The last mile is an arithmetic problem, not a technology problem
Rural broadband programmes keep rediscovering the same cost curve. The density threshold where private deployment stops has barely moved in twenty years.
Every few years a new round of public money is directed at rural broadband, and every round runs into the same wall in roughly the same place. The wall is not technical. Fibre is well understood and the equipment is commoditised. The wall is cost per passing against households per mile.
The cost curve
Building fibre along a road costs a fairly predictable amount per mile, dominated by labour, permitting and whether the route can be aerial on existing poles or has to be buried. Buried rural routes through ledge are dramatically more expensive than aerial ones, which is why terrain explains more cost variation than distance does.
Divide that per-mile figure by the number of premises the mile passes and you get cost per passing. At thirty homes per mile the number is modest and a commercial operator will build it unprompted. At three homes per mile it is an order of magnitude higher. At one home per mile, which describes a great many rural road segments, no subscription revenue at any plausible price recovers the build.
The subsidy therefore has to cover the gap, and the gap grows non-linearly exactly where the need is greatest.
Why the maps keep being wrong
Coverage data has historically been collected by asking providers which areas they serve, at a geographic granularity coarse enough that a single served premise could mark a whole block group as covered. This systematically overstated coverage and, because funding eligibility depended on being unserved, systematically excluded the places that needed money.
Granularity has improved substantially with premise-level reporting and challenge processes. The residual problem is that advertised speed is not delivered speed, and a location marked as served at a given tier may be served at that tier only at three in the morning.
The ownership models
Incumbent subsidy. Fastest to deploy, because the operator already has crews, poles and billing. The recurring criticism is that public money funds an asset the public does not own, and that service standards after the build are whatever the market supports, which in a monopoly segment is not much.
Municipal or district ownership. The asset stays public. Governance is genuinely hard: a small authority needs engineering, finance and operations capability it does not naturally possess, and the failures in this model are nearly always operational rather than financial.
Open-access middle mile. Public money builds the backbone and any retail provider can use it. This is the most elegant arrangement and the slowest, because it only works if enough retail providers actually show up, and in thin markets they often do not.
What determines whether a build finishes
Two unglamorous things. Pole attachment. Access to existing poles, the timeline for make-ready work and who pays for it will delay a project more reliably than anything else, and it is a regulatory question rather than a construction one. Take rate. A build financed on an assumed sixty per cent subscription rate that achieves forty per cent is in trouble regardless of how well it was constructed, and take rate depends on retail price, which depends on the operating model chosen years earlier.
Programmes that treat the build as the finish line consistently underperform the ones that budget for five years of operation.