The rapid acceleration of artificial intelligence is forcing a fundamental reappraisal of the digital infrastructure that underpins modern economies. A new study commissioned by Nokia suggests that technology and business leaders on both sides of the Atlantic believe existing network architectures are no longer sufficient to support the next phase of AI adoption.
Based on responses from around 2,000 technology and business decision-makers across the United States and Europe, the research points to an emerging consensus: without substantial evolution and sustained investment in network infrastructure, the promise of the AI supercycle risks colliding with physical and architectural constraints. The respondents included telecommunications providers, data centre operators and enterprises already deploying or planning to integrate AI into their operations, highlighting how widely the issue is now recognised.
The study frames AI not simply as a software challenge, but as a structural one. As AI systems become more deeply embedded across industries, the demands placed on networks are changing in ways that current designs were never intended to accommodate.
How AI is reshaping network requirements
The research identifies a shift in the basic traffic patterns that networks must support. Traditional networks were optimised for downlink-heavy consumer use cases, such as web browsing and video streaming. AI workloads, by contrast, are increasingly uplink-intensive. Applications ranging from autonomous vehicles and smart manufacturing systems to surveillance platforms and remote healthcare diagnostics generate vast volumes of data at the edge, which must be transmitted upstream for processing and analysis.
At the same time, AI is driving more distributed data flows, tighter latency requirements and higher expectations around resilience, security and energy efficiency. These characteristics place stress on both fixed and mobile networks, particularly as real-time inference and feedback loops become integral to operational systems.
Pallavi Mahajan, chief technology and AI officer at Nokia, argues that the first wave of the AI supercycle has already altered how industries innovate, but that future waves will demand networks that are fundamentally more AI-native. In her view, connectivity, capacity and low-latency performance are no longer incremental upgrades, but essential building blocks for how devices, organisations and societies function as AI adoption deepens.
Diverging pressures in the US and Europe
While the underlying challenges are shared, the research highlights different regional dynamics. In the United States, where AI deployment and mass-market adoption remain relatively advanced, 88 per cent of respondents expressed concern that network expansion may not keep pace with the scale of AI investment. Priorities identified by US respondents include optimising bi-directional data flows, expanding fibre capacity, enabling real-time training feedback and deploying low-latency edge infrastructure to support increasingly distributed workloads.
In Europe, the findings point to a more immediate gap between ambition and readiness. Eighty-six per cent of enterprise respondents said current networks are not yet equipped to handle widespread AI adoption. Despite two-thirds already running AI in live environments, more than half reported experiencing issues such as downtime, latency or throughput constraints linked to rising data demands.
European respondents emphasised the importance of regulatory conditions in shaping the pace of infrastructure development. Calls for more consistent regulatory simplification, aligned spectrum policies, adjustments to competition frameworks and greater investment in energy-efficient, AI-ready networks reflect concerns that fragmented markets could slow the continent’s ability to scale AI effectively.
Infrastructure as a competitiveness issue
Beyond technical considerations, the research positions network evolution as a matter of national competitiveness and long-term digital leadership. As AI systems become integral to critical services and industrial processes, the capacity and reliability of underlying networks take on strategic importance.
Nokia is using the findings to encourage closer collaboration across the connectivity ecosystem and more predictable regulatory environments that support timely investment. The message implicit in the research is that AI’s trajectory will be shaped as much by physical infrastructure as by advances in models and algorithms.
As organisations move from experimentation to large-scale deployment, the AI supercycle is revealing a simple constraint. Without networks designed for uplink-heavy, low-latency and energy-efficient operation, the next phase of AI-driven transformation may be harder to realise than the technology itself suggests.
