Indiana nuclear power forecast: gas may arrive before 6 GW

Indiana’s new nuclear proposal carries a more immediate signal for the power market: gas generation may come first. AtkinsRéalis, the state and Hoosier Energy have signed an agreement to explore up to eight CANDU reactors with as much as 6 gigawatts of capacity. The evaluation also includes smaller gas plants ahead of the nuclear buildout.

That sequencing changes the Indiana nuclear power forecast. The headline number describes a possible long-term destination; it does not establish electricity available to a factory or data center on a particular date. TECHi’s assessment is that the next meaningful forecast upgrade should come from a funded site, contracted customers and a credible delivery schedule—not another increase in proposed megawatts.

What the Indiana nuclear agreement changes

The announcement, issued at 5:08 p.m. Eastern on October 19, is a memorandum of understanding to evaluate potential deployment. It sets no firm operating date, construction budget or named AI customer. Financing, offtake arrangements, licensing, environmental review and grid connections remain dependencies.

Those omissions do not make the agreement meaningless. Bringing a technology supplier, a state government and a wholesale electricity cooperative into the same evaluation creates a route to answering practical questions together. It can align a reactor proposal with customers and infrastructure. But cooperation is an early development asset, not evidence that lenders have accepted construction risk or buyers have accepted a power price.

The distinction matters to anyone comparing competing power announcements. A capacity target measures ambition. A signed electricity contract measures a buyer’s willingness to pay under specified terms. A construction decision shows that somebody is prepared to fund the project. Operating output proves the system can deliver. Each step removes a different uncertainty, and none should be silently substituted for the next.

The gas bridge is the forecast’s most important detail

The partners are considering smaller simple-cycle or combined-cycle gas plants before the reactors, including possible gas-to-nuclear development at retired or retiring generation sites. No gas capacity total or delivery deadline is disclosed.

For a customer deciding where to build, that creates two separate investment questions. The first is how to secure electricity for the initial facility. The second is whether a future nuclear supply arrangement offers better long-term economics, reliability or emissions performance. A developer might find the first proposition attractive even while the second remains unresolved.

The commercial test is therefore whether the bridge works on its own. If a gas project needs a future nuclear project to justify its financing, delays in the reactor program could also weaken the earlier phase. Conversely, independently financeable generation with committed customers could serve demand even if the nuclear campus changes size. These are possible financing structures, not arrangements the parties have announced.

A bridge also needs an exit strategy. Future disclosures should explain whether gas assets would continue operating alongside reactors, be repurposed or face a shortened economic life. Without that information, readers cannot calculate the cost of transition or assume that early gas investment automatically becomes useful nuclear infrastructure. Sharing land is different from sharing every piece of equipment.

This is the same analytical distinction behind TECHi’s examination of the SLB–Liberty data-center power alliance: a plausible industrial combination becomes financially measurable only when the customer, scope and payment obligations become visible.

How much electricity could 6 GW actually produce?

Gigawatts measure generating capacity; terawatt-hours measure energy produced over time. If the entire proposed 6 GW were eventually built and averaged a 90% capacity factor for a full year, it would generate approximately 47.3 TWh. The calculation is 6 GW × 8,760 hours × 0.90, divided by 1,000 to express the result in TWh.

The 90% figure here is an illustrative assumption, not a performance commitment from the partners. At an assumed 80%, the same nameplate capacity would yield about 42.0 TWh; at 95%, about 49.9 TWh. These figures describe a fully commissioned fleet in a normal 365-day year. They do not account for a staged opening, transmission losses or contractual allocation among customers.

That distinction prevents two forecasting errors. First, a campus built in phases would not contribute its eventual annual output from the first reactor’s opening. Second, annual energy is not a guarantee that a particular customer can draw its full required load every hour. Outages, network constraints and the terms of the supply contract still matter.

It would also be premature to divide the headline capacity by the reactor count and present the result as a selected unit design. Both figures are upper bounds in an evaluation. A credible model needs the actual configuration, commissioning sequence and expected operating assumptions before it can turn the campus concept into annual supply.

The NRC record shows preparation, not a construction approval

There is independent evidence that regulatory preparation is underway. The Nuclear Regulatory Commission’s August 27 meeting notice describes a pre-submittal discussion of the Enhanced CANDU 6 quality-assurance program. That supports the existence of regulatory engagement; it does not approve an Indiana plant.

The NRC’s explanation of pre-application topical reports makes the boundary clear. These reports allow particular safety-related subjects to be reviewed separately from a specific license application. Approved reports can subsequently be incorporated by reference, subject to their conditions. Review timing depends on the scope and complexity of the material.

For the forecast, the useful progression is from engagement to accepted submissions, completed technical reviews and the approvals applicable to an actual project. Counting a meeting as a license would skip the central uncertainty. Counting every preparatory step as worthless would miss genuine progress. The appropriate question is which specific uncertainty a document resolves, and which requirements still remain.

AI demand supports the opportunity, not a guaranteed buyer

The demand backdrop is substantial but uncertain. Lawrence Berkeley National Laboratory’s U.S. data-center energy report, updated in June 2026, models a reference case of 649 TWh of U.S. data-center electricity use in 2030. Its compounded uncertainty cases span 521–843 TWh. These are modeled scenarios, not contracted consumption or a probability interval.

That range is a reason to investigate additional supply. It is not evidence that a particular Indiana project has secured demand. National consumption can grow while individual campuses are delayed, relocate or choose different power arrangements. A reactor developer needs commitments in the place and period where its electricity will become available, at a price sufficient to support the investment.

The strongest future signal would be an identifiable buyer taking a durable obligation: a minimum purchase, a long contract term or a clear contribution to dedicated infrastructure. A nonbinding expression of interest answers a different question. It demonstrates attention, while leaving the risk of a customer’s changed plans largely unresolved.

Readers should also avoid translating electricity demand directly into an AtkinsRéalis earnings forecast. The announcement does not disclose a project contract value or a revenue-recognition schedule. Engineering work, equipment, construction management and operating services can have different timing and economics. Potential industry demand is not the same as revenue attributable to one supplier.

Indiana’s existing sites offer a starting point, not a shortcut

Indiana’s Office of Energy Development nuclear overview says work with Purdue identified eight retired coal-plant sites as promising nuclear infrastructure candidates. That is a separate planning exercise: eight candidate sites must not be confused with the eight reactors contemplated in the new agreement. The announcement does not designate those eight locations as its selected campus sites.

Existing energy sites can make a proposal worth studying because land, a workforce and grid infrastructure may already exist. The value still depends on site-specific suitability. A prior generating connection is not proof that the network can accept any proposed replacement capacity without upgrades, and a community’s experience with one kind of plant does not settle its views on another.

Cost allocation deserves equal attention. A project could serve a major new customer while requiring infrastructure used by the wider system. The critical questions are who pays upfront, who guarantees recovery if demand disappoints, and which obligations survive a customer’s departure. Those contract terms will tell households more about their exposure than a headline promising affordable power.

There is a concrete Indiana benchmark for that disclosure. AES Indiana says its separate Google agreement has a minimum 15-year term and makes Google responsible for its power costs and required new infrastructure. The utility describes financial assurances, minimum demand commitments and exit provisions. Those are AES’s descriptions of a different project, not terms of the CANDU proposal, and Google is not identified as a buyer for the proposed reactors. They nevertheless show the kinds of contractual protections readers can look for as this evaluation develops.

Three scenarios for the Indiana power outlook

TECHi’s base case is a development-stage program in which site studies, commercial negotiations and regulatory work provide the next evidence. The announced nuclear capacity should remain outside a committed-supply forecast until there is a documented basis for including it. That is a modeling discipline, not a prediction that the project will fail.

In the stronger case, the partners disclose a viable site, funded work, a creditworthy power buyer and milestones consistent with the licensing process. An independently viable gas phase could support earlier demand while the nuclear proposal advances. The forecast could then add capacity in dated stages, with the assumptions visible rather than treating the full 6 GW as one event.

In the weaker case, buyer commitments, construction economics or grid requirements fail to align. The program could shrink, take longer or stop after evaluation. A gas phase could proceed independently, or neither phase could become a funded project. No defensible probability can be assigned to those outcomes from this announcement alone.

The next release that materially changes this forecast should identify who will buy the electricity, who will finance it and when a specific tranche can credibly operate. Until then, Indiana has a significant new nuclear option to investigate—and a proposed gas bridge whose commercial terms may reveal the nearer-term power story first.