Crusoe's Energy Procurement Advantage

Crusoe is trying to turn power procurement from a pass through cost into a core product advantage. Instead of renting space in grid connected facilities and paying whatever local wholesale and utility rates happen to be, it has historically put machines next to energy that other operators cannot easily use, like flared gas, and is now extending that logic into dedicated on site generation and remote renewable campuses, which can make power cheaper and more available when GPU demand spikes.1 2 3 4 5

• The original version of this model was very literal. Crusoe took natural gas that oil producers would otherwise flare, converted it into electricity on site, and used that to run compute. Internal research ties that setup to power costs as low as about 1/13th of standard cost in some deployments, because the gas is hard to transport and often a liability for the producer.2 3
• That advantage matters because power is one of the biggest differences between Crusoe and rivals like CoreWeave and Lambda. Crusoe builds or owns more of the energy stack itself, while co location models can open faster with less upfront capital but carry ongoing lease expense. Internal research notes lease payments can be about 60% of opex for those models, which makes cheap power more valuable over time than it first appears.2 4
• Crusoe is now scaling the same playbook from edge sites to giant campuses. It signed for 29 GE Vernova aeroderivative turbines expected to supply nearly 1 GW, and a separate 13 unit PROENERGY fleet totaling 650 MW for hyperscale projects. In parallel, its atNorth footprint in Iceland reached about 57 MW after a 24 MW expansion, pairing low cost power with high end Blackwell and GB200 capacity.1 4 5 6

The next step is clear. As AI infrastructure shifts from finding spare GPUs to securing firm power at scale, operators that can bundle land, generation, cooling, and compute into one deliverable should capture a growing share of large cluster buildouts. Crusoe is moving from being a cheap power specialist to being a power first hyperscale AI infrastructure company.1 4 5