Get network infrastructure right or everything else is moot
By Pallavi Mahajan
Chief Technology and AI Officer
08 January, 2026
In today’s world, AI dominates the conversation. Consumers want it, businesses want it, and politicians want it. But how much of the conversation is being directed towards the most pressing issues?
Every day we see articles focused on upskilling employees or introducing new subjects on curriculums to avoid a talent shortage.
We see questions about how we can reduce the energy consumption of the data centers that train and house the brains of the technology.
New rules are being discussed and implemented to create guardrails and then debated as to whether they encourage or inhibit innovation.
These are all incredibly valid and valuable conversations to be had. The outcomes of these discussions will shape the AI era.
But what about the network?
The elephant in the room
Network infrastructure is a topic people return to often, yet it feels oddly missing from today’s headlines. Maybe that’s because connectivity and digital services have become so seamless and ever-present that we barely notice them anymore — we simply assume they’ll keep working.
But that complacency is exactly the problem.
The more invisible infrastructure becomes, the more fragile the system is. What’s changed is not our dependency, but the stakes. Connectivity is no longer a “nice-to-have” humming in the background, it’s the substrate on which new industries, new public-sector services, and new scientific breakthroughs are being built.
And our research confirms it is a significant challenge.
After speaking with more than 2000 decision makers and technologists, ranging from infrastructure providers to AI adopters, the results are mixed at best.
In short, they lack the confidence in the infrastructure which underpins everything digital. In the U.S., only 12% of the survey’s respondents show no concern in the ability for infrastructure to scale in line with the requirements of AI. Europe were a little bit more confident, but at only 24%, it is still a significant concern.
Get the latest insights, download the AI will Break the European internet report
Get the latest insights, download the Build first, lead forever report
This begs the question as to why we aren’t putting connectivity and network infrastructure as the number one priority to bring the AI supercycle to life? It was the most common hurdle to success in our research.
This is not to say the other topics are not important, but it is simply a case of timelines.
Yes, you need a skilled workforce, a sustainable approach to energy management and rules to safeguard customers and companies alike. But without the right digital infrastructure environment, is this not all moot?
Infrastructure is the first domino for the AI supercycle. It supports demand.
What we have today is infrastructure built for the last 20 years of consumer-led connectivity – it is download orientated. AI plays by different rules. It demands far greater uplink capacity to move massive datasets, rock-solid reliability for continuous inference, lower latency for real-time decision loops, and security baked in at every hop.
Above all, AI needs networks engineered for token-level throughput, optimized for machines talking to machines at unprecedented scale.
Without the right digital infrastructure in place, everything else is largely irrelevant.
Infrastructure gives the brains a way to talk
The emergence of AI creates a series of new conversations to be had concerning the network.
Extended coverage, because will not AI succeed if it only works some of the time.
Rearchitecting networks to allow for greater capacity, flexibility and performance, as well as uplink improvements because in mobile networks AI needs to be fed information, not just talk. Spectrum must be considered, as must latency, jitter, traffic prioritization and congestion. The list is long.
Firstly, you must consider the overarching coverage environment for mobility use cases.
If you want AI to be in the real-world to deliver consumer use cases such as autonomous driving, or accessible to employees who happen to be outside an office environment, the mobile network must be capable of supporting.
Secondly, you must consider the fixed networks.
Businesses are going to be heavy users of AI, but it doesn’t mean all of them will invest in private cloud to host compute power on-premises. The volume of data which flows in and out of a business will be significant in these scenarios. And this doesn’t have to be a significant percentage uplift on what the status quo is today.
It can grow incrementally, but that doesn’t make the process of upgrading fixed infrastructure any easier.
Finally, you must consider the networks we don’t see.
Data needs to flow from mobile sites to data centers, then it might have to move throughout the data center, or even to applications hosted in different data centers. All within the blink of an eye.
The old model of a single, monolithic AI sitting inside one data centre is already obsolete. What’s emerging instead is a mesh of specialised agents (reasoning agents, retrieval agents, vision agents, orchestration agents, safety monitors, optimisation engines) each running in different compute clusters, perhaps in different sites.
These invisible networks are becoming critical as AI asks questions we have not yet had to answer.
The network is the first thing to get right
No single point outweighs the others, but there is a sequence in which they matter, a chain of dependencies that ultimately determines whether some points matter at all.
AI isn’t powered by optimism; it’s powered by silicon, quality data and human ingenuity. But none of this functions without the right infrastructure beneath it.
If infrastructure stays out of the spotlight, we risk treating the most critical layer of the digital economy like background plumbing; invisible until something breaks.
This is exactly the moment to bring the network back into view, not as a cost centre but as the strategic advantage that sustains the entire AI supercycle.
Without the right connectivity and network conditions in place, there is no AI supercycle.
About Pallavi Mahajan
Pallavi Mahajan, Nokia’s Chief Technology and AI Officer, leads Nokia Bell Labs, Technology and AI Leadership, and Group Security to drive innovation in core technologies, strengthen AI and security capabilities, and create differentiation through open ecosystems and strategic partnerships. With deep expertise in networks, software, and AI, she has scaled multi-billion-dollar portfolios and shaped industry-defining shifts at Intel, HPE, and Juniper Networks. A holder of six patents and a passionate advocate for women in tech and grassroots sports, Pallavi champions collaboration to unlock the next wave of growth.
Connect with Pallavi on LinkedIn
