On October 5, TeraWulf announced that it had signed an amended and restated electric service agreement with Kentucky Power, an American Electric Power company, doubling the contracted power capacity at its Muskie Data Campus in eastern Kentucky from 500 megawatts to 1 gigawatt. The amendment also advances planned delivery of the campus’s second 500 MW phase from 2030 to 2029. It was, by data center announcement standards, a crisp piece of news: a bigger number, a faster date, a named utility partner.
Read more carefully, though, and the announcement is a study in the gap between contracted capacity and actual load. Today, Muskie draws nothing. The first 500 MW phase is targeted to begin ramping in 2028. The second phase’s new 2029 date is explicitly subject to Kentucky Public Service Commission approval and Kentucky Power’s construction schedule. And talk of eventually supporting up to 2 GW is an evaluation, not a commitment. A contracted gigawatt is a real milestone. It is not a gigawatt of load.
What the agreement actually says
The core is a contract amendment. TeraWulf and Kentucky Power replaced their existing electric service agreement with a new one covering twice the capacity: 1 GW instead of 500 MW. The most consequential change is not the headline number but the schedule. When TeraWulf acquired the Muskie site in May 2026 for phased development as an artificial intelligence and high-performance computing campus, the plan called for the first 500 MW in 2028 and the second 500 MW in 2030. The amended agreement pulls that second phase forward by roughly a year, to 2029.
That acceleration comes with a condition that matters more than the date. The 2029 delivery is subject to approval by the Kentucky Public Service Commission and to Kentucky Power’s construction schedule. The commission has not yet approved the amended agreement; until it does, the second 500 MW is a signed contract awaiting regulatory blessing, not a done deal. The first phase, too, remains a target — “begin ramping in 2028” is the company’s language, and ramping is a process, not a switch flip.
On infrastructure, Kentucky Power is developing a 765-kV / 345-kV substation connected to the regional transmission network to serve the site. The voltage levels tell you something about the scale. The 765-kV system is the top of AEP’s extra-high-voltage backbone, the tier of transmission built to move bulk power across regions. Stepping that down to 345-kV for a single campus only makes sense for loads measured in the hundreds of megawatts. It is also the kind of build that takes years, which is why the substation schedule — not the press release — will determine when Muskie can actually energize.
TeraWulf also said it continues to evaluate whether the campus could support up to 2 GW over time. That sentence deserves the same careful reading as the rest of the release. Two gigawatts has not been contracted, has not been studied in public, and would require additional utility planning, infrastructure, and agreements. It signals ambition to prospective customers and investors, but it should not be added to any tally of Kentucky’s incoming load.
Contracted, approved, and energized: three different things
The data center industry’s capacity announcements have become a thicket of numbers that sound interchangeable and are not. Three categories matter, and Muskie illustrates all of them. Contracted capacity is what a developer and a utility have agreed to in writing: the utility commits to make a stated amount of power available on a schedule, and the developer typically commits to pay for it, post financial security, or both. It is a paper allocation of future grid capability. It is also the number that shows up in headlines, because it is the largest number available at the earliest stage.
Approved capacity is what a regulator has blessed. In Kentucky, large-load electric service agreements go before the Public Service Commission, which tests them against a simple but demanding question: will existing customers be harmed? The commission examines cost allocation — who pays for the substation, the line upgrades, the market and transmission costs — and the financial protections keeping ratepayers from absorbing the bill if the project stalls. An agreement that fails that test does not get built on the terms proposed.
Energized capacity is the only category measured in electrons: load actually flowing to a campus through completed infrastructure to buildings that exist. The path from contracted to energized runs through regulatory approval, substation and transmission construction, campus construction, and the phased ramp of computing equipment. At every step, schedules slip. A 1 GW contract signed in 2026 that energizes in phases across 2028 and 2029 is, if anything, an optimistic but plausible version of that path. Many contracted gigawatts never complete it.
The distinction matters because the industry’s announced numbers have inflated faster than its energized load. Developers file interconnection requests for multiple sites, sign agreements to hold capacity while they shop for tenants, and announce round-number capacity figures that describe the site’s ultimate potential rather than its funded build. None of that is improper — holding optionality on power is rational when power is the scarcest input — but it means headline capacity is a measure of developer ambition and utility willingness, not of load the grid must soon serve.
Why Appalachian Kentucky
Muskie sits in Grayson, in Carter County, within the EastPark Industrial Park in eastern Kentucky. The pitch for a site like this is straightforward once you understand the interconnection problem. The scarcest resource in data center development is no longer land or capital; it is a place where the grid can deliver hundreds of megawatts on a timeline matching AI deployment plans. Eastern Kentucky has something most candidate regions do not: a 765-kV backbone running through it, built decades ago to move power out of Appalachian coal country to distant load centers.
That history is the point. The transmission network that once carried coal-fired generation out of the mountains is now the asset attracting the industry replacing it. AEP built its 765-kV system in the 1960s and 1970s to move bulk power from large mine-mouth coal plants, and the lines, rights-of-way, and substation sites remain. Where the old system has headroom — or can be expanded at one large substation rather than through miles of new greenfield transmission — a gigawatt-scale campus becomes feasible on a schedule impossible in a region starting from scratch.
