On November 6, during its third-quarter earnings call, PPL Corporation announced the most consequential data-center power deal a Pennsylvania utility has struck — and it is not a deal to deliver power to a data center. It is a deal to build the power plants. Invitium Energy, PPL’s joint venture with Blackstone Infrastructure, has agreed to all key commercial terms with QTS to supply approximately 2 gigawatts of generation capacity for QTS’s data center project in Salem Township, Luzerne County, Pennsylvania. The portfolio pairs two combined-cycle natural gas plants with fuel cells and battery storage. Fuel cells and batteries could come online as early as 2029; the gas units are expected in 2032 and 2033.
The announcement matters less for the megawatts than for the model. For two years the industry’s answer to hyperscaler load has been “bring your own power”: the developer assembles its own supply while the utility provides wires. PPL is proposing the inverse. The utility holding company builds the generation itself, through an unregulated joint venture, and sells the output to the data center under a long-term contract. The customer brings the load; the utility brings the power plant.
The terms, and the one that matters most
PPL disclosed most of the commercial terms on the call. The term sheet provides a 20-year initial term with provisions for multiple five-year renewals, a fixed monthly payment with annual escalation covering the return of and on invested capital, and variable payments for fuel when the plants run. QTS has also signed a cost reimbursement agreement covering specified development costs, which backstops PPL’s spending if the project does not proceed. Management put PPL’s share of the investment in the QTS project at $2.5 billion to $3.5 billion. The venture has already spent $40.4 million assembling more than 500 acres in Luzerne County, management said.
The term that matters most is the one PPL stated plainly: the term sheet is nonbinding, subject to customary conditions precedent. What was announced on November 6 is an agreement on all key commercial terms, not a signed power purchase agreement, and the conditions between here and a binding contract are the same ones that have killed data-center power deals before. The cost reimbursement agreement matters because the deal is not yet final: QTS, not PPL’s shareholders alone, bears specified development costs if approvals fall through.
That structure — fixed payments for capital, variable payments for fuel, a long initial term with renewal options — is a tolling arrangement, and it is worth naming it as such. QTS is not buying electricity at a market price. It is paying a fixed monthly charge for the right to call on dedicated capacity, plus fuel when the plants run. The venture takes construction and operating risk, QTS takes long-term demand risk, and fuel-price risk sits with the customer. It is a familiar structure in independent power; what is new is who is on the other side of it.
The fourth model
Fracmoo has now covered four distinct answers to the question of who builds the power for a hyperscaler, and Invitium is the fourth. The first answer is bring-your-own-power: the customer assembles supply itself. Google’s expanded agreement with Voltus pays for more than a gigawatt of batteries, thermostats, and industrial curtailment across PJM through 2040 — flexibility as capacity, bought by the customer. At the other end of the same logic, xAI is running rebuilt jet engines at a plant in Southaven, Mississippi, to power its Memphis campus, and Venture Global has signed frame agreements for up to 16 gas-fired power islands of 720 megawatts each, putting the first public price arithmetic — roughly $1,490 per kilowatt — on on-site gas.
The second answer is the regulated build, covered on this desk in Indiana last month. AES Indiana won unanimous commission approval to serve Google’s 390-megawatt Monrovia campus with regulated solar, battery storage, and transmission, at a $737.3 million project cost, under a 2025 state law requiring large-load customers to cover 80 percent of assigned costs. The third answer is the affiliate: NIPSCO’s GenCo, approved by the same commission in June, is a separately structured generation company building up to 3 gigawatts — two 1.3-gigawatt combined-cycle plants and a 400-megawatt battery — for Amazon, at roughly $7 billion, with the corporate firewall doing the ring-fencing.
Invitium is none of those. It is not the customer building its own supply, not the regulated utility building rate-based generation, and not a ring-fenced affiliate of a regulated utility. It is an unregulated joint venture between a utility holding company and an infrastructure investor, building merchant-style generation under a long-term bilateral contract with a single customer. Pennsylvania law bars PPL’s regulated utility, PPL Electric Utilities, from owning generation, so the venture exists precisely because the regulated route is closed. The generation is priced by contract rather than by a regulator, and it sits on an unregulated balance sheet rather than in rate base.
Why Pennsylvania, and why now
The demand case for the venture is not in doubt. PPL Electric Utilities’ Pennsylvania data-center pipeline reached 35.3 gigawatts in advanced planning stages in the third quarter, the eleventh consecutive quarterly gain, up from 31.8 gigawatts in the second quarter. Signed service agreements totaled 11.4 gigawatts, and about 8.5 gigawatts of data-center load is under construction — up 2 gigawatts in a single quarter. In Kentucky, where PPL also operates, the economic development pipeline grew to 14.7 gigawatts through 2032, of which 12.8 gigawatts is data-center related. Much of the pipeline will never materialize, but the under-construction figure is real load, arriving years before the generation meant to serve it.
Salem Township is the sharp end of that pipeline. The township in Luzerne County already hosts the Susquehanna nuclear plant and the Moxie Freedom combined-cycle gas plant, and Amazon bought Talen Energy’s data-center campus next to the nuclear plant in 2024. QTS assembled its own position in the township earlier this year, paying a reported $586 million for about 1,700 acres. One township now holds a nuclear plant, a gas plant, an Amazon campus, a QTS campus, and a proposed 2 gigawatts of new generation — the load landing where the generation already is.
