Compute Without Residents: Certified remote operations for data centres in polar, remote and orbital locations

AI compute is moving to places where few or no people live. Examples include a proposed 1.5 GW campus at Kangerlussuaq in Greenland, cold-climate sites across the Nordic and Arctic regions, and satellites carrying AI accelerators in low Earth orbit. These sites offer cooling, land and energy, but they lack a resident skilled workforce, and orbital sites have no physical access at all. They will be run largely from elsewhere. This paper argues that the critical capability for such infrastructure is certified remote operation: privileged access granted to named, certified people and AI agents, inside documented boundaries, with AI-generated operational advice challenged and every action recorded in tamper-evident form. It reviews precedents from unmanned offshore platforms, remotely operated mines and lights-out data centres. It identifies the risks specific to remote compute, among them connectivity, insider and supply-chain threats, AI operations agents, jurisdiction and local consent. It proposes a Certified Remote Operator track built on the three existing GCTI certifications (Clean Engagement, Reflection Skills, and Verification and Provenance), with follow-the-sun coverage across a multi-time-zone network. It closes with advantages, disadvantages and a pilot. The author leads the entities described; the paper is a proposal and test plan. Declared interests: the author is the founder of GCTI, Managing Member of E.M.B.O.A LLC and Representative Member designate of Emboa Japan GK (in formation). Discussion paper, version 1.0.