Intent, Commitment and the Gap Between

On August 20, 2025, the Louisiana Public Service Commission voted four to one to approve Entergy Louisiana's plan to build three natural gas power plants to serve a single customer: a Meta data center campus in Richland Parish called Hyperion. Meta had announced the campus in December 2024. The regulatory vote, eight months later, was in many ways a firmer commitment than the announcement, because it put a utility, its regulator and its ratepayers on the hook for infrastructure built to Meta's needs. This lesson is about the gap between those two kinds of event.

A campus announcement commits a company to very little in a legal sense. It is a statement of intent, and intent can be revised, scaled back, delayed or moved. That is not a criticism of the companies. It reflects how large projects work: they are announced before the power is contracted, before the equipment is ordered and often before the financing is arranged, because announcing early helps secure local approvals, tax incentives and utility attention.

Hyperion shows what firmer commitments look like as they accumulate. After the utility approval came financing. In October 2025, Meta announced a joint venture with funds managed by Blue Owl Capital to develop the campus, under which Blue Owl's funds would hold an 80 percent interest and Meta 20 percent, with Meta providing construction and property management. The companies described total development costs of roughly 27 billion dollars for buildings and long-lived power, cooling and connectivity infrastructure. A financing of that size, with outside investors committing capital, is a different order of evidence from a press release.

The Stargate site in Abilene, Texas, illustrates the other direction. The campus is developed by the company Crusoe on land in the Lancium Clean Campus, and Oracle operates capacity there for OpenAI using Oracle Cloud Infrastructure. The first buildings were reported operational in 2025, and OpenAI and Oracle publicly showcased the site that September. Then, on March 6, 2026, Bloomberg reported, citing people familiar with the matter, that Oracle and OpenAI had dropped plans to expand the flagship site beyond its planned buildings after negotiations over financing and OpenAI's changing requirements stalled. The existing campus continued; the expansion did not proceed as planned.

Notice what that episode does and does not show. It does not show that Stargate failed, or that demand for capacity collapsed; the same reporting described other Stargate sites continuing. It shows that the plan for a specific site changed after it had been widely discussed. A reader who had recorded the original plan, with its date and source, would have noticed the change immediately. A reader who had absorbed only the headline would have carried an outdated picture for months.

So what does make an announced campus more likely to be built? Faraday's read is that the signals worth watching are the ones that cost somebody money or legal exposure. A utility filing and a regulator's approval, like the Louisiana vote, commit a third party. A signed financing commits investors. Building permits and site grading show that construction has begun. Orders for transformers and turbines, when disclosed, show that long-lead equipment is reserved. A named tenant with a signed contract, like Anthropic at Project Rainier, shows that someone intends to pay for the space. Each of these is more informative than a larger headline number.

The weaker signals are worth knowing too. A memorandum of understanding with a state, an option on land, a statement that a company is exploring a site, and a gigawatt figure for the ultimate buildout are all early and reversible. They are not meaningless, but they describe what might happen rather than what has been arranged.

The discipline, then, is to grade each campus by the strongest evidence behind it, not by the size of its announcement. A modest campus with a utility contract, a financing and a named tenant is further along than an enormous one with only a press release. Record each new piece of evidence with its date and source, and revisit the grade when it changes.

Hyperion and Abilene will themselves age. By the time you read this, one may have grown and the other may have changed hands or purpose. That is exactly why they are useful examples. The method of grading them does not depend on how their stories end.