The siting logic extends beyond the wires. An industrial park parcel brings zoning, road access, and a community accustomed to industrial neighbors. Eastern Kentucky brings something less tangible but politically significant: a region that has watched coal employment decline for decades and has reason to welcome large-scale investment. That welcome is not unconditional — Kentucky’s data center debate has been contentious enough that the governor set statewide ground rules, discussed below — but the economic development case for a gigawatt campus in Appalachian Kentucky is easier to make than the same case in a wealthy suburb.
There is a tension worth naming plainly. A campus drawing a gigawatt will employ far fewer people per megawatt than the coal plants and industrial facilities that once anchored the region’s economy — a hyperscale facility might employ a few hundred permanent staff, well under one job per megawatt. Construction brings thousands of temporary jobs, and the tax base contribution is real, but communities weighing a proposal face a facility whose resource appetite is enormous and whose permanent payroll is modest. That arithmetic is part of why the regulatory framework around these projects has tightened.
The regulatory gauntlet
Muskie’s second phase must clear two gates before its 2029 date means anything. The first is the Kentucky Public Service Commission, which reviews the amended agreement. Based on the commission’s handling of TeraWulf’s earlier Kentucky agreement, the review will focus on the terms protecting existing customers: who bears the market, transmission, and customer-specific infrastructure costs; what financial security TeraWulf posts; and whether the agreement’s structure prevents Muskie’s costs from leaking into other customers’ rates. The commission’s job is not to judge whether a gigawatt campus is a good idea. It is to make sure Kentuckians who never asked for it do not pay for it.
The second gate is the framework set by Governor Andy Beshear’s Executive Order on Data Center Development (2026-494), issued August 6, 2026. The order requires developers of large data center projects to demonstrate three commitments. First, no detrimental impact on ratepayers or the environment: developers must submit an energy plan to the Energy and Environment Cabinet showing how they will meet their energy needs without imposing additional costs on other ratepayers, the commission is directed not to let a utility raise other customers’ rates to recover data-center costs, and the cabinet must deny permits for projects that would harm air or water quality, water supply, federally jurisdictional wetlands, or natural resources. Second, developers must commit to paying their full and fair share of state and local taxes, including those supporting local communities and schools. Third, they must commit to engaging transparently with local community leaders and residents, including addressing how the project will create local job opportunities.
TeraWulf said Muskie’s development will proceed consistent with the executive order’s requirements. That phrasing is doing real work. The order sets a statewide baseline — in the governor’s words, it does not say yes to any data center, but it does say no to any center that cannot meet its requirements — while leaving the ultimate local decision to local communities. For Muskie, the energy plan, ratepayer protections, and community engagement are not optional extras. They are part of the approval path, and the commission and cabinet will be watching whether the commitments hold as the project moves from paper to construction.
The Justified precedent
Muskie is not TeraWulf’s first time through this process in Kentucky, and the earlier case is the best guide to how the amended agreement will be received. In August 2026, the commission approved a separate electric service agreement for TeraWulf’s Justified Data Campus in Hawesville, supporting up to 482 MW. The commission found that the agreement’s terms adequately protected existing customers and appropriately allocated the costs and risks of serving the project.
Justified differs from Muskie in an important way. The Hawesville campus redevelops a former Century Aluminum smelter site, reusing electrical infrastructure that once served the smelter’s roughly 482 MW load. The grid capability already existed; the regulatory question was how to allocate its costs, not whether to build it. Muskie, by contrast, requires Kentucky Power to build a new 765-kV / 345-kV substation — a different risk profile of cost overruns, schedule delays, and stranded investment if the campus’s load ramps slowly. The commission will want those risks assigned to the developer, not to other customers, and the Justified order suggests it knows how to demand that.
The Justified approval also shows that the process can move. The commission reviewed the agreement and issued its order within the same month the project’s timeline demanded it. If the Muskie amendment follows a similar path, the regulatory gate need not be the long pole. But “can move” is not “will move,” and the Muskie agreement’s larger scale and new-build substation give regulators more to examine.
What to watch
Three milestones will tell whether Muskie’s contracted gigawatt is becoming an energized one. The first is the commission’s action on the amended agreement: watch for the filing, the cost-allocation terms, and the financial protections required. The second is Kentucky Power’s substation construction — the physical critical path. The third is a tenant announcement. As of the October 5 release, TeraWulf has named no customer for Muskie, and that absence is the most important fact in the announcement.
Treat the company’s characterization of customer interest accordingly. Chief executive Paul Prager said customer interest in Muskie has been “tremendous.” That is a chief executive describing his sales pipeline, not disclosing signed contracts. In an industry where developers routinely market capacity contracted but not yet leased, “tremendous interest” is consistent with anything from a nearly full campus to exploratory calls. The verifiable facts are the contract, the schedule, and the conditions. Everything else is marketing until a tenant’s name is attached.
The same discipline applies to the 2 GW figure. Evaluating a site’s potential to support twice its contracted capacity is prudent planning; announcing it is also a way to keep Muskie in the conversation about the largest AI campuses. But 2 GW at Muskie would require additional utility planning, additional infrastructure, and additional agreements — each of which would have to clear the same regulatory gates as the current amendment. Count it as ambition, not inventory.
A contracted gigawatt in eastern Kentucky is still a meaningful signal: the economics of AI infrastructure now justify building a new extra-high-voltage substation in Appalachian coal country, a major utility is willing to commit to serving the load, and a developer is willing to put financial security behind the commitment. Those are real facts about where the industry is going. But the grid keeps score in energized megawatts, and on that scoreboard, Muskie is at zero — with a long, conditional path to one.