IT Brief UK - Technology news for CIOs & IT decision-makers
United Kingdom
ThermAI turns GPU server waste heat into home heating

ThermAI turns GPU server waste heat into home heating

Wed, 18th Nov 2026 (Today)
Sofiah Nichole Salivio
SOFIAH NICHOLE SALIVIO News Editor

ThermAI has developed a home heating system that uses waste heat from GPU servers, with Airsys backing the project as lead investor and technical partner.

Its HeatBridge system is designed to capture heat from AI and research computing and use it to provide free heating and hot water to participating households. The initial rollout is aimed at social housing residents and other households struggling with energy bills, with pilot projects due to lead to 100 homes in 2027.

The model links a domestic heating installation to a six-GPU ThermAI Node hosted in the home. Universities, research centres, AI start-ups and engineering firms would access the computing remotely through ThermAI's FlightDeck platform, while heat from the server is transferred into a water circuit for household use.

The approach addresses growing debate over the energy demands of AI infrastructure and the amount of heat wasted by conventional server installations. ThermAI and Airsys are betting that the economics of GPU computing can also support cheaper domestic heating, particularly in homes facing fuel poverty.

How It Works

The HeatBridge system is built around ThermAI's LiquidDock server cassette, which uses Airsys' LiquidPack spray-cooling system. Instead of immersing servers in liquid, it sprays an electrically non-conductive fluid directly onto the hardware inside a sealed cassette.

An Airsys heat exchanger then transfers the recovered heat into a water circuit. Outlet water temperatures can reach up to 60C, and Sunamp heat batteries store the energy until it is needed by the household.

The servers are being built by 2CRSi through its Manchester-based subsidiary and use NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, each with 96 GB of memory. ThermAI said the HeatBridge system is UK patent pending.

Participating households would receive heating and hot water free of charge. If backup heating is needed, ThermAI would also cover that cost. The company cited a typical UK household bill of about £144 a month for gas and electricity at current price-cap rates as an indication of potential savings.

Funding and Partners

ThermAI has completed a pre-seed funding round led by Airsys, with participation from venture capital firm Augur VC and angel investors. Financial terms were not disclosed.

Airsys, which specialises in thermal management for digital infrastructure, said the investment marks its first move into heat reuse. It sees the household model as a way to test whether the same approach could later be used at larger sites, including data centres connected to district heating networks.

Christopher Smart, Founder and Chief Executive Officer of ThermAI, said the business was designed from the outset to integrate computing and heat recovery. "We're designing the computing platform and the heating system together, so heat reuse is part of the infrastructure from the start," Smart said. "Our aim is to make that work as a commercially viable model that can scale."

Dan Chapman, President of Augur VC, linked the proposal to both demand for GPU access and pressure on household energy costs. "ThermAI brings together a commercial opportunity in GPU computing and a practical way to make homes more affordable to heat," Chapman said. "We see strong potential in the model and are pleased to support Chris and the team."

Research Pilot

The University of Nottingham is among the organisations lined up for an early compute pilot. Professor Jonathan Hirst of the university's School of Chemistry has signed a letter of intent to take part, with the pilot expected to support GPU-intensive research including drug discovery and molecular dynamics.

The involvement of an academic user is important because the model depends on sustained remote demand for specialist computing. ThermAI is positioning FlightDeck as a UK-hosted platform for repeatable jobs, reusable datasets and clearer visibility of computing costs.

Professor Hirst said the idea could bring local social value as well as research access. "ThermAI's approach is a highly sustainable one that would benefit local communities," Hirst said. "Initiatives like this could play an important role in maintaining and improving the sustainability of organisations such as ours."

Housing Focus

The first housing deployments will focus on social landlords as providers work to improve the energy efficiency of rented homes. In England, landlords are under pressure to bring social rented homes to EPC band C by 2030, with support from the government's Warm Homes Plan.

Smart said lower-carbon housing measures also need to cut running costs for residents. "Decarbonising homes should also make them more affordable to heat, not just improve EPC ratings," he said.

For Airsys, the attraction lies in proving that heat recovery can be designed into computing systems rather than treated as an afterthought. Yunshui Chen, Founder and Chief Executive Officer of Airsys, said: "Heat reuse is the next frontier for thermal infrastructure, and ThermAI shows what it looks like when you design for circularity from the server up. For the first time, our spray-cooling technology is not only keeping servers cool, but turning their heat into a vital utility for the home. If we can prove this circular model in a single household, the same approach can scale to data centres feeding district heating. That is what 'Balance the Environment' demands of us: treating every watt of energy as a resource that does double duty."

Smart said the choice of spray cooling had been central to the concept from an early stage. "When we saw spray cooling and higher coolant temperatures, it instantly clicked," he said. "For something that has to live inside a home, it suits what we're building."