Everyone Buried Nokia Too Early

The company that lost the smartphone war may have quietly positioned itself for the AI infrastructure war.

Here is a number that should not make sense. The company most people remember as smartphone-era roadkill made about €2 billion in operating profit last year, sits on roughly 20,000 patent families, and just cashed a $1 billion cheque from one of the most valuable companies on Earth.

That company is Nokia. Yes, that Nokia.

When Finland owned the world’s pockets

Nokia did not start in phones. It began as a Finnish paper mill in the 1860s and passed through rubber, cables and industrial gear before it ever found mobile telephony. By 2007 it was an empire. It booked €51.1 billion in revenue, close to €8 billion in operating profit and more than €7.2 billion in bottom-line profit, on roughly 40% of the world’s handset market. Two in every five phones sold on Earth were Nokias.

It did not miss the smartphone. It missed the new battlefield.

The lazy story is that Nokia never saw smartphones coming. It did. It had Symbian, the Communicator, the N-series, email, maps, cameras and early touchscreens. What it lacked was not technology. It was clarity about where the power had moved.

Nokia still treated the device as the product. Apple treated it as the door into an ecosystem. Google gave Android away and signed up an army of hardware makers. Developers followed the users, users followed the apps, and the profit followed the platform. Battery life still mattered, but software mattered more. Distribution still mattered, but developers mattered more. Nokia kept polishing the shell after the soul had moved somewhere else. By 2012 its phone arm was losing about €1.1 billion a year. Microsoft then bought the handset business for roughly €5.4 billion, and within eighteen months wrote down about $7.6 billion, more than it had paid, and cut thousands of jobs. It had bought the phones, more or less, to bury them. The public watched the handset vanish and assumed the company had gone with it.

It had not. It had just moved to the part of technology you never look at.

The treasure nobody could see

Here is what the phone story always skips. While the world mourned the handset, Nokia was sitting on one of the largest patent hoards in technology.

The portfolio now runs to around 20,000 patent families, each holding several individual patents, built on more than €150 billion of research since 2000. Over 7,000 of those families are essential to the 5G standard, which means roughly one in three of Nokia’s inventions is wired into the way any modern phone connects at all. More than 250 companies pay to license this work, including the rivals that beat Nokia in the shop. Samsung won on Android, and still writes Nokia a cheque every year. That licensing business alone pulls in around €1.4 billion a year, most of it high margin, for inventions Nokia largely made a decade or more ago. It is the closest thing in tech to being paid rent on the air.

This is the quiet power of infrastructure. You do not have to sell the phone if you own the physics inside it. Patents are much harder to disrupt than products, because a challenger has to either route around you or pay you.

Nokia lost the body. It kept the DNA.

Nobody notices infrastructure until it breaks

So Nokia became a network company: radio equipment, fibre, internet routing, optical networks, private wireless, patents and research. Less glamorous than a phone. Far harder to replace. Your call may travel through Nokia gear without you ever knowing, and increasingly so do factories, ports and power utilities. By 2025 this supposedly dead company still turned over about €20 billion, made €2 billion in comparable operating profit, and was busy buying its way deeper into optical and data-centre networking through its Infinera takeover. Not dead. Invisible. Then AI changed the price of being invisible.

The second AI race is happening outside the data centre

Most people picture AI as one loop: you type a question, a giant model runs in a distant data centre, an NVIDIA chip does the maths, an answer comes back. That works for a chatbot. It works badly for a drone, a factory robot, a self-driving car or a power grid. A chatbot can wait half a second. A machine moving at 100 km/h cannot.

Sending every decision to a data centre hundreds of kilometres away is often too slow, too costly or too fragile. Some of the intelligence has to sit close to the machine. That is edge AI. Instead of phoning distant headquarters every time a traffic light needs to change, you put enough brain at the junction to decide on the spot.

Which is where AI-RAN comes in. RAN, the Radio Access Network, is the towers, antennas and radios that connect your device to the network. Today it mostly carries traffic. AI-RAN wants it to do more: run the network smarter with AI, run AI apps close to users and machines, and let operators sell computing power, not just data. The tower stops being a traffic junction and starts becoming a small, distributed computer.

NVIDIA builds the brain. Nokia already owns part of the road.

That is why the NVIDIA deal matters. In October 2025 NVIDIA agreed to put $1 billion into Nokia and build commercial AI-RAN products for 5G-Advanced and 6G, with T-Mobile lined up to run the first live trials in 2026. NVIDIA brings the accelerated computing. Nokia brings the wireless technology, the telecom software, the operator relationships and equipment already bolted into networks worldwide. Analysts at Omdia put the wider RAN opportunity tied to this shift at more than $200 billion cumulatively by 2030.

The number is not the point. The business model is. A tower has always been an expensive way to sell connectivity. If the same site can also run AI, the operator can sell edge computing, private AI and industrial intelligence. The tower moves from cost centre to computing platform.

The part that turns serious: defence

On a battlefield none of this is theoretical. AI is worthless the second the signal drops, and modern enemies work hard to make it drop, with jamming, cyberattacks and drones of their own. Connections that keep working through all of that are Nokia's oldest speciality. In November 2025 Nokia and Finland's sovereign fund put €100 million into NestAI, a European defence-AI lab, and in July 2026 the first tools arrived: command-and-control running on Nokia's deployable 5G, mission planning that maps its own coverage, and threat detection stitched from Nokia's sensing radios and NestAI's tracking. All of it built to NATO standards, and designed to keep working in the "denied" conditions where the enemy is actively trying to kill the network.

Nokia's whole business was always connections that do not break. The stakes just moved from dropped calls to lost wars.

An opportunity, not a coronation

None of this is won. Ericsson, Samsung, Huawei, Mavenir, the big cloud firms and the operators themselves are all chasing smarter, more open networks, and NVIDIA wants the whole industry to adopt AI-RAN, not just Nokia. Nor is the money guaranteed. Telecom operators are famously tight with capital, new network designs take years to roll out, and the performance gains have to justify the bill. Nokia has been here before, holding excellent technology and still losing the market.

But this time its old strengths line up with the new fight. Nokia lost when software climbed above the device. It may win as AI sinks below the cloud and into physical infrastructure. The first smartphone war was fought on the screen. The next AI war may be fought underneath it, across towers, factories, ports, robots and defence networks where intelligence has to move now, not eventually.

We used to laugh that a Nokia could survive a fall from the nineteenth floor. The next durability test is harder.

Can Nokia survive history twice?