Is Nokia Becoming Europe’s AI Infrastructure Company?
Why one telecom company’s transformation may reveal where artificial intelligence is heading next
Artificial intelligence does not become infrastructure merely because more computing power is built. It becomes infrastructure when intelligence can move—between data centres, factories, machines, hospitals, ports and public systems. That makes the network layer increasingly decisive.
For most of the artificial-intelligence era, attention has concentrated on what happens inside the data centre. Nvidia supplies the processors. Hyperscalers assemble enormous computing environments. Model developers turn that capacity into increasingly powerful systems. But intelligence confined to centralised computing facilities remains only partially connected to the economy around it.
Factories do not operate inside data centres. Neither do ports, electricity grids, hospitals, vehicles or telecommunications systems. If artificial intelligence is to become embedded in the physical economy, computing power must be connected to the places where data is generated and decisions are executed. That requires networks capable of transporting enormous volumes of information with sufficient speed, reliability and security. It also requires those networks to become more intelligent themselves.
The next phase of artificial intelligence may be determined not only by who builds the largest models, but by who connects intelligence to the physical world.
Nokia’s strategic transformation should be understood within this emerging architecture. The Finnish company is still widely perceived as a telecommunications equipment manufacturer. Yet its combination of optical networking, IP routing, mobile infrastructure, data-centre connectivity and edge computing increasingly places it inside a much larger technological system.
Nokia may not be building Europe’s answer to the American hyperscalers. It may be building part of the infrastructure through which artificial intelligence leaves the hyperscaler behind.
From telecommunications to AI infrastructure
For decades, telecommunications networks were treated primarily as transportation systems. Their task was to carry voice, messages and data between users, devices and computing centres. Network quality was measured through coverage, capacity, speed and reliability. Intelligence largely remained at either end of the connection. Artificial intelligence begins to change that division.
AI systems generate vast and continuous flows of data. Training models requires large centralised computing environments, but using those models across the physical economy produces a different set of requirements.
An industrial robot may need to respond within milliseconds. A vehicle cannot always wait for instructions from a distant cloud region. A hospital cannot allow every sensitive data stream to leave its local environment. An electricity grid must detect and respond to changing conditions almost instantaneously. In such environments, the location of computing becomes as important as the quantity of computing.
Some decisions can still be processed centrally. Others must take place closer to the machine, organisation or infrastructure generating the data. The result is a more distributed computing architecture in which data centres, telecommunications networks, edge facilities and connected devices begin to function as parts of the same system.
AI becomes infrastructure when intelligence can move beyond the data centre.
That changes what a network is. It no longer merely connects computers. Increasingly, the network determines where computing happens, which data moves, what must remain local and how intelligence is distributed across the system.
The network becomes the computer
Nokia’s work on AI-RAN illustrates this transition. Traditional radio access networks provide the infrastructure connecting mobile devices to the wider telecommunications system. AI-RAN introduces computing capabilities that can support both network operations and artificial-intelligence workloads.
In collaboration with Nvidia, Nokia is exploring how telecommunications infrastructure can combine connectivity with accelerated computing. Processing capacity can then be placed closer to users, machines and industrial environments instead of being concentrated exclusively inside distant hyperscale data centres. That creates several possibilities.
Artificial intelligence can help optimise the network itself, dynamically allocating capacity and improving energy efficiency. At the same time, the underlying infrastructure can potentially support external AI applications operating at the edge. The distinction between telecommunications equipment and computing infrastructure begins to weaken.
A base station may continue to provide connectivity, but the location can also become part of a distributed computing environment. Network infrastructure can host inference, process local data or support applications requiring extremely low latency.
This does not mean that every telecom site will become a miniature data centre. Nor does it mean that centralised cloud computing will disappear. It means that the architecture of artificial intelligence is likely to become more geographically distributed.
The cloud remains important. But it is joined by a new layer between the hyperscale data centre and the connected device. That layer is the intelligent network.
Optical networks and the movement of intelligence
The AI infrastructure debate often begins with processors. Yet processors are useful only when they can receive data and communicate results.
As computing clusters grow, the connections within and between data centres become a critical constraint. Thousands of processors must exchange information at enormous speed. Data must move between storage systems, computing environments and geographical locations without turning the network into a bottleneck. This makes optical networking increasingly important.
