25 June 2025
|
Darren Watkins, Chief Revenue Officer at VIRTUS, on why smarter switching is transforming efficiency.
As AI and high-performance computing continue to reshape the data centre landscape, operators are under immense pressure to reduce power consumption and maintain thermal stability. Cooling, long seen as a support system, has become a central pillar of sustainable data centre design. But one critical factor is often overlooked: the network switch itself, and how its performance directly affects cooling demands.
The overlooked link between switching and cooling
Every byte of data that moves within a data centre must pass through network switches, and these switches are far from passive. In traditional designs, switching involves constant conversion of data signals between optical and electronic formats - a process that not only consumes significant power but also generates substantial heat.
This heat doesn’t disappear, it becomes part of the data centre’s thermal load, increasing the burden on cooling systems. As workloads become denser and more dynamic, especially in AI-rich environments, the switching layer is contributing to overall heat output more than ever. This means that any gains in switching efficiency can have a direct and measurable impact on cooling performance.
A new approach: tackling heat at the source
To address this challenge, operators are rethinking not just cooling technologies but the very architectures that produce heat. One of the most promising breakthroughs comes from Finchetto, a UK-based innovator that has developed a fully passive, optical switch capable of reducing energy use, and by extension, heat generation by up to 50 times compared to conventional electronic switches.
By keeping data entirely within the optical domain, Finchetto’s system eliminates the need for repetitive optoelectronic conversions that traditionally generate heat. The result is a dramatic reduction in thermal output at the network layer which allows cooling systems to work more efficiently and freeing up capacity for higher compute densities.
Finchetto has developed a fully passive, optical switch capable of reducing energy use
Innovations in cooling tech
While smarter switching tackles heat at the source, cooling systems themselves are evolving rapidly to meet new demands. Among the leading innovations are:
- Direct-to-chip liquid cooling, which brings coolant directly to the hottest components and drastically improves heat transfer rates compared to air cooling.
- Immersion cooling, where servers are submerged in dielectric fluids that efficiently absorb and dissipate heat, enabling much denser configurations.
- Rear-door heat exchangers and in-rack cooling, offering modular solutions that intercept heat at the rack level before it can impact the wider environment.
- AI-driven cooling control, which dynamically adjusts thermal management in real time to match actual workloads and environmental conditions.
These cooling technologies, when paired with low heat switching innovations, create a powerful synergy reducing heat at both the point of generation and the point of extraction.
Heat reuse: turning waste into opportunity
Another significant trend is the rise of heat reuse schemes. As data centres become larger and more integrated into urban environments, operators are increasingly looking for ways to repurpose waste heat to benefit local communities. Efficient cooling, and critically, lower overall heat production through innovations like optical switching, makes it far easier to harness usable heat without excessive energy overhead.
District heating networks, greenhouses and industrial facilities are all prime candidates for receiving repurposed heat, helping operators achieve sustainability targets while delivering tangible community benefits.
The case for integrated design
In the past, cooling, power and network teams often operated in silos. But as workloads evolve and energy pressures mount, an integrated approach is essential. Smarter cooling alone can’t solve the problem if other parts of the infrastructure, like switching, continue to produce avoidable heat.
The most efficient data centres today are those that align their cooling strategies with wider infrastructure choices, enabling each layer of the system to support overall efficiency. Optical switching is a prime example of how rethinking network design can cascade into lower cooling loads, smaller backup power requirements and improved facility performance.
Preparing for high-density workloads
The rapid rise of AI, big data, and high performance computing (HPC) means rack densities will only continue to grow. McKinsey claims that rack density is expected to rise to 30 kW by 2027 as AI workloads increase and training models like ChatGPT can consume more than 80 kW per rack. To stay ahead, operators must build facilities that are both adaptable and efficient. This means investing not only in advanced cooling but also in low-heat, high-efficiency networking that minimises thermal output from the outset. Modular cooling designs, flexible pod layouts, and infrastructure that supports new thermal management techniques are now essential for long-term success.
Cooling has always been central to data centre performance, but its role is evolving. By embracing next generation switching technologies alongside leading cooling systems, operators can dramatically improve both energy efficiency and thermal performance.
Ultimately, the next generation of data centres will be defined not just by how much compute they can deliver, but by how intelligently and sustainably they can keep their cool.
