As artificial intelligence accelerates, so too does our responsibility to scale and support it sustainably, writes LiquidStack’s Director of Offer & Capture Management, Kevin Roof. The global data centre liquid cooling market is forecast to triple by 2030, reaching nearly $16 billion, fuelled by the explosion of AI workloads and next-generation GPUs that will be capable of drawing multi-kWs each. With this surge, operators face a dual imperative: keep pace with performance while proving climate alignment through measurable reductions in energy and water consumption.
Kevin Roof
Every new generation of AI hardware drives greater computational performance and with it, greater thermal demand. This exponential rise in heat output has led to the end of the air-cooled era.
The future belongs to liquid cooling: cleaner, denser, and vastly more efficient. Compute density has already surged past 120 kilowatts per rack and is forecast to exceed 600 kilowatts within just a few years. Traditional cooling can’t keep up. A new class of thermal infrastructure is needed and crucially, one that scales as fast as AI itself, without increasing its environmental cost.
Before the solution comes the challenge: data centres today are not just power hungry, they are space, water and resource constrained. At the edge, facilities are squeezed by urban limits; at scale, they’re under pressure from governments and investors demanding proof of sustainability. Cooling has become not just a technical issue, but a licence to operate.
It’s not just about removing heat but about orchestrating it intelligently, efficiently, and sustainably. Where conventional systems reject heat as waste, liquid cooling reclaims it as a resource, enabling low-temperature heat recovery and circular energy models that support district heating or industrial reuse. This shift from cooling hardware to cooling platforms is what defines the next phase of AI infrastructure.
LiquidStack developed GigaModular™, a new generation of coolant distribution architecture, built to allow data centre operators navigate this pivotal moment within the industry. Scalable from 2.5 to 10 megawatts, its modular design lets operators expand capacity in tune with real demand. By presenting a genuine pay-as-you-grow model paves the way to eliminate the waste of overbuilt infrastructure and minimise the risk of over investment. The system’s intelligent separation of pump and control modules streamlines maintenance, while front-access serviceability allows deployment even in tight or urban environments. Each unit is engineered for lifecycle efficiency. From manufacturing to operation, the GigaModular™ cuts embodied carbon, conserves materials, and simplifies maintenance.
But what truly differentiates GigaModular™ is how it behaves within the wider data centre ecosystem. Unlike isolated CDUs stitched together on site, it operates as a single, resilient system. Its synchronised controllers and dual-ring MRP loop ensure continuous uptime and load balancing, with no single point of failure. Every module communicates in harmony, adjusting dynamically to changing thermal conditions and ensuring that performance scales as seamlessly as the compute it protects.
As data centres evolve into globally distributed, AI-driven ecosystems, thermal strategy becomes as critical as compute strategy. Whether in hyperscale campuses or space-constrained edge sites, operators will be measured not only by how much power they can deliver, but by how sustainably they manage every watt.
Liquid cooling is the foundation for that future – enabling performance at density, sustainability at scale, and reliability by design. Cooling is no longer the supporting act—it’s the enabler of AI’s next great leap.
Because in the age of intelligent compute, keeping it cool isn’t just an engineering challenge, it’s the defining capability of the digital era.
Learn more here: https://liquidstack.com/
For more news: https://essmag.co.uk/category/news/
