City connectivity is the layer underneath every other one: the fibre, the poles and the radio networks carrying the sensor reading, the camera feed and the train's signal. The question that decides its fate is never how fast it is — it is who owns the pipe, and what they may do with it.
This guide draws on 31 verified projects in 23 cities across 16 countries. Connectivity is the layer cities most often build themselves and keep for decades — so the ownership decision taken at the start still governs what is possible twenty years later.
Four models recur, and they are not interchangeable. The city owns the cable and sells nothing: Stockholm's Stokab has leased dark — unlit — fibre on equal, non-discriminatory terms to any operator since 1994, reaching over 90% of regional households, with 100+ operators on it; Cape Town does the same smaller. The municipal utility becomes the ISP: at Chattanooga's EPB, the same strands run about 1,200 automated switches that reroute power around faults and sell households gigabit internet. A neutral host builds once for everyone: Smart Docklands became Dublin's first neutral-host 5G testbed in 2017, and London let a 20-year concession reuse Underground ducts. Or a concessionaire owns it and sells attention: LinkNYC gives away gigabit Wi-Fi from kiosks paid for by on-screen advertising, not taxpayers.
| City | Model | What the record shows |
|---|---|---|
| Stockholm | Municipal dark fibre, since 1994 | Self-sustaining on lease fees: SEK 844m net sales, SEK 246m pre-tax profit in 2024 |
| Chattanooga | Utility fibre + retail broadband | US$169m revenue bonds plus a federal grant; a 2024 study puts net benefit at $5.3bn |
| Cape Town | Open-access municipal fibre | 848 km, 400+ Wi-Fi zones serving roughly 200,000 users a month |
| Medellín | Public Wi-Fi and inclusion, since 2007 | 411 free hotspots, 40m+ connections in 2025 (city figures) |
| Brazzaville | Donor-financed national data centre | €66.55m — €52.47m from the African Development Bank, €14.50m from the state |
Less than the headline numbers suggest — and the strongest figure is the one nobody involved paid for. Chattanooga's $5.3 billion in net community benefit since 2011 comes from a 2024 University of Tennessee at Chattanooga study: genuinely independent, and unusual here. Stokab's much-quoted return of roughly SEK 16 billion against SEK 5.4 billion invested comes from a 2012 study commissioned by Stokab — its audited accounts are the harder evidence.
Two entries carry no evidence grade at all. Zurich's ewz.LoRaWAN, run by the city utility since 2018, has published no outcome data — its documented uses are a 185-space parking pilot and 16 sensors on chairs measuring how two squares get used. That is an honest picture of a shared LoRaWAN early on, as is Vienna's. Shenzhen's 46,480 base stations by August 2020 are verifiable; the "world's first" claim is not.
Because consolidation is the business case. Shanghai removed about 40,000 poles between 2018 and 2020 — a 60% reduction — and raised 16,000 multifunctional poles absorbing roughly 19,000 sets of equipment, under a 2019 rule sending 5G base stations onto light poles rather than new masts. Seoul's S-Pole merges streetlight, traffic signal, CCTV, public Wi-Fi and sensors into one unit; a study across 1,027 poles put installation costs about 23% below separate fixtures. Tokyo's poles add 5G, Wi-Fi 6 and AI cameras, and Washington D.C. hangs 239 Wi-Fi access points off a 75,000-streetlight retrofit.
Say it plainly: the mast giving you free Wi-Fi is usually the mast watching you. Mexico City's C5 shows where that ends: 113,814 cameras in March 2026, and a Data Cívica and R3D finding that about 60% of crimes happen within 200 m of one, while police use footage in under 1% of investigations. (See street lighting and sensors and IoT.)
Algiers is the clearest case: a 2017 contract to wire 500 intersections delivered 22 of the 200 first-phase junctions, stalled on unresolved fibre-optic links and a national-security refusal to switch the cameras on, and lost its Spanish partner. Officials called it a failure by 2021. Two rail entries show the subtler version: on San Francisco's Muni Metro, a train that loses its data link must run manually at half speed, slowing every train behind it; New York's replacement signalling costs roughly $25m per track mile, on an MTA projection of nearly 50 years.
The practical takeaway: connectivity is the longest-lived thing a city buys here, and almost the only layer it still builds itself. Stockholm's 1994 decision not to become an ISP is still paying; Algiers' fibre is still dark. Decide the ownership model first — everything downstream, including who may look through the camera on the same pole, follows from that one choice. See our essay on who owns your smart city.
Dark fibre is unlit cable — the strand without the electronics that light it — leased to operators who bring their own equipment. Stockholm's Stokab has leased it on equal, non-discriminatory terms to any operator since 1994 and has never sold retail internet itself, which lets a competitive market of over 100 operators share one physical network instead of digging separate ones. Barcelona's 22@ district took the same approach from 2000, laying dark fibre under about 37 km of rebuilt streets.
Stokab is self-sustaining on lease fees rather than subsidy, reporting SEK 844m in net sales and SEK 246m profit before tax in 2024. Chattanooga's EPB financed its network with US$169m of municipal revenue bonds plus a federal smart-grid grant, and a 2024 University of Tennessee at Chattanooga study valued the combined fibre and smart grid at $5.3 billion in net community benefit for Hamilton County since 2011. Note the provenance: the Chattanooga study is independent, while Stokab's often-quoted SEK 16 billion socio-economic return came from a study Stokab itself commissioned.
Usually, yes — that is the business case. Seoul's S-Pole merges streetlight, traffic signal, CCTV, public Wi-Fi and IoT sensors into a single unit, and a city study across 1,027 poles found installation costs roughly 23% lower than mounting the fixtures separately. Tokyo's smart poles carry 5G, Wi-Fi 6 and AI cameras; Shanghai consolidated lighting, signals, signage, public-security and traffic cameras and communications equipment onto one comprehensive pole. A city procuring public Wi-Fi on street furniture is generally also procuring surveillance infrastructure, and should govern it as such.
Shared radio infrastructure that any operator can use, rather than each building its own. Dublin's Smart Docklands became the city's first neutral-host 5G testbed in 2017, a city–telecom partnership our record describes as among the first of its kind worldwide rather than a national first. London applies a related idea at scale by letting a 20-year telecoms concession reuse Underground ducts to build a fibre backbone alongside mobile coverage.