The grid context is what makes the timing acute. PJM’s capacity auction for the 2028/2029 delivery year, held July 14, cleared at its $325 per megawatt-day price cap and still came up 6,831 megawatts short of the reliability requirement — the second consecutive year the auction failed to procure enough capacity. Only about 317 megawatts of new generation cleared the auction. PJM’s own simulation showed the auction would have cleared at $554.72 per megawatt-day without the cap. The region’s response runs through a backstop: PJM filed its Reliability Backstop Procurement with federal regulators on July 31, proposing a one-time procurement to close the shortfall, with a bid window that ran September 30 through October 21 and results due December 2. Federal regulators rejected PJM’s large-load rule on October 9, pushing the question of who pays — and who gets curtailed first — to state regulators.
The timeline gap
Against that backdrop, the Invitium timeline deserves a hard look: it is the deal’s weakest feature. The fuel cells and batteries could come online as early as 2029. The combined-cycle gas units — the bulk of the 2 gigawatts — are expected in 2032 and 2033. That is six to seven years from the announcement, and it is worth asking what powers the QTS campus in the meantime.
The announcement does not say, and the gap is the central open question of the deal. One possibility is phased energization: the campus ramps in stages on grid power while the dedicated generation is built, with the fuel cells and batteries bridging part of the load from 2029. Another is that the campus’s own ramp is slower than the headlines suggest, and the 2032 and 2033 dates match the load’s arrival better than they appear to. A third is that the campus energizes on grid power and the dedicated generation arrives as backfill — meaning load already under construction draws on a system that just failed its capacity auction two years running.
None of those possibilities is disqualifying. Every large-load deal in PJM has a version of this gap, because a combined-cycle gas plant takes five to seven years from commitment to operation and a data center takes two to three. The infrastructure timeline, not the building timeline, determines when a facility can operate. What is notable is that the gap is disclosed in the announcement itself — 2029 for the fast pieces, 2032 and 2033 for the gas — rather than buried in a filing. The question is whether QTS, the township, and the grid are comfortable with the years in between.
Who bears the risk
The risk question is where the model will be judged, and the early evidence cuts both ways. For existing customers, the structure is designed to be reassuring: the generation sits outside the regulated utility, so its costs do not enter rate base, and the customer pays a fixed monthly charge for twenty years, with a cost reimbursement agreement covering development spending if the deal falls apart. Pennsylvania’s governor has moved the same way: an August 18 executive order requires developers of projects over 25 megawatts to commit to the state’s GRID requirements and secure local approval before the environmental agency reviews a permit, and directs the utility commission toward rules curtailing data centers before other customers in emergencies and making them bear backstop capacity costs.
For shareholders, the deal is the growth story PPL told on the same call. Third-quarter ongoing earnings came in at $0.54 per share, up from $0.48 a year earlier and a cent ahead of the consensus estimate, on revenue of $2.50 billion. PPL narrowed its 2026 ongoing earnings guidance to $1.92 to $1.96 per share, introduced a 2027 forecast of $2.06 to $2.14, and raised its long-term annual earnings growth target to 7 to 9 percent through at least 2030, up from 6 to 8 percent. The capital plan for 2027 through 2030 rose to $25 billion, a $2 billion increase driven by an additional combined-cycle unit in Kentucky. Funding that plan requires about $3.0 billion of additional equity on top of more than $1.5 billion already executed through forward sales. The Invitium venture is, in that telling, the unregulated engine of a regulated growth story: demand that strains the regulated grid becomes, through the joint venture, earnings growth the regulated utility could never capture on its own.
The tension between those two tellings is the real story of the deal. The structure protects ratepayers from the cost of the generation, because the generation is not in rate base. But it also means the upside — the fixed monthly payments, the return on $2.5 billion to $3.5 billion of PPL-share investment — accrues to shareholders and to Blackstone, not to the customers whose grid the data center will draw on in the years before the gas units arrive. That is a defensible allocation; it is also a choice. Invitium’s structure avoids the cost-allocation fight that Monrovia’s rate-based share provoked, by keeping the generation off the regulated books entirely — which is precisely why the argument will instead be about the years before 2032, when the load is on the grid and the generation is not yet built.
What to watch
Four milestones will tell whether this announcement becomes a power plant. The first is a binding contract: the conversion of a nonbinding term sheet into a signed agreement. The second is permitting — air permits for two combined-cycle plants and a fuel-cell installation in a township already hosting four data-center campuses, under an executive order that makes local approval a precondition for state review. The third is the turbine order: the conversion of Invitium’s 5-gigawatt reservation position — which PPL says supports $6 billion to $9 billion of potential PPL generation investment through 2032 — into firm equipment orders with delivery dates that hold the 2032 and 2033 targets. The fourth is the capacity market: the December 2 backstop procurement results, and the December capacity auction for the 2029/2030 delivery year, will show whether PJM’s shortfall is closing or widening while the plants are still on paper.
The deal’s significance does not depend on any one milestone landing cleanly. Even if the QTS project slips, PPL has now shown the industry a template: a utility holding company and an infrastructure investor, building dedicated generation for a hyperscaler under a twenty-year tolling-style contract, outside the regulated utility, in the PJM territory where the capacity market is failing to produce new supply. If the template works, every other utility holding company in PJM with a data-center pipeline will study it, because it solves the problem their regulated utilities cannot solve: how to build generation for a customer that needs gigawatts, without asking regulators or ratepayers to pay for it. And if it does not work, the 2032 and 2033 dates will be the reason: the gap between the load’s arrival and the generation’s arrival is the one part of the announcement no contract structure can close.