the same order.
That is what makes Indiana the template fight. The state did not choose one model. It approved three, and the one that survives the next rate case, the next rehearing, and the next county vote becomes the answer every other state copies. Monrovia, as the first filing under the new law, gets to go first. It will not get the last word.