Even for me, as an AI and emerging tech expert who has been working for the past decade on AI and emerging tech for critical infrastructure sectors such as energy and transport, it is easy to forget that digital communications is critical infrastructure, and in fact, one of the most underpinning layers there is.
Today was a chance to deep dive on this at the AI & Advanced Communications Conference. Hosted at IET London: Savoy Place and organized by UKTIN, the event brought together some of the absolute “who’s who” of the UK telecoms and AI landscape, from the teams at UKTIN and Digital Catapult to academic veterans, government reps, and giants like Amazon and Nvidia.
The depth of the discussions was matched only by the amazing networking; having been immersed in this world during 6-7 years at UKRI and Digital Catapult, it was great to see so many former partners, colleagues, and friends in the room (if I missed you, let’s DM to say hi!).
I wanted to share some of my perspectives on what we heard today and where we are heading, I have even fed my notes to NotebookLM to create an infographic!
It is clear that the UK’s telecommunications sector, past its heydays of the 90’s and ’00s, has reached a strategic turning point. AI could be the catalyst that spurs a new cycle, allowing the UK to move back up the ladder in the telecom world with intelligent, software-defined networks.
The Strategic Scope: From AI-on-Top to Native AI Networks
The core conversation is shifting decisively from “How do we add AI to our networks?” to a more profound architectural evolution.
As Dritan Kaleshi (Digital Catapult) masterfully framed it, we are now navigating a three-way intersection: “AI for networks, networks for AI, and AI in networks to address use cases.” This highlights a fundamental shift in our thinking: we aren’t just looking at automation or Agentic AI bolted onto existing systems as an “overlay.” Instead, we should be thinking about a move toward Native AI Networks: non-deterministic, adaptive environments where machine learning is “fused” into the protocol stack from day one.
In this paradigm, the physical layer itself becomes intelligent. We discussed concepts like Integrated Sensing and Communication (ISAC), a key focus for Prof. Harald Haas (TITAN Hub) and Prof. Dominic O’Brien (HASC Hub), where the network doesn’t just move data but actively senses the environment. Furthermore,
Prof. Julie McCann (CHEDDAR Hub) introduced the fascinating concept of “over-the-air computing,” where the wireless channel itself performs computations as data is transmitted. This is a move toward a network that acts as a distributed computer, where intelligence isn’t a feature you call via an API, but a foundational utility embedded in the hardware and software layers to manage real-time resource allocation and self-healing loops.
The Open Ecosystem vs. Proprietary Black Boxes
A major debate centered on whether the future of telecoms will be built on open models or proprietary “black boxes”. Amanda Brock, CEO of Open UK, argued that openness is a “no-brainer” in the AI environment because it allows for rapid iteration and building on existing innovation, citing how open models have already significantly reduced the cost and compute requirements for Large Language Models (LLMs). She compared the infrastructure of the internet to a “pizza base,” noting that while software infrastructure is now over 76% open source, the telecoms sector has historically been one of the slowest adopters due to risk-averse gatekeepers in legal and procurement. However, startups like Madevo highlight that the unique value proposition now lies in taking these open-source models and building specialised, efficient layers on top of them.
Evolution of Standards
The pace of AI development creates a “dissonance” with traditional ten-year telecommunications cycles. David Boswarthick of ETSI acknowledged that the way standards were made for 4G is no longer fit for purpose in the 6G era. To survive, standardisation must become lighter, faster, and more adaptive, shifting from 700-page word documents to “living code” and collaborative software development groups. ETSI is now focusing on “standardising just enough” in critical areas like data quality, ontologies, and security guardrails to ensure infrastructure does not fail.
Trust, Testing, and the “Black Box” Problem
As AI enters network control loops, it presents a challenge to the UK’s reputation for resilient, well-assured networks. Dritan Kaleshi from Digital Catapult warned that there is currently no unified end-to-end telecom AI test and verification framework. Because AI adapts over time, traditional deterministic staging is no longer sufficient.
The UK is addressing this through major R&D infrastructure initiatives, and today we heard from the people driving these missions:
- JOINER: Prof. Dimitra Simeonidou (University of Bristol) told us about this national-scale experimentation platform connecting 14 labs to provide distributed GPU resources for AI deployability assessments.
- CHEDDAR: Prof. Julie McCann (Imperial College London) shared insights on this hub’s focus on resilience and trustworthy AI, exploring “over-the-air compute” where the network itself performs calculations to control autonomous systems like drone swarms.
- TITAN: Prof. Harald Haas (University of Cambridge) explained the vision for a “network of networks” that is self-healing and self-configuring, ensuring connectivity even in the “high altitude economy” of drones and robots.
- HASC: Prof. Dominic O’Brien (University of Oxford) detailed how AI is being used to transform the spectrum scarcity problem into an automation and data problem, enabling more efficient spectrum sharing.
Temporary Hype or Economic Super Cycle?
From an investment perspective, the room grappled with whether we are witnessing a passing bubble or the dawn of a genuine “industrial age of AI.” David Pollington noted that we are in a true economic super cycle where the fundamental unit cost of intelligence, that is the cost of inference, is dropping by a staggering 5 to 10-fold every single year. Unlike the superficial tech hypes of the past, this transformation is anchored by massive algorithmic efficiencies and GPU innovations that offer tangible, measurable value to our national infrastructure.
As Pollington articulated, this deflationary pressure on intelligence is set to unlock use cases that were previously economically unviable. When intelligence becomes “too cheap to meter,” the network can support high-density, low-latency applications that were once pipe dreams such as ubiquitous, real-time coordination for massive robotic swarms, battery-free “ambient IoT” devices that harvest energy from the network itself, and the “high altitude economy” of drones requiring constant, self-healing connectivity in complex environments.
The Path Forward: Bridging the Scale-Up Gap
However, a sobering concern remains: while the UK leads Europe in foundational research and early-stage funding, we face a critical lack of home-grown AI infrastructure and large-scale growth capital. This creates a “valley of death” for our most promising startups. Without a strategic effort to build sovereign compute power and domestic scale-up capital, we risk watching our best innovations, and the economic value they generate, be swallowed by the US market as companies are either bought up or move across the Atlantic to find the resources they need to scale.
Bridging this gap was a definitive consensus of the conference: for the UK to remain competitive, we must move beyond viewing the network as a “cheap pipe.” Instead, we must integrate telecoms into the heart of the AI ecosystem and grab that as a vast business opportunity. This means proactively offering distributed compute and low-latency offloading capabilities to support the next generation of distributed AI, robotics, and autonomous systems. The goal is no longer just connectivity, it could become the foundational intelligent fabric of the next generation of connected, intelligent assets and infrastructure.
Final thoughts
While the UK may have lost some of its way in terms of raw industrial capacity over the decades, the sheer caliber of innovation on display today provides hope that this shift toward AI-native infrastructure is more than just a technical challenge, it is a potentially large enabler and market opportunity. Let’s hope this new wave of innovation spurs the next generation of home-grown unicorns in the sector.
I would love to compare notes with colleagues across the ecosystem!
#AI #6G #Telecoms #Innovation #UKTech #DigitalInfrastructure #UKTIN #IETLondon #OpenSource #OpenUK #ETSI