The network is no longer merely carrying intelligence. It is beginning to participate in it.
Nokia’s optical and IP networking businesses therefore should not be viewed as separate from its artificial-intelligence strategy. They form part of the infrastructure required to connect increasingly large and distributed computing environments. The more powerful AI systems become, the greater their dependence on the movement of data.
This creates a subtle reversal in the technology stack. Computing was once considered the active layer, while connectivity appeared to be a supporting utility. In the emerging AI economy, the performance of the network can determine how much computing capacity can actually be used. Bandwidth, latency and network architecture become constraints on intelligence itself.
Nokia’s importance may therefore lie less in owning a single dominant AI product than in occupying several of the interfaces through which computing systems are connected: between processors, between data centres, between clouds and telecommunications networks, and between centralised AI and the physical economy. These interfaces are rarely visible to the public. Strategically, however, they are becoming difficult to ignore.
Europe’s overlooked layer
Europe is generally portrayed as trailing the United States in artificial intelligence. At the model layer, this diagnosis contains considerable truth. American companies dominate hyperscale cloud computing, advanced AI processors and many of the world’s most influential foundation models. Europe possesses promising companies and research capabilities, but lacks platforms with equivalent scale.
Yet artificial intelligence is not a single market. It is a technological architecture composed of multiple interdependent layers.
Europe may not control the model layer, but it still controls critical points of connection.
Europe remains strong in several of them: semiconductor equipment, photonics, industrial automation, telecommunications infrastructure, power systems, enterprise software and advanced manufacturing.
These capabilities do not remove Europe’s dependence on American computing platforms. But they do complicate the idea that Europe has no meaningful position in artificial intelligence.
Nokia represents one of these less visible positions. Its technologies help determine how computing environments communicate and how intelligence reaches industrial and public systems. Ericsson occupies a related position. European photonics companies contribute components that enable high-speed optical communication. Siemens and Schneider Electric connect software and automation to factories, buildings and energy systems.
Together, these capabilities form part of the architecture through which AI becomes operational.
Europe’s potential advantage may therefore lie not solely in reproducing the American model economy, but in controlling parts of the infrastructure required to deploy intelligence across the physical world. That is not a substitute for computing power. It is a different source of leverage.
The Nvidia question
Nokia’s collaboration with Nvidia demonstrates both the opportunity and the limitation of Europe’s position.
Nokia contributes networking expertise, radio infrastructure and relationships with telecommunications operators. Nvidia provides the accelerated computing platform around which much of the current AI economy is being organised.
The partnership can allow Nokia to move deeper into AI infrastructure. But it also risks embedding American compute architecture more firmly inside European networks. This creates a strategic tension.
If Nokia succeeds, Europe gains a stronger position at the intersection of telecommunications and artificial intelligence. Yet part of the underlying computational value may still flow through Nvidia’s hardware, software ecosystem and technical standards.
The network may become European in operation while remaining partly American in computational logic. That does not make the collaboration undesirable. Complex technology systems are always built through interdependence. Complete technological autonomy is neither realistic nor necessarily efficient.
The more important question is whether Europe retains sufficient control over critical functions inside that interdependent system.
Can European operators determine how infrastructure is configured? Can sensitive workloads remain within trusted environments? Can alternative processors and software platforms eventually be integrated? Who controls orchestration, standards and access to the data generated by the network? And which company captures the long-term value created when connectivity and computing converge?
Sovereignty is not achieved by excluding every foreign technology. It is achieved by preventing dependency from becoming structural submission.
Telecom operators face a choice
Nokia’s transformation also raises questions for Europe’s telecommunications operators.
For years, operators have invested heavily in infrastructure while much of the economic value generated through that infrastructure accrued to digital platforms. Telecom companies financed networks; technology companies used those networks to build globally scalable services. Artificial intelligence could repeat that pattern.
Operators may provide the physical sites, fibre connections, spectrum and customer access required for distributed AI. Hyperscalers and chip companies could nevertheless capture most of the value through cloud platforms, processors and software ecosystems. Alternatively, telecom operators could use their infrastructure to develop a more substantial position in the AI economy.
They possess geographically distributed assets, trusted relationships with businesses and governments, local data environments and infrastructure close to end users. These capabilities could support sovereign cloud services, industrial edge computing and AI applications requiring local processing. But that would require operators to think beyond connectivity.
They would have to treat their networks not merely as channels through which other companies deliver intelligence, but as computing environments capable of hosting, managing and governing intelligence themselves.
Nokia can provide part of the architecture. Whether operators use it strategically is a different question.
Infrastructure becomes sovereignty
Digital sovereignty has often been discussed as a choice between European and foreign software platforms. Artificial intelligence makes that debate more physical.
Who owns the fibre connecting AI data centres? Who provides the optical systems carrying data between computing clusters? Who operates the radio infrastructure through which machines and people connect? Where does inference take place? Which data must travel to the cloud, and which can remain close to its source?
These are not secondary technical questions. They determine who can access critical systems, who can interrupt them, who can observe the resulting data and who can decide how intelligence is distributed across society.
As artificial intelligence becomes embedded in factories, ports, hospitals, transport systems and public administration, the network becomes part of the governance architecture surrounding AI.
The strategic question is not who connects Europe, but who captures the value moving through its networks.
Connectivity determines reach. Latency determines which applications are possible. Network design determines where data travels. Infrastructure ownership determines where strategic dependency begins.
This is why Nokia’s evolution matters beyond its financial performance or corporate strategy. The company occupies a layer that will increasingly shape how artificial intelligence enters European society.
Nokia’s unfinished transformation
Calling Nokia an AI infrastructure company would still be premature if the phrase implied that its transformation were complete.
Traditional telecommunications remains central to the company. The commercial model for AI-RAN is still developing. Telecom operators face investment constraints, and the economic division between network vendors, chipmakers, hyperscalers and operators remains unsettled. It is also unclear how quickly edge AI will develop at scale.
Centralised data centres benefit from enormous efficiencies. Many applications do not require real-time local processing. The movement towards distributed intelligence will therefore be uneven, driven by specific industrial, security and regulatory requirements rather than by a universal replacement of the cloud. But corporate transformations often become visible before their final market structure is clear.
Nokia’s growing presence in data-centre networking, optical systems, IP infrastructure and AI-enabled radio networks suggests that the company is moving towards a broader role. It is no longer positioned only as a supplier to the telecommunications industry. It is increasingly positioned between compute and connectivity. That may become one of the most consequential places in the AI stack.
The network layer
The first phase of artificial intelligence was organised around models. The second has exposed the importance of processors, energy and data centres. The next phase will be about connection.
Intelligence will have to move from centralised computing clusters into factories, vehicles, energy grids, hospitals and cities. It will have to operate across different jurisdictions, respond within milliseconds and process sensitive data without always transferring it to a distant cloud.
That requires more than algorithms. It requires an infrastructure capable of deciding where intelligence resides, how it travels and when it must act locally.
Nokia’s transformation offers an early view of that emerging architecture. The company may never possess the cultural visibility of an OpenAI or the market power of an Nvidia. Its role is quieter and more structural.
Nokia helps build the layer through which artificial intelligence becomes connected, distributed and physically present.
Europe’s position in AI may ultimately depend not only on whether it can produce the largest model or the most powerful processor. It may also depend on whether European companies remain indispensable at the points where computing meets infrastructure.
Nokia cannot solve Europe’s AI dependency alone. But it may demonstrate that the network is not simply the space between computers.
Increasingly, the network is becoming part of the computer itself.
This article is part of Who Builds Europe’s AI?, an Altair Media Perspective series exploring the technological, industrial and strategic foundations of European artificial intelligence. From algorithms and compute to networks, energy and digital sovereignty, the series examines the interconnected layers that together determine Europe’s AI future.
Credit
Illustration: Altair Media — conceptual visualisation
Caption
The network becomes part of the computer. As artificial intelligence moves beyond centralised data centres, optical networks, telecom infrastructure and edge computing will determine how intelligence reaches factories, hospitals, energy systems and public infrastructure. Nokia’s transformation reveals why the network layer is becoming a strategic part of Europe’s AI architecture.
