Twenty years of sensor networks and dashboards haven’t solved urban congestion, energy waste, or emergency response. The reason is simpler than you’d think — and the fix is finally here.

Twenty years ago, “smart city” was the term that was going to transform urban life. Sensors everywhere. Real-time data. Dashboards that would let city managers see everything happening across millions of square metres of urban infrastructure. Governments invested heavily. Consultancies built entire practices around it. Pilot projects multiplied.

And yet, most cities today still deal with the same structural problems they had in 2005. Traffic congestion. Energy waste. Slow emergency response. Infrastructure maintained reactively rather than proactively.

At the Tele2 IoT Talks Paris event in June 2026, Onur Kasaba, Managing Director of Tele2 IoT, put it plainly after more than 20 years in the industry:

A smart city essentially meant a city with a dashboard. Unfortunately. It improved efficiency to some extent, but it didn’t really move the needle on the fundamental level — because we didn’t have what we have today: the execution capability of artificial intelligence.

Why Data Alone Couldn’t Close the Execution Gap

The original smart city model was built on a sound premise: if you can see what’s happening, you can manage it better. Sensors were deployed across traffic systems, water networks, energy grids, and public spaces. Data flowed into centralised platforms. Dashboards showed everything.

The problem was what happened next. Humans still had to look at the dashboards. Interpret the data. Coordinate responses. Send teams. Wait for results. The decision-to-action loop was too slow to make a meaningful difference in most scenarios. You could see a water leak forming.

But fixing it still took the same number of calls, work orders, and truck rolls it always had. Collecting data was never the bottleneck. Executing on it was.

How AI Turns Data Into Action

Artificial intelligence doesn’t just add more data. It adds the ability to act on data without human intervention at every step — and to act in seconds, not hours.

Emergency services route dynamically based on live traffic and incident data. Energy systems self-optimise hour by hour from actual consumption patterns. Traffic lights adjust in real time to reduce congestion before it builds. Water systems detect anomalies and trigger maintenance workflows before failure occurs.

This isn’t AI as a decision-support tool. It’s AI as an execution layer. The city stops being a passive observer and starts behaving like an active, self-managing system.

The Digital Twin Milestone That Makes 100% Accuracy Real

One of the most significant technology moments at CES 2026 was the demonstration of digital twin capability at 100% accuracy. Nvidia’s Cosmos platform allows any environment — including complex urban infrastructure — to be digitalised at a fidelity that was simply not possible before.

What does that mean in practice? A city can model a proposed infrastructure change, simulate its impact under real-world conditions, and validate the outcome before a single worker arrives on site. The physical result can be tested virtually.

This is why city planners are now projecting up to 60% reductions in energy consumption through AI-guided planning and execution. The simulation is reliable enough to act on.

AI-Native Smart City Infrastructure: What It Looks Like in Practice

The next generation of smart infrastructure in smart city is designed around different principles than the last. It is deliberately distributed rather than centralised — because experience has shown that a single point of failure in a centralised system can take down critical services across an entire city, from traffic management to water metering.

It’s locally hosted where possible — municipalities care about data sovereignty, and cloud costs have risen substantially. And it operates in real time: 5G IoT networks, now widespread across most major cities globally, support decision loops across thousands of connected sensors simultaneously.

A Live Example: Road Accidents Tracked in Real Time

One of the clearest examples of this already working is a live emergency beacon network across Spain, operating on Tele2 IoT connectivity, that tracks road accidents in real time across the country. Emergency services are notified and routed instantly.

It is a relatively narrow use case — but it demonstrates the principle: data collected from IoT devices, processed immediately, triggering action without manual intervention. That loop, applied across water systems, air quality monitoring, energy grids, and traffic management, is what the AI-native city looks like in full operation.

Why Governments Are Paying Serious Attention to Smart City Technology Now

The urgency has a quantitative dimension that gets lost in the technology conversation. Urban populations are at historic highs, on infrastructure designed for population densities from the 1960s. Operational inefficiencies in city systems now cost economies up to 4% of GDP. And cost of doing nothing is rising faster than the cost of acting.

At the same time, the technology has become accessible: sensors are cheap, AI is no longer an enterprise-budget item, and networks are fast enough. The United Nations has launched a formal governance framework for the “Cityverse” — the AI-augmented, digitally twinned version of urban life moving from concept to policy agenda. Over 200 government delegates attended CES 2026 to engage with AI city infrastructure — not as tourism, but to work out what they need to regulate, procure, and build.

Cities That Adopt Smart City Tech First Will Be Very Hard to Catch

Smart city was never a bad idea — it was a good idea deployed before the execution layer existed. That execution layer now exists.

Cities and infrastructure teams that recognise that distinction early — and build AI-native IoT systems rather than layering AI on top of existing dashboards — will create compounding advantages in energy cost, public safety, service quality, and operational efficiency that slower movers will find very difficult to close.

Key Takeaways

The smart city concept wasn’t wrong – it was too early. The sensor data has always been there. What’s new is the ability to act on it autonomously, in real time, without a human in the middle of every decision loop. That shift changes what city infrastructure can actually do.

AI-native cities are a connectivity design problem as much as a software one. Distributed architecture, real-time sensor loops, digital twins at 100% accuracy – none of it functions without a network that can carry high volumes of data reliably, at low latency, across thousands of endpoints simultaneously. Smart city AI is only as fast as the connectivity underneath it.

Data volumes are scaling up, not stabilising. Every new sensor, every autonomous decision loop, every digital twin added to a city system increases the load the network needs to handle. Connectivity capacity and reliability are active design requirements for every smart city project, not a solved baseline.

The infrastructure already exists for early applications. The live emergency beacon network across Spain – running on Tele2 IoT connectivity, tracking road accidents in real time and routing emergency services automatically – is a working example of what the AI-native city model looks like in practice. It is narrow in scope. The underlying approach is not.

As smart city and IoT deployments move to AI-driven execution, connectivity becomes the foundation. Curious how Tele2 IoT can support your deployment? Contact us.