Computing infrastructure providers, data processing, web hosting, and related services

The demand is the least ambiguous on the entire list, and it is the cause of the opportunity the grid study at 221121 examines. The problem is what it costs to serve. Amazon Web Services earned US$128.7B in 2025, from US$107.6B in 2024, inside a company that spent US$131.8B on property and equipment in the same year [A] — a capital programme larger than the segment's whole revenue. Equinix is the closer comparable, because colocation is what a non-hyperscaler could plausibly sell, and it tells the same story: $9.217B of 2025 revenue on roughly $4.3B of capital expenditure, about forty-seven cents of capex for every revenue dollar, with $3.655–4.155B guided again for 2026 [A]. In Canada the unit of entry is now quoted the same way — Innovation, Science and Economic Development Canada describes Bell's Saskatchewan AI hub as up to $52.5B of capital investment for up to 900 MW [A]. Against that, the 1,163 Canadian establishments in this code are small: 538 employ fewer than five people [A]. They are not building capacity, they are managed-hosting, web-hosting and data-processing firms reselling someone else's. An entrant cannot buy its way to a cost position here, because the cost position is set by counterparties spending more in one year than an entrant could raise in a lifetime, and the scarce input — interconnected grid power — is allocated years ahead to whoever is already in the queue. What is left enterable sits on top of that infrastructure: sovereign or regulated hosting, managed migration, specialised workloads. That is a labour business wearing an infrastructure name. The grid study at 221121 is how a small operator sells into this wave without building it.