LEE MYALL POWER, COMPUTE, AND FIBRE
Neos Networks found that 82% of UK data centre operators have delayed site builds or expansions due to fibre availability. The result is a growing risk of mis-sequencing. Data centres may look ready on the surface but experience delays, constrained go-live capacity, or costly retrofits because the optical foundations are not in place early enough. The industry desperately needs to shift the conversation from coverage to readiness. National fibre availability tells us very little about whether AI infrastructure can really scale in practice. What is crucial is whether the right corridors and clusters have the capacity, diversity, and provisioning speed to bring new sites online without friction. WHY THIS IS NOT ANOTHER RECKLESS FIBRE BOOM There is the argument that if fibre is so critical, the solution is simply to build more of it. However, the market has learned hard lessons from previous infrastructure cycles, and that experience is shaping today’s investment behaviour. In the early 2000s, the industry underwent a fibre overbuild cycle driven by speculative optimism. Networks were hastily deployed ahead of real utilisation. The result was stranded assets, collapsing returns, and a painful correction. Today’s environment is fundamentally different. Capital is more expensive, investors are less tolerant of speculative build, and both operators and infrastructure providers are under pressure to prove demand-led deployment rather than footprint for footprint’s sake. And AI itself is reinforcing that discipline. Rather than being distributed evenly, AI demand concentrates in specific corridors, including hyperscale data centre clusters, cloud on-ramps, interconnect hubs, and emerging AI Growth Zones. Recent research once again proved this trend, with 96% of UK data centre operators saying that AI Growth Zones are influencing expansion and site selection. There is also strong signals for geographical diversification, with more operators pointing to the North of England and the Midlands (39%) for new investment (greater than the 23% of operators expecting growth in established hubs in Greater London). That means the smartest investment is not about building the broadest footprint, but targeting optical fibre readiness – capacity, diversity and scalability – in the places AI workloads will actually land. A CASE STUDY IN STRATEGIC BUILD: PROJECT REACH An example of this disciplined approach is Neos Network’s partnership with Network Rail on Project Reach, which leverages the mainland UK railway network to deploy new
high-count fibre along critical corridors. Project Reach enables faster and less disruptive builds, avoiding the complexity of wayleaves, road closures and local authority coordination that can delay traditional deployments in the highway. It also improves predictability, with structured build windows supporting more accurate delivery timelines, increasingly valuable when data centre energisation schedules depend on connectivity infrastructure being ready on time too. And by focusing investment on backbone routes that connect key regions and interconnect points, Project Reach aligns fibre delivery with real market demand, helping ensure high performance fibre infrastructure is built where AI infrastructure growth is being realised. The principle is broadly applicable: reduce friction, build where demand is forming, and prioritise readiness over raw route kilometres. HOW THE INDUSTRY CAN IMPROVE OPTICAL READINESS None of this can be solved by fibre builders alone. The optical ecosystem, which includes vendors, operators, data centre providers, standards bodies, and policymakers, all have a significant role in making optical infrastructure more scalable, interoperable and faster to deploy. Several themes are becoming increasingly important as AI data centre demand grows: 1. Interoperability and standards as foundations AI data centres must communicate across diverse environments and provider networks. Using standardised protocols ensures these links remain stable and manageable as capacity increases. For example, open optical standards such as Open ROADM, standardised pluggables and common management interfaces allow data centre operators to light capacity across different vendors’ equipment and multiple networks without complex bespoke integration. This makes it far easier to scale interconnect capacity, introduce new hardware generations, or reroute traffic as AI clusters expand. Following industry standards simplifies the integration of different hardware and software, making hardware refreshes faster and allowing for more flexible routing. In a shifting market, this prevents vendor lock- in and ensures the infrastructure can be reconfigured as needs change. 2. Automation and faster provisioning The AI era demands an operational model that enables connectivity to be provisioned and scaled more quickly than traditional telecoms processes allow. Optical provisioning and service activation increasingly need to move closer to a cloud-like experience: fast, predictable, and controllable.
Automation is a key lever here. Not just to improve efficiency, but to enable a completely different consumption model for fibre connectivity. 3. Visibility and observability As dependencies grow, assurance becomes critical. AI workloads require predictable performance, and operators need deeper optical visibility to maintain it. Observability, including richer telemetry and proactive monitoring, is becoming a central part of meeting resilience expectations. In practice, this means having real-time insight into optical signal quality, latency, packet loss and utilisation across routes and interconnects, rather than relying solely on fault alarms after something breaks. With this level of visibility, operators can detect degradation early and reroute traffic or add capacity before AI workloads are impacted. 4. Resilience by design Finally, resilience is not an add-on feature. Data centre connectivity architectures must be designed for diversity from the outset: multiple routes, multiple interconnect options, and clear failover capability. AI won’t tolerate fragile connectivity. WHAT THE UK CAN LEARN, AND WHAT NEEDS TO CHANGE Other regions that have succeeded in scaling data centre capacity at pace have tended to tightly align infrastructure planning. Fibre, power, planning permissions and interconnect ecosystems are treated as connected parts of a single readiness strategy. The choice is not between data centres and fibre. We need to stop running them as two separate workstreams on separate timelines. Because AI infrastructure only works when the fundamentals move together – power, planning and optical connectivity – and when readiness is designed in from the start, not bolted on after the build. Now is the time to turn ambition into execution. If we focus on the corridors where demand is actually forming, build in resilience and route diversity, and ensure common standards and interoperability from the outset, we can scale capacity quickly as demand evolves. We can build an AI powerhouse. Lee Myall, CEO, Neos Networks Lee Myall joined Neos Networks in September 2023 as CEO. His goal is to drive continued growth and value for Neos Networks, ensuring it continues to be a leading telecoms and digital infrastructure provider for UK businesses. Lee has spent nearly 25 years in privately funded technology businesses. Prior to joining Neos Networks, Lee held many senior roles including CEO of leading data centre operator, Kao Data, CCO of Epsilon and various senior roles at Interoute.
www.opticalconnectionsnews.com
9
ISSUE 43 | Q1 2026
Made with FlippingBook flipbook maker