A correction first, because the number in the headline needs one. NiSource’s October 16 announcement did not say its new Microsoft agreement will save NIPSCO electric customers $1.75 billion. It said the Microsoft deal builds on previously disclosed expected savings of $1.4 billion over the life of all of the company’s data center customer contracts, lifting the total to approximately $1.75 billion. The headline figure is cumulative, it is a company projection, and it covers the whole data center portfolio — of which Microsoft is the newest and largest addition. None of that makes the deal less significant. But if you are going to ask what “customer savings” actually means, you should start by knowing what the number measures.
The deal itself, announced Friday from NiSource’s Merrillville headquarters, is the largest load agreement the company has signed and arguably the clearest expression yet of the template converging across the industry: the hyperscaler pays for its own infrastructure, incumbent customers are insulated from the incremental costs, and the utility keeps the savings narrative front and center. Under a 20-year term beginning in 2028, NiSource will supply power to Microsoft’s previously announced northwest Indiana campus (reported as the La Porte development) through NiSource Generation Company LLC — GenCo — with Microsoft fully funding all generation and transmission required to serve its load. The GenCo-owned pool portfolio grows by approximately 1,100 MW, and alongside the agreement GenCo signed a long-term tolling agreement for 600 MW of battery storage to be developed by an affiliate of AGP Group.
The mechanics of “savings”
The press release contains the most important sentence in the whole deal, and it is not the one with the dollar figure. Through GenCo, NiSource says, the company is serving new large-load customers under a structure “designed to protect existing Northern Indiana Public Service Company electric customers from incremental costs and return value from large-load customers’ use of the existing electric system.” That is two mechanisms, not one, and each deserves a separate reading.
The first mechanism is protection: Microsoft fully funds all generation and transmission required to serve its load. The new 1,100 MW of generation resources, the transmission infrastructure to connect them, the storage under the tolling agreement — none of this goes into the rate base that ordinary NIPSCO customers pay. This is the same principle DTE applied to its storage contracts in Michigan (customer-funded, kept out of general rates) and that Duke encoded in its North Carolina settlement (upfront payments for connection assets, deposits against shared upgrades). The industry has converged on a rule: the customer that causes the incremental investment pays for it.
The second mechanism is where the $1.75 billion comes from, and it is subtler. “Return value from large-load customers’ use of the existing electric system” means the new load also pays toward the fixed costs of the system it joins — the plants, wires, and operating apparatus that already exist and already have to be paid for. When a large new customer takes service from a pool that includes existing assets, its payments contribute to those assets’ fixed costs, spreading them across more sales and reducing the share borne by existing customers. That contribution — not a rebate check, but a spreading of fixed costs over a bigger denominator — is what NiSource is projecting as savings: approximately $1.75 billion over the life of the data center contracts, or as much as $140 per customer annually.
There are honest caveats to the arithmetic, and they belong before the utility’s critics raise them. The $1.4 billion base and the $1.75 billion total are projections, not measurements; they assume the contracts perform as written over their full terms, which run into the 2040s. The $140-per-customer figure is an annualized average, not a line item anyone will see on a bill — savings flow through rate cases and fuel proceedings as lower costs than would otherwise have been charged, not as credits. And the denominator matters: NIPSCO serves about 500,000 electric customers, so per-customer figures are sensitive to how benefits get allocated across customer classes over two decades.
The GenCo chassis: pool, toll, and who builds
The institutional detail that distinguishes this deal from DTE’s or Duke’s is the GenCo vehicle itself. NiSource Generation Company LLC is the company’s generation arm; under the announced structure, NIPSCO and GenCo together provide the transmission and generation infrastructure, while the GenCo-owned pool portfolio — increased by roughly 1,100 MW — serves Microsoft’s data center alongside NIPSCO’s existing customers. The release is explicit that the pool approach provides “a safe, reliable, and coordinated power supply” and that the new assets will upgrade NIPSCO’s electric transmission system, improving reliability and resilience for customers across northern Indiana.
The pool concept is doing real work here. The 600 MW of battery storage contracted under the tolling agreement with the AGP Group affiliate is not dedicated to Microsoft alone: the release says it will be “added to GenCo’s pool resource assets designed to serve multiple large-load customers.” The storage is a shared platform asset, funded in connection with the Microsoft deal but available to the next hyperscaler that signs. That is a different design choice than DTE’s, where the storage sat in a customer-specific 15-year contract. NiSource is building a portfolio against a pipeline.
The filing behind the press release adds texture — and risk. As disclosed in the company’s October 16 8-K, the tolling agreement covers a 600 MWAC / 2,400 MWh standalone battery project (Atlas Storage 1 LLC, an indirect subsidiary of AGP Capital Holdings Pte. Ltd.) connecting to the MISO transmission system, with aggregate storage payments of about $2.2 billion over a delivery term of roughly twenty and a half years, the first phase expected in commercial operation by December 1, 2027. GenCo supplies the charging energy, directs dispatch, and takes the capacity, discharged energy, ancillary services, and environmental attributes — the classic tolling split: the owner builds and owns the battery, the tolling counterparty decides when it charges and discharges and keeps the economics.
The conditions in the filing deserve a second read: the agreement’s effectiveness is conditional on AGP Group executing the EPC contract and a qualifying interconnection agreement by January 31, 2027. The 600 MW that anchors the reliability case does not yet have its construction contract or grid interconnection finalized. GenCo holds an exclusive option to buy the project at intervals — a backstop — but near-term execution risk sits with a third-party developer hitting milestones, with first-phase commercial operation expected December 1, 2027, just ahead of Microsoft’s 2028 service start.
