Why smart cities fail
🤖 This essay was generated with AI from the sources linked throughout. It has not had a complete human review — treat it as a starting point and follow the sources. How this works.
Google's city of the future died citing a pandemic — after two years of privacy revolt. The best-documented smart-city failures follow six distinct patterns, and almost none of them are about technology.
On 7 May 2020, Alphabet's Sidewalk Labs walked away from its city of the future on Toronto's waterfront, citing "unprecedented economic uncertainty" from the pandemic. The explanation was true and almost beside the point: by then the project had survived two years of privacy revolt, a #BlockSidewalk campaign and a data-governance proposal so contested it became an academic case study. The most instructive thing about smart-city failures is that the technology almost never breaks. Something else does — and across the documented cases in this atlas, it breaks in one of six recognisable ways.
Six ways a smart city dies
| Pattern | What it looks like | Documented cases |
|---|---|---|
| 1. Never operated | Hardware installed, nobody owns the running of it | Dhaka, Algiers |
| 2. Trust collapse | Data and surveillance questions answered too late | Sidewalk Toronto, San Diego streetlights |
| 3. Quiet target burial | Ambition shrinks; nobody announces the funeral | Masdar, Songdo |
| 4. Pilot purgatory | The demo works; the handover to normal operation never comes | German AV shuttles; the federal transfer gap |
| 5. Vendor exit | The platform a city depends on stops being a business | Cisco Kinetic for Cities |
| 6. No business model | It works, then runs out of money — no users, or no way to pay for them | Whim, Bycyklen, Metaverse Seoul |
Pattern 1: installed, never switched on
The purest failures in this atlas involve no failed technology at all. Dhaka automated its traffic signals and the police kept directing traffic by hand — the machines stood there, ignored, for want of training and maintenance ownership. (A fresh attempt at semi-automated signals began at seven intersections in August 2025; whether anyone runs it is, once again, the whole question.) Algiers contracted 500 camera-linked intersections in 2017; 22 were ever wired, none fully activated, because the security services refused to authorise the cameras and the Spanish partner eventually left. In both cases the procurement succeeded and the institution around it didn't. This is the pattern our fact-checkers now probe first on every entry: not "does it work?" but "who runs it?"
Pattern 2: the trust collapse
Sidewalk Toronto is the canonical case, and its file is thicker than the pandemic explanation suggests. The frictions were documented long before Covid: the #BlockSidewalk campaign formed in February 2019, and the company's flagship governance idea — an independent "urban data trust" to hold the neighbourhood's data — drew enough criticism that peer-reviewed work now analyses it as a failure in civic accountability. When the end came, the company — having committed roughly $50 million to planning — cited economics; MIT Technology Review summed up the longer arc — years of public controversy first. The sequel is telling: Waterfront Toronto relaunched the site in 2022 as "Quayside 2.0," a zero-carbon, mass-timber housing plan with no sensor-city framing at all; construction is expected to get underway in 2026, with the first residents planned for 2031.
San Diego ran the same pattern at streetlight scale: thousands of camera-equipped "smart streetlights", sold in 2016 as an energy project, quietly became a police video resource — and when that surfaced, the mayor ordered the system off in September 2020. The grimly comic detail: the city initially couldn't fully switch the cameras off, because they shared power with the lights. What makes San Diego more than a cautionary tale is the third act, covered below.
Pattern 3: the quiet target burial
Masdar City launched in 2008 as the world's first zero-carbon city, a roughly $22-billion plan. After the financial crisis, zero-carbon became "carbon-neutral", on-site energy became partly off-site, the deadline slid from 2016 toward 2030 — and only a fraction has been built. Songdo tells a milder version: about 212,000 residents against a planned 265,000, with commercial space slow to fill. Neither project "failed" on any single day, which is exactly the problem: scaled-back ambitions rarely get the honest post-mortem a cancelled project forces. Research projects bury targets the same way: Amsterdam's Roboat spent five MIT-backed years (2016–2021) building full-scale driverless canal boats for passenger transport, waste collection and water sensing; the spin-off that survived it now sells crew-assistance automation for a single GVB ferry, not the autonomous water taxis the vision promised. (Our showcase-or-use-case piece covers why cities keep building these — and what showcases legitimately buy.)
Pattern 4: pilot purgatory
Documented at length in the research: pilots that never become services, because scaling needs different partners, money and governance than demonstrating. Van Winden and van den Buuse's study of Amsterdam's own pilots states the base rate plainly — many smart-city projects fade out after the pilot stage and fail to generate scalable solutions. Germany offers the measured version: its €820M programme is funding 73 model-city projects, and the federal institute BBSR's dedicated "staged plan" project focuses on scaling proven solutions so that even non-funded municipalities can adopt them. The country's autonomous-shuttle scene — some twenty pilots completed or ongoing in Munich, Hamburg and Ulm alone, whose own association president pleads for regular service instead of ever more test labs — is the pattern on wheels.
