From Tenant to Principal: The Structural Inversion
Hold this thesis: the buyers of compute have seized the means of commissioning it. Labs were tenants — line items in a landlord's forecast. Now they are principals, reaching upstream to own the scarce resource: energized land on a schedule that matches a model roadmap. The trigger is lead time. A tenant inherits the landlord's queue; a principal owns its own. When being a year late to a capability is fatal, owning the schedule beats owning the software. The consequence is a balance-sheet mutation — software companies taking on utility-scale capital, power contracts, and construction risk. Read every lab infrastructure move through this frame: it is not demand landing in a stack, it is a new principal manufacturing a stack that did not exist.
Knowledge check
What single constraint most directly explains why frontier labs moved from renting capacity to commissioning it themselves?
Show the answer
Lead time on the physical inputs — power, interconnection, and long-lead equipment like turbines and transformers. A tenant who waits for a landlord to clear those queues inherits the landlord's timeline; a principal who places the orders and secures the interconnection position directly converts a passive wait into an owned schedule. In a race where being a year late to a capability is fatal, controlling that schedule is worth more than any software advantage, so labs reached upstream to seize the constraint.