Who bears the construction risk
The pitch framing — “the hyperscaler pays for its own infrastructure” — is accurate about who funds the assets and incomplete about who executes them. Microsoft’s commitment is to fully fund the generation and transmission required to serve its load; that is a payment obligation, the load-side equivalent of DTE’s minimum monthly charges and Duke’s upfront payments. But the procurement and construction sit with NiSource’s GenCo: it contracted the 1,100 MW of generation resources, it signed the storage toll, and it owns the relationship with AGP Group.
The company’s own forward-looking statements acknowledge where the risk lives: receipt, timing, and terms of required regulatory approvals; the ability to place generation and transmission assets into service on time and consistent with initial cost estimates; the ability to obtain the significant additional financing required; customers’ performance under the contracts; and potential changes in MISO’s accreditation treatment of capacity resources. That last item is worth pausing on. MISO has been tightening how it accredits capacity resources — batteries included — which means the capacity value GenCo is counting on from the storage could be re-rated by the market operator after the contracts are signed.
So the risk allocation, stated plainly: Microsoft assumes the long-term payment risk (twenty years of funding the assets built for its load). NiSource assumes the execution risk (building them on time and on budget, obtaining approvals, financing the build). Existing NIPSCO customers are, by design, outside both risks — protected from incremental costs and positioned to receive the fixed-cost contributions that make up the savings figure. If construction slips or costs overrun, the question becomes how far Microsoft’s “fully funds” commitment stretches — and the release does not specify. Expect that question on the third-quarter earnings call on October 28, where NiSource has promised further operational and financial details.
The market’s first read was favorable: NiSource shares rose about 2.8% in premarket trading on the announcement, per Reuters, as investors priced in twenty years of contracted load with the cost risk placed on the customer. Microsoft shares were roughly flat — for a company of Microsoft’s scale, fully funding a gigawatt-scale power supply is a hedge against the regulatory and reputational risk of being seen as the reason household bills rose.
The template, three ways
Put the three October deals side by side and the convergence is unmistakable, even as the institutional machinery differs. DTE’s approach, as detailed in this desk’s October 17 piece, serves near-term hyperscale load from existing capacity plus customer-funded storage — 1.4 GW for Oracle under a 19-year contract with an 80 percent minimum billing demand, a separate 15-year customer-funded storage contract, nearly $2 billion of incremental storage investment, and Michigan commission approvals with conditions that put unrecovered costs on DTE rather than other ratepayers. Storage before new steel; defer the generation to the back of the plan.
Duke’s approach, from the October 10 piece, is tariff architecture rather than bespoke contracts: a North Carolina settlement with the Public Staff — joined, remarkably, by Amazon, Google, Meta, and Microsoft themselves — setting 50 MW / 150 MW thresholds, moving large loads onto a High Load Factor rate schedule, requiring upfront payment for dedicated infrastructure, imposing a 75 percent minimum bill and exit fees ranging from $25 million to over $587 million. The rules are general; the allocation methodology for shared upgrades is still to be decided.
NiSource’s approach is the GenCo pool: new generation built for the load (1,100 MW), storage contracted as a shared platform asset (600 MW toll), the customer fully funding the incremental infrastructure while contributing to the fixed costs of the existing system, all inside a twenty-year agreement with the savings narrative quantified up front ($1.75 billion). Where DTE leans on existing capacity and Duke leans on tariff design, NiSource leans on new build — which is also why NiSource’s construction-risk exposure is the largest of the three.
The common thread is the political economy, and the hyperscalers’ behavior explains it. Microsoft’s Jeff Riles, senior director of energy markets, put the industry’s position in the release: “Microsoft has always been committed to paying our own way and identifying long-term solutions like this one to ensure that our data centers don’t increase electricity prices for other ratepayers.” That sentence has now appeared, in substance, in every one of these deals — from Duke’s settlement signatories to DTE’s contract terms to NiSource’s funding structure. The hyperscalers have concluded that certainty and social license are worth the premium of paying their own way. For utilities, the lesson of the month is that the customer-pays structure is no longer a negotiating position; it is the entry ticket.
What to watch
The calendar is unusually concrete for a deal this new. NiSource reports third-quarter 2026 results on October 28, with the promised operational and financial details — watch for the contract’s payment mechanics, any disclosed minimum-bill or collateral terms, and how the company describes the GenCo pool’s regulatory treatment. The storage agreement’s January 31, 2027 condition deadline is the first hard milestone: EPC contract and interconnection agreement, or the 600 MW toll is not yet a deal. First-phase commercial operation is expected by December 1, 2027; Microsoft’s service start follows in 2028.
Beyond the milestones, the questions that will decide whether the $1.75 billion survives contact with reality. First, regulatory approvals: the release’s own forward-looking statements cite required approvals without specifying them, and Indiana’s Utility Regulatory Commission has yet to weigh in on the GenCo pool’s cost allocation — the mechanism that turns Microsoft’s payments into other customers’ savings. Second, the campus itself: the power contract is one approval among many, and Microsoft’s construction timeline has to converge with GenCo’s asset timeline. Third, MISO accreditation: if the market operator re-rates battery capacity value, the economics of the $2.2 billion toll shift.
The bottom line: NiSource has written the most ambitious version of the industry’s new template — new generation, not just existing capacity; a shared storage platform, not a single-customer asset; a quantified savings promise, not just a protection pledge. CEO Lloyd Yates’s framing — “Affordability is not separate from economic development; it is a key pillar of our GenCo model” — is the thesis in one sentence. The thesis is now funded for twenty years. The test, as with DTE’s storage build and Duke’s allocation methodology, is execution: the assets have to be built, the approvals have to land, and the customer has to pay. The $1.75 billion is a projection. The contract is a fact. The next eighteen months will show how much of the former the latter can deliver.