John Ithier, Europe Director,GCN, Colt Technology Services
Across Europe and beyond, a new generation of cloud providers, AI infrastructure specialists and neoscalers are rapidly expanding to meet demand for AI training, inference and high-performance computing. According to IDC, the global AI infrastructure market is forecast to reach $758 billion by 2029, with more than 84% of spending already taking place in cloud and shared environments, driven primarily by hyperscalers, cloud service providers and digital service providers. This reflects the rapid growth of specialised AI infrastructure providers and the increasing demand for AI-optimised platforms.
Much of the conversation has focused on GPUs, power availability and data centre capacity. But another challenge is increasingly determining how quickly organisations can scale: the network.
As AI workloads become more distributed across multiple facilities and geographies, the infrastructure required to move vast amounts of data between locations is becoming just as important as the compute itself. Data centre leaders expect interconnect bandwidth requirements to grow sixfold over the next five years, driven largely by AI workloads and distributed model training. For neoscalers, this creates a challenge that is often overlooked. The issue is not simply bandwidth, it's uncertainty.
The future is becoming harder to predict
Historically, network planning was relatively straightforward. Traffic increased steadily. Growth patterns were reasonably predictable. Capacity could be forecast and expanded over time. AI has changed the rules. A new customer deployment, GPU cluster expansion or AI service rollout can create significant increases in traffic almost overnight. Growth cycles that once took years can now happen in months. For infrastructure teams, that makes forecasting increasingly difficult. At the same time, demand for optical networking capacity is accelerating. Research indicates cloud providers are investing heavily in optical interconnection infrastructure to support AI growth, while suppliers are reporting growing backlogs and extended lead times as demand continues to outpace supply.
The result is a growing disconnect between business ambition and infrastructure certainty. Many organisations know they will need more capacity; they simply don't know when.
Why traditional connectivity models are under pressure
This uncertainty exposes the limitations of traditional approaches to connectivity. Many fast-growing organisations begin with wavelength services, purchasing capacity as and when demand increases. For predictable environments, that model can work well. But AI growth is rarely predictable. As demand becomes more dynamic, organisations can find themselves reliant on provider provisioning cycles, equipment availability and operational lead times. Every upgrade becomes another project, another approval process and another waiting period. The consequence is that networking infrastructure can become the very thing slowing growth. Ironically, at a time when businesses are moving faster than ever, many networks remain constrained by models designed for a more predictable era.
The shift from capacity planning to capacity certainty
This is prompting a shift in thinking. For years, networking teams focused on forecasting demand as accurately as possible. Today, leading organisations are recognising that forecasting alone is no longer enough. The more strategic objective is building infrastructure that can absorb uncertainty rather than attempting to predict it. In other words, the conversation is moving from capacity planning to capacity certainty.
The real question is no longer: "How much capacity will we need in three years?"
It's: "If demand doubles next quarter, can we respond immediately?"
For neoscalers operating in highly competitive markets, that distinction is becoming critical.
Because when growth opportunities emerge, infrastructure cannot be the thing holding the business back.
Why dedicated optical infrastructure is gaining momentum
As a result, many organisations are rethinking how they build and scale connectivity. Dedicated optical infrastructure offers a fundamentally different approach. Rather than purchasing bandwidth incrementally, organisations establish a scalable optical foundation capable of supporting future growth, with capacity available when needed. This provides something increasingly valuable in today's market: certainty.
Certainty that network capacity will be available when demand changes. Certainty that expansion plans are not dependent on provisioning queues. Certainty that infrastructure can keep pace with business growth. For organisations supporting AI workloads, where the consequences of delay can be significant, that certainty can become a competitive advantage.
Control without complexity
Neoscalers face an additional challenge. Unlike the largest hyperscalers, many operate with lean infrastructure teams. In some cases, a handful of networking specialists are supporting rapid expansion across multiple data centres, regions and customer environments. Their challenge is not only about gaining control. It's about gaining control without adding complexity. This is why managed optical network models are becoming increasingly attractive. A Managed Optical Fibre Network (MOFN) combines the benefits of dedicated optical infrastructure with the simplicity of a fully managed service. The organisation retains control over how capacity is utilised and scaled while avoiding the operational burden of building, managing and operating the underlying optical network itself. The result is a model more closely aligned to how modern neoscalers operate: lean teams, ambitious growth targets and constant pressure to move faster.
Why Colt is uniquely positioned to support this shift
This is where the provider behind the infrastructure matters. Capacity certainty is only possible if the underlying network can support it. Colt's advantage starts with ownership and control of extensive fibre infrastructure across Europe's most important digital markets. Unlike models dependent on multiple third parties, Colt owns and operates large parts of the network foundation itself, helping reduce complexity and increasing delivery confidence. But fibre alone is no longer enough. Today's AI ecosystems are built around data centres, cloud platforms and interconnection hubs. Colt's deep connectivity into key carrier-neutral facilities and major data centre ecosystems enables customers to connect where they need to be, whether that is within the FLAP markets, the Nordics, Germany, Iberia or emerging AI growth locations.
This is particularly important as AI infrastructure becomes increasingly distributed across multiple facilities and regions. Colt combines infrastructure ownership with experience designing and managing complex optical environments. For neoscalers that lack large in-house optical engineering teams, this provides access to specialised expertise without sacrificing strategic control.
The outcome is more than connectivity. It is a scalable, resilient and future-ready optical foundation designed to evolve alongside the business.
Reach out if you’d like to learn more about Colt’s MOFN solution.
