Tucson's utility vote led AWS to abandon its 290-acre Project Blue data center

The instructive part of Project Blue is where in the sequence it died. Land can be optioned and cleared quietly; water and utilities cannot, because they run through a body that answers to voters, and in this case those voters were in a drought-stricken district ^[1]. Vox describes the site as already under construction when the council voted ^[2]. A developer who secures acreage first and the hookup later has converted local politics into sunk cost ^[3].

The systemic version of the same constraint is the queue. Vox puts the wait to connect to the power grid at 12 years, alongside what it calls an explosive populist backlash ^[5]. Twelve years counted from a decision made now runs past the end of this decade, which is the window inside which SpaceX says it might assemble its first data center in orbit ^[15]. For anyone buying capacity, that arithmetic makes the scarce input an interconnection slot and a water source whose owner is under no obligation to sell.

The board-deck version says orbit routes around all of it, and it is incomplete in one specific way. On the aerospace engineer Andrew McCalip's analysis, a one-gigawatt data center in space costs about $51.5bn against $16bn on the ground ^[9]. Divide the two and orbit runs at 3.2 times terrestrial cost, a premium of $35.5bn for every gigawatt ^[13]. Jason Aspiotis, who leads an orbital data center startup, told Vox the physics is not in dispute and the engineering is solvable with time and money, but that "the gray area is, well, what about the economics?" ^[10]

Scale the figure the industry has attached to the idea and the gap becomes easier to hold in view. SpaceX's eventual ambition is 100 gigawatts of new solar energy harvested for AI each year, which Vox calls a very iffy promise and equates to one-fifth of all US electricity consumption last year ^[7]. At today's orbital cost basis, and holding cost per gigawatt flat in a way nobody actually claims it would, that is roughly $5.15 trillion of capital ^[14]. Attached to engineering that Vox itself flags as unprecedented ^[16], a number that size is better read as a statement about how blocked the ground route feels than as a construction budget.

One council in one desert district is not, on its own, a national siting constraint: the source offers no tally of comparable denials, so the breadth of the veto is not established here. The answer is in who moved. Amazon Web Services withdrew rather than contest the outcome ^[3], and Blue Origin's filing for tens of thousands of possible orbital data centers arrived in the same period ^[4].

The piece leaves two questions, and only one of them is decidable this quarter. The near question is whether water and interconnection terms are committed in writing before land is committed in concrete, and which jurisdictions will do that. The 2030s question is whether launch costs fall far enough to close a 3.2x premium ^[13], which the experts Vox spoke to treat as the open item now that feasibility is not ^[12]. Choosing a water-constrained site this year on the assumption that the utility question resolves itself later is a decision to learn at the end of the process what Project Blue learned ^[3].