AI Data Center Gives Up on the Power Grid, Signs 20-Year Lease With the Plant Next Door
Facing a four-to-seven-year wait to connect to the Texas grid, a data center developer instead built next to a gas plant and signed directly with it — a workaround so common it now has its own acronym and its own equity stake for the utility.
New Era Energy & Digital announced on September 21 that it has secured a 20-year power purchase agreement with Vistra’s Luminant subsidiary, guaranteeing 200 to 207 megawatts for the first phase of its Texas Critical Data Center project outside Odessa. The power will come from a 1,180-megawatt natural gas plant that Vistra already operates, located, per the press release, “immediately adjacent” to the data center site. This is the part of the story that’s easy to miss on a skim: the plant was already there. New Era didn’t build near cheap power by coincidence. It built next door to it on purpose, specifically so it could skip a step that has become the actual bottleneck in the AI buildout — not chips, not money, but the electric grid itself.
That step is grid interconnection, the process by which a new facility gets formally plugged into the regional transmission network, inspected, queued, and approved. In Texas and most of the country, that process currently takes four to seven years. A frontier AI data center, by contrast, is expected to be online in a fraction of that time, because the chips inside it will be at least one, and probably two, generations obsolete by the time a standard interconnection queue clears. So New Era did the logical, slightly absurd thing: it located its facility at the fence line of an existing power plant and signed a private, direct contract for electrons that never have to touch the public grid at all. Power delivery is expected in the third quarter of 2027 — roughly a third of the standard queue’s minimum wait, achieved by the simple expedient of not getting in the queue.
The deal terms tell you how routine this workaround has already become. Vistra isn’t just selling power for two decades; it’s also taking a 5 percent non-voting equity stake in the portion of the data center project it powers, plus rights of first refusal on future phases of the site and first offer rights on New Era’s other projects elsewhere. New Era, for its part, is on the hook for roughly $198.8 million in collateral obligations and still has to convert its contracted megawatts into actual signed tenant leases before the 2027 delivery date arrives — meaning the power is secured well before the customers who will actually use it are. This is now a recognizable genre of deal: utility gets guaranteed 20-year revenue plus a stake in the upside; developer gets power certainty it can point to when courting AI tenants; nobody involved has to wait for a public utility commission to finish its paperwork.
The site itself is not small. New Era’s Texas Critical Data Center sprawls across 493 acres in the Permian Basin and is designed to eventually reach 1.4 gigawatts of total capacity — nearly seven times the power secured in this initial agreement — meaning this deal is best understood as the opening move in a much larger campaign to lock down enough dedicated generation to run a facility that, at full build-out, would consume more electricity than some small countries. It’s also a Permian Basin data center running on natural gas, a detail worth sitting with for a moment: the same oil-and-gas region famous for flaring off “stranded” gas it couldn’t economically transport anywhere is now positioned to burn that gas on-site to train language models instead, which counts, technically, as putting stranded assets to productive use.
None of this is unique to New Era. It’s the emerging default architecture of the AI infrastructure boom: developers increasingly treat the public grid not as infrastructure to plug into but as a queue to be routed around entirely, using direct “behind-the-meter” or fence-line contracts with generators who’d rather sign a guaranteed 20-year deal with one enormous customer than wait for the grid operator to sort out who gets the next available interconnection slot. It is, in its way, a quietly damning verdict on the state of the American grid: the fastest-growing, most capital-flush category of electricity demand in the country has concluded, collectively and independently, that the public system for connecting new demand to new supply doesn’t work fast enough to be worth using.
The generous read is that this is innovative dealmaking, unlocking power that would otherwise sit unused behind a bureaucratic queue. The less generous read is that the country’s electrical grid has become such a well-known point of failure that an entire industry has simply stopped trying to use it, one gas plant fence line at a time. Both readings are true simultaneously, which is the most 2026 sentence in this whole story.
Sources: