The AI supercycle explained

The AI supercycle is the latest major technology wave - a multi-decade period where artificial intelligence becomes embedded in every industry, service and device.

As AI shifts from narrow applications to real-time, immersive and autonomous systems, it drives new demands for capacity, speed, and trusted connectivity. This is accelerating the need for AI-native networks designed to handle far more data, far more intelligence, and far more complexity.

For Nokia, the AI supercycle isn’t defined by clever demos, but by three on-going shifts:

• AI Inference diffusion from centralized LLMs to more distributed multi agent systems
• Compute expansion: a step-change in demand for capacity, optimized for AI-specific workloads, and distributed closer to the endpoint
• Network evolution: the move to deterministic, high-capacity, low-latency networks that allow AI to work everywhere, not just in the cloud

Once these three shifts converge, AI becomes a general-purpose technology that drives long-term growth across the entire economy.

AI has gone from research curiosity to industrial engine in a matter of years. Previous technology euphoria’s have taken decades to embed in society and the economy, but AI is on a fast track like nothing we have seen before.

The clues are obvious when you think about it. Record GPU shipments, rapid data growth, 5G Standalone buildouts, edge-cloud convergence, and a once-in-a-generation surge in venture and national investment.

But the real reason the supercycle is here is simpler: AI broke out of the data centre.

Inference is now happening on devices, in factories, in vehicles, across supply chains, and in the network itself. It is in accountancy software, and voice assistants on our smartphones. It has become normalized. That shift is irreversible.

This is what you could call the diffusion phase, the moment when AI becomes embedded everywhere and the underlying infrastructure must be re-engineered to keep up.

A supercycle focuses on a continuous development. It happens when progress stops behaving linearly and starts behaving exponentially, when every breakthrough seeds the next, creating momentum you can’t easily unwind.

More powerful models drive new applications, which drive heavier infrastructure investment, which fuels the next leap in capability. Better tools attract more talent, more talent creates better tools, and the cycle tightens. That’s when an industry stops evolving slowly and starts pulling itself forward.

That feedback loop is what turns a moment into a decade-long cycle. The more successful AI becomes, the greater the acceleration.

If compute is the engine of AI, connectivity is the chassis that keeps it on the road.

The often-unspoken secret of AI is that it’s useless if it can’t reach the data, devices, and users it needs reliably, securely, and in real time. Connectivity is the link between locked off intelligence, and real-world benefit.

This is why Nokia keeps emphasising several areas which need to be prioritized:

• 5G Standalone as the baseline for AI-driven mobility, network slicing, and deterministic latency
• 6G pathfinding focused on sensing, extreme reliability, and native AI integration
• Optical transport to move massive training and inference traffic between hyperscale and edge, as well as between locations. With multi-agent AI accelerating, Data Center Interconnect becomes essential to sustain the surge in east–west traffic between compute sites, ensuring distributed models stay synchronised, low-latency, and able to operate as one unified intelligence
• Edge compute to anchor inference closer to where data is generated, delivering real-time performance while keeping sensitive information sovereign and under local control
• Quantum-safe networking to harden the backbone before AI-native threats arrive

The network remains critical to ensuring AI progresses. Simply put, without the vital link between intelligence in the cloud and the end-user (not to mention the link between different AI and cloud environments), the AI supercycle simply won’t happen.

The AI supercycle is already visible in macro indicators: rising national AI budgets, data-centre investment growth, spectrum policy reform, and a scramble for talent. PwC estimates AI could add $15.7 trillion to global GDP by 2030, but only if countries have the infrastructure to absorb it.

Regions with dense fiber, strong 5G SA coverage, modernised grids, and open innovation ecosystems will pull ahead. Regions without them will find AI adoption patchy, expensive, and slow.

Nokia has recently conducted research with more than 2000 technologists and business decision makers in the U.S. and Europe to understand how prepared the regions are for the AI supercycle.

While there is significant ambition to deploy the technology in both markets, infrastructure cannot support the economic objectives. In the U.S., 88% of the survey respondents were concerned infrastructure would limit AI scaling, while in Europe, 78% were concerned.

The research confirms the simple statement highlighted above; without the right networks in place, the AI supercycle does not happen.