Patagonia's 3,620-megawatt AI pitch starts as 460 megawatts still hunting financing

Supply-advantaged is the bucket Cryptopolitan, reading McKinsey, drops Patagonia into: a market that sells electrons to workloads willing to sit a long way from their users, rather than one selling proximity to an existing digital ecosystem, and on that reading the job is not to displace Northern Virginia, London or Singapore but to take the power-hungry tail ^[6]. The pitch is coherent, but the arithmetic is where it gets interesting.

Getting Pampa Energia's Loma de la Lata site to its stated 500 MW needs nearly $900mn of further electrical infrastructure, which is $1.8mn per megawatt for the power side before anyone buys a building or a rack ^[7][12], and the 20-40 MW pilot under discussion is between 4% and 8% of that capacity, the size you build when you are buying an option rather than taking a position ^[7][13]. FlexDomes' $1.4bn against a 120 MW opening phase is $11.7mn per megawatt ^[8][14], about six and a half times the Pampa figure ^[15], because one number is grid kit and the other is a campus with the grid kit inside it.

Add the two disclosed asks and Patagonia is hunting roughly $2.3bn ^[16]. Set that against PwC's forecast of $800bn of global data-centre spending in 2026 and the whole corridor is chasing 0.29% of a single year's capex ^[5][17]; PwC's path from that $800bn to $1.8tn by mid-century is a 2.25x rise, with $31.6tn cumulative to 2050 ^[5][18]. Patagonia does not need the trend to hold. It needs one signature.

Green Capital's answer to the transmission problem is to route around it, leasing 501 square miles south of Trelew for wind and solar so the project leans less on the national grid ^[9], which pencils out at about 6 MW of eventual ambition per square mile of land ^[20] and relocates the binding constraint from a transmission queue to turbine and panel capex on its own balance sheet.

The anchor cited for the corridor is the letter of intent OpenAI and Sur Energy signed in October 2025 for the first Stargate in Latin America, which Argentine authorities size at 500 MW ^[11]; no capital figure for it is on the record ^[23]. Line that up against the first phases actually described and the three named projects together open at 440 to 460 MW, less than the single anchor, while their stated ceilings sum to at least 3,620 MW ^[21][22].

There are three ways this breaks. The LOI converts into a signed contract and the 500 MW drags fiber, transmission and the smaller developers behind it; or financing stalls, the 20-40 MW pilot is the only thing that gets poured, and Patagonia turns out to be a stranded-gas power story wearing a compute label; or Green Capital's self-generation works, transmission stops mattering, and the constraint becomes how fast turbines arrive. Cryptopolitan's own summary concedes the point that most of this is proposals, and that financing, fiber, transmission and signed customers decide ^[3].

The reading the numbers support is the middle one, priced at pilot scale. What would break it is specific and checkable: a multi-year offtake with a named counterparty and a committed fiber route, at which point the $900mn and the $1.4bn stop being asks and start being capex ^[7][8].