A city converting its street lighting to LED has a second decision to make after the fixture: how the controllers talk to the management platform. Mesh and cellular are the two architectures in common use, and both do the job. They diverge on the two things that decide what the network costs over the life of a fixture.
The first is obsolescence. Mobile generations are retired on the operator’s schedule, not the city’s. 2G and 3G are gone; each generation that follows has a working life measured in single-digit years. A streetlight fixture does not — which means a cellular-based controller has to be planned for replacement more than once before the fixture it sits on wears out. That is a capital cost with a schedule nobody in the city controls.
The second is the recurring bill. Cellular runs on licensed spectrum, and the operator charges per connected device. At the scale of a municipal lighting network that becomes a meaningful annual line item, and it is set by a party the city does not negotiate with from a position of strength. A mesh network runs on unlicensed spectrum between the poles themselves, so adding a node adds no subscription. Lighting control is not a data-hungry application — schedules, energy readings and fault alerts are small messages — so the bandwidth a mesh provides is not the constraint.
The same argument extends past lighting. Once a mesh reaches every pole, sensors added to it cost nothing per месяц to keep connected: environmental and air-quality monitoring, water or snow levels, motion and temperature. That is the practical case for treating the lighting network as infrastructure rather than as a single-purpose system.