Pattern 5: the vendor walks away
In December 2020 Cisco discontinued Kinetic for Cities, the flagship platform of its smart-city business, leaving customer cities to wind down with support timelines rather than a successor product — the formal end-of-life notice offered "no replacement". The lesson is structural, not about Cisco: a city's operating system can be someone else's discontinued product line. Cities that build on open standards and own their data pipelines — the model of our LoRaWAN towns like Bad Nauheim — carry less of this risk than cities that bought a dashboard.
Pattern 6: the money never added up
The fifth pattern is a vendor choosing to walk away; the sixth is quieter and, lately, more common — the service itself never had a way to pay for its own fifth year. Sometimes the users simply never come. Copenhagen, the Capital Region of Denmark and Hitachi launched the City Data Exchange in 2016 as the world's first marketplace to buy and sell urban data; it never attracted enough providers or buyers to become a working two-sided market, and the partners closed it in 2018 citing weak financial sustainability. Seoul's city-run Metaverse Seoul, launched in January 2023 as the world's first local-government metaverse public-service platform, drew only a few hundred visits a day at its best and was shut down in October 2024 — after about 21 months — as the metaverse hype it rode gave way to AI. Both were, in the end, a technology in search of a demand that never arrived.
The harder version is when the users do come and the economics still won't close. Copenhagen's Bycyklen — one of the first bike-shares with a tablet on every handlebar — grew to about 933,000 rides in a year, then its operator filed for bankruptcy in December 2022 the moment Copenhagen, Frederiksberg, Rødovre and the national rail operator withdrew their subsidies; the service shut down and was never replaced. Helsinki's Whim, widely credited as the world's first commercial mobility-as-a-service app, is the sharpest case of all: an operator-commissioned Ramboll evaluation of its first year found Whim subscribers took 73% of their trips by public transport against 48% for the average resident — a real modal shift, exactly what the policy wanted — and its parent MaaS Global still went bankrupt in March 2024, with the acquirer declining to relaunch it in Helsinki. A smart service can work, be used, move the needle, and still die because no one built it a way to pay for itself. So the atlas's operating test has a second half: not only who runs it in year five, but who pays for it.
— Ben Green, The Smart Enough City (MIT Press, 2019) — the "tech goggles" critique behind most of these patterns
The honest counterweights
Three findings keep this from becoming a doom piece. First, failure is the base rate, not the exception: the Standish Group reports that fewer than 10% of large IT projects fully succeed — figures whose methodology is itself contested, but which suggest smart-city flops are ordinary public-technology failure, not a special curse. Second, not every futuristic city is vaporware: Toyota's Woven City completed its first construction phase in 2024 and welcomed its first residents in September 2025 — albeit hand-picked employees in a corporate test environment (a few hundred residents planned for phase one, outside visitors from fiscal 2026), not an open city. Third, and most usefully: processed failure becomes tuition. Stockholm converted opposition that once ran as high as 70% into majority support by running an honest trial and letting people feel the result. Barcelona absorbed its "No Superilla" protests and iterated the superblock model into the Green Axes plan — though that scale-up has since stalled at 4 of roughly 21 planned streets, a reminder that "processed" doesn't always mean "finished." And San Diego's streetlights came back in December 2023 — under a surveillance ordinance, with annual council re-approval — that the original rollout never had. The failure produced the governance; the governance made the technology viable.
What actually separates the graveyard from the atlas
Reading the six patterns together, the common thread is that every one is an institutional failure wearing a technological costume: no operator (1), no trust framework (2), no accountability for targets (3), no handover plan (4), no exit strategy from a vendor (5), no way to pay for it (6). Which is why the single test this atlas applies to every documented entry does most of the diagnostic work on its own: who operates it — and pays for it — in year five, and who answers when it stops working? Cities that can answer, like Sydney, whose SCATS signal system has had the same public road authority behind it since 1975, rarely appear in articles like this one. The same question, asked of who holds a system's data, decisions and exit rights when a third party builds it, is the subject of a companion piece on who really owns your smart city.
Projects mentioned
Sources
43 checkable claims in this essay are annotated — select any dotted-underlined passage to see exactly what its source says. The evidence links were prepared with AI and are not yet individually verified — follow the sources.
