As digital infrastructure continues to evolve at pace, there’s a growing focus on power, sustainability, and compute performance.
From hyperscale cloud providers to AI-driven startups, everyone wants faster, greener, and more scalable systems. But while discussions around energy use and thermal efficiency are rightly at centre-stage, one part of the data centre stack has been allowed to drift into the background – network switching.
Switching rarely gets much attention. It’s often treated as an operational necessity – one of those things that is plugged in and expect to work. But the reality is that legacy switching architectures are increasingly becoming a bottleneck. They draw more power than many realise, produce significant amounts of heat, and are poorly suited to the demands of high-density, high-speed environments.
For organisations building next-generation infrastructure, this blind spot is beginning to matter.
More data, more switching, more pressure
The volume of data passing through data centres has increased dramatically over the last decade. And it’s not just the quantity that’s changed – it’s the nature of the traffic. There is more east-west movement between nodes, more real-time requirements, and heavier use of graphics processing unit (GPU) clusters that depend on low-latency data flows.
Every bit of this data passes through switches. In traditional systems, these switches convert data signals between optical and electronic formats to make routing decisions. This constant back-and-forth doesn’t just consume power; it also generates heat. And as demand rises, so does the cost – both financially and in terms of energy efficiency.
In some high-performance environments, switching now accounts for a surprising share of total power usage. According to NVIDIA, switching in data centres handling dynamic AI workloads typically makes up 8% of energy consumption. It’s also increasing the load on cooling systems, which need to work harder to manage the thermal output of inefficient network gear.
Shifting the conversation
Ultimately, the industry has reached a point where it’s no longer enough to focus on compute and cooling in isolation. Switching needs to be considered as a first-order infrastructure concern and a core part of the strategy to reduce energy use, enable greater density, and future-proof performance.
This shift is particularly relevant in the context of AI, high-frequency trading, and real-time media environments, where performance demands are high and speed is critical.
A new approach to switching
This is where optical switching is beneficial. Unlike conventional systems, optical switches keep data in the light domain from start to finish. There’s no conversion between light and electricity, which means much lower energy consumption and less heat generation. And because there’s no electronic processing involved, latency is significantly reduced.
Finchetto, a UK-based innovator, has developed an all-optical switch that operates at the packet level and requires no power to perform the switch itself. It’s already being deployed in data centres where efficiency is paramount. By introducing switching technology that consumes up to 50 times less power than traditional alternatives, costs and energy consumption can be reduced.
Critically, these switches are designed to be installed in existing rack-level environments and can work with standard protocols like Ethernet. This makes them a practical upgrade without requiring a complete infrastructure overhaul.
The knock-on effect: cooling and cost
What makes optical switching especially compelling is its impact beyond the network layer. Less power consumed by switches means less heat produced. That immediately reduces the load on the cooling system, freeing up capacity and potentially allowing for denser rack deployments.
It also makes more sustainable design strategies – such as heat reuse – easier to implement. Lower, more predictable heat profiles simplify integration with district heating networks, greenhouse warming, or even internal energy recycling initiatives.
Avoiding the infrastructure trap
It was once safe to assume that switching was a problem solved, but that is no longer the case. Legacy switches were not built for the demands that are being placed on them today – and certainly not for what’s coming in the next three to five years.
The shift to optical switching is not about chasing novelty, it’s about correcting an imbalance. Power is being addressed, cooling is being overhauled, and next-generation compute investment is recognised, so it makes sense to bring the network up to speed too.
How to get started
Taking advantage of optical switching doesn’t involve throwing out what works. New switching solutions are designed to be adopted incrementally. Operators can begin by deploying them in high-density pods, AI clusters, or experimental workloads. These environments are where the energy and performance payoffs are most pronounced.
Over time, the benefits justify wider rollout. And because the technology is built to work within familiar topologies, it avoids the risk and disruption that come with more radical rearchitecting.
The key is to start with a clear understanding of where switching inefficiencies are most acute and then build a roadmap for improving them that aligns with broader business and sustainability goals.
A small change with big impact
Data centres today are under more pressure than ever to do more with less. And while switching might not be the most exciting part of the infrastructure, it’s one of the most overlooked sources of waste. By reducing switching-related energy use and easing pressure on cooling systems, operators can improve performance and cut costs, whilst increasing sustainability.
About the Author
Darren Watkins is Chief Revenue Officer at VIRTUS Data Centres. VIRTUS Data Centres (VIRTUS) is a trusted global leader in the data centre industry and the UK’s leading data centre company, committed to delivering reliable, resilient and responsible digital infrastructure for our customers. As part of ST Telemedia Global Data Centres (STT GDC), VIRTUS Europe is the home of the cloud in London, and at the heart of Europe.
