Three projects account for all 1,525 megawatts local moratoriums have actually delayed

Add the pieces and the total checks out: 1,525 megawatts held up by local restrictions plus about 800 megawatts in New York gives 2,325 ^[4][6][1]. The denominators are where the argument lives. Roughly 20 gigawatts of planned capacity sits inside a restricted local boundary, and 7.6 percent of it is delayed ^[4][2]. New York's order touches about 1.4 gigawatts and delays 0.8, or 57 percent ^[6][3]. A state permitting freeze leaves no parcel line for a developer to step across.

In Brownsville there was one. Days before the city approved its moratorium, according to SemiAnalysis, the city and Elon Musk "deannexed" 444 acres and brought the project out of the moratorium's jurisdiction ^[14]. The firm's test for whether a restriction counts is narrow by design: it has to apply to a specific project, cover that project's parcel, and block an approval the developer still needs ^[15].

Six percent of the 2027 pipeline sounds like a rounding error ^[4]. It is not, once you notice which six percent. Of the 38 gigawatts SemiAnalysis expects delivered in 2027, 22 are already in vertical construction, which leaves the 16 gigawatts still in planning as the only pool a permit decision can reach, and 2.3 against 16 is one megawatt in seven ^[7][8][5]. The firm says that planning bucket is in majority already financed and doing siteworks ^[8].

The grid side of this is counted in a different unit. SemiAnalysis puts the ERCOT queue pause at perhaps three to four months of administrative delay for base-load projects, and says it means nothing at all for projects that never needed the grid ^[11]. Months, not megawatts, so the two constraints cannot be ranked against each other on this record. The same report counts 75 gigawatts of firm behind-the-meter equipment orders, more of them bound for Texas than any other state ^[12], following a June forecast of behind-the-meter power carrying more than half of new US capacity from 2028 ^[13]. Against that order book, 2.3 gigawatts of moratorium delay is 3 percent ^[6].

One vendor is carrying all of this. The count comes from a subscription model built on real-time and historical satellite imagery of more than 6,000 data centers ^[9], and SemiAnalysis says the forecast has barely moved over the last 12 months despite bi-weekly client updates ^[10]. A model that does not move is either well calibrated or committed to its prior, and from outside you cannot tell which.

The sharper objection is in the definition. Because a restriction only counts when it blocks an approval a named project still needs ^[15], a moratorium that deters a site search before any application is filed never enters the 2.3 ^[3]. Pennsylvania's new permit guardrails and Oregon's freeze on state-land deals sit outside the number too, since SemiAnalysis limited it to the two categories where it can establish a direct project-level impact ^[1][3]. In my view 2.3 gigawatts is the floor of the political cost, and the 16 gigawatts in planning is where the next increment turns up. If projects already in vertical construction start appearing on the delayed list, or the 38-gigawatt figure for 2027 finally moves, that floor was set too low.