- Why we're no longer pursuing the Quayside project — and what's next for Sidewalk Labs open source ↗ archived copy
- Alphabet's Sidewalk Labs to turn Toronto area into a model smart city open source ↗ archived copy
- Critics Vow to Block Sidewalk Labs' Controversial Smart City in Toronto open source ↗ archived copy
- Is civic data governance the key to democratic smart cities? The role of the urban data trust in Sidewalk Toronto open source ↗ archived copy
- Toronto wants to kill the smart city forever open source ↗ archived copy
- Adjaye Associates designs mass-timber building covered in plants for Toronto's waterfront (Quayside 2.0 unveiling) open source ↗
- Quayside — project page open source ↗ archived copy
- Citing Privacy Concerns, Faulconer Shuts Down 'Smart Streetlights' open source ↗ archived copy
- San Diego Can't Actually Turn Its Smart Streetlights Off open source ↗ archived copy
- Smart Streetlight, License Plate Reader Technology Helping Solve Crimes and Keep San Diegans Safe open source ↗ archived copy
- Green dreams: Abu Dhabi rethinks car-free, waste-free city open source ↗
- Rethinking the low-carbon city: eco-urbanism after Masdar open source ↗ archived copy
- Songdo International City at the Crossroads (extended abstract) open source ↗ archived copy
- The self-driving boats of Amsterdam: Roboat completes its research program open source ↗ archived copy
- Next phase for Roboat: to boost future-proof urban mobility as a spin-off company open source ↗ archived copy
- Smart City Pilot Projects: Exploring the Dimensions and Conditions of Scaling Up open source ↗ archived copy
- Modellprojekte Smart Cities — über uns open source ↗ archived copy
- Stufenplan Smarte Städte und Regionen open source ↗
- Rund 15 Modellprojekte für autonome Shuttles in Deutschland open source ↗ archived copy
- End-of-Sale and End-of-Life Announcement for the Cisco Kinetic for Cities Enterprise SKUs open source ↗
- Cisco explains its smart city software exit open source ↗ archived copy
- City Data Exchange open source ↗ archived copy
- Cities want more from their data, including money open source ↗ archived copy
- Seoul gov't launches world's 1st public services platform in metaverse open source ↗ archived copy
- 메타버스 서울, 서비스 종료 (Metaverse Seoul service terminated) open source ↗
- Kim Hyun-gi, Chairman of Seoul Metropolitan Council: Metaverse Seoul daily visitors down to 273 open source ↗
- Danish Electric Bike-Sharing Dodges Failure open source ↗ archived copy
- Bikeshare Danmark A/S indgiver konkursbegæring (bankruptcy notice) open source ↗ archived copy
- 'Netflix of Transportation' App Reduces Car Use and Boosts Bike and Bus Use, Finds MaaS Data Crunch open source ↗ archived copy
- MaaS Global declares bankruptcy open source ↗ archived copy
- umob acquires MaaS Global, Finnish creator of mobility app Whim open source ↗ archived copy
- The Smart Enough City — Ben Green open source ↗ archived copy
- CHAOS Report open source ↗ archived copy
- Toyota Woven City Officially Launches as a Test Course for the Future of Mobility open source ↗ archived copy
- Woven City Phase 1 construction completed; launch planned for 2025 open source ↗ archived copy
- The Stockholm congestion charges: an overview (CTS Working Paper 2014:7) open source ↗ archived copy
- Green Axes Eixample — atlas entry open source ↗
- SCATS — About Us open source ↗ archived copy
- Bangladesh: Clean Air and Sustainable Environment Project — completion report open source ↗ archived copy
- Automated traffic signal: Money thrown down the drain open source ↗ archived copy
- Pilot project shows promise, delays stall wider rollout open source ↗
- Feux tricolores à Alger : Chiali explique l'échec et plaide pour des autoroutes à étages open source ↗ archived copy
- Algiers Intelligent Traffic Management — atlas entry open source ↗
Frequently asked
Why do smart city projects fail?
The documented cases cluster into six patterns: hardware never operated, trust collapse over data and surveillance, quietly buried targets, pilots that never scale, vendor exits, and services with no viable business model (Whim, Bycyklen, Metaverse Seoul). Technology itself is rarely the cause — operating models, governance, trust and money are.
What happened to Sidewalk Toronto?
Cancelled 7 May 2020, officially over pandemic economics — after two years of documented privacy backlash, the #BlockSidewalk campaign and a failed "urban data trust" proposal. The site returned in 2022 as Quayside 2.0, an explicitly non-smart-city housing plan.
Do smart cities fail more than other IT projects?
Probably not: Standish Group data (methodologically contested, but directionally echoed elsewhere) shows large IT projects fully succeed less than 10% of the time. Smart-city failure looks like ordinary large-project failure — with cameras.
Which smart city services shut down for lack of money?
Several failed on economics, not technology: Helsinki's Whim (parent MaaS Global went bankrupt in March 2024, despite a genuine modal shift to transit), Copenhagen's Bycyklen bike-share (bankrupt December 2022 after subsidies were withdrawn), Copenhagen's City Data Exchange (closed 2018 for lack of a market), and Seoul's Metaverse Seoul (shut down October 2024 after minimal use). Each worked; none could pay for itself.
Can a failed project still be worth it?
If the failure is processed, yes: San Diego's cameras returned under new surveillance ordinances, Stockholm's contested trial became a template, Barcelona iterated past its protests. The difference between failure and tuition is whether governance changes afterwards.