HexSeed Technology has raised more than £600,000 — around €700,000 — in early-stage funding to develop diamond-coating technology aimed at the thermal problems now constraining AI data-centre power systems.
Carbon13 led the round, with participation from the Net Zero Technology Centre and Vento Ventures. The company has also secured an Innovate UK Partnership Grant.
The bottleneck is heat
As data-centre power density has climbed, the practical ceiling on performance has moved from silicon capability to the ability to remove waste heat. Every watt dissipated as heat is a watt paid for twice — once in the power bill and again in the cooling load — and thermal limits force designers to derate components that would otherwise run harder.
Diamond is the most thermally conductive bulk material available, which makes it an obvious candidate for spreading heat away from a semiconductor junction. The obstacle has always been process temperature: conventional diamond growth is hot enough to damage the finished devices it would be applied to. HexSeed's stated approach is low-temperature deposition, which would allow diamond layers to be applied to devices that could not survive existing methods.
Gallium nitride and the next step
The company's initial target is gallium nitride, the wide-bandgap semiconductor increasingly used in high-efficiency power conversion. GaN switches faster and loses less energy than silicon, but it concentrates heat in a very small area — precisely the failure mode a high-conductivity spreading layer is meant to address.
HexSeed's work draws on research collaboration with the University of Bristol. Its next milestone is demonstrating the coating on commercial GaN devices. Until that is complete, the performance benefits remain an engineering objective rather than a proven result, and no efficiency figures have been published.
Why early materials companies attract patient capital
Deep-tech materials businesses present a distinctive risk profile. Development timelines are long, capital requirements step up sharply at the pilot-production stage, and commercial validation depends on qualification cycles controlled by customers rather than the company itself. Against that, a process that works becomes very difficult to displace, because it is designed into products with lifecycles of a decade or more.
That combination — slow to prove, durable once proven — is one reason such companies frequently appear in the portfolios of family offices and mission-linked investors, whose holding periods are not set by fund life. Carbon13's participation reflects the climate framing: reducing the energy overhead of computing infrastructure is one of the more measurable decarbonisation levers available in a sector whose electricity demand continues to rise.
The customer set is unusually concentrated. Power-electronics components reach data centres through a small number of module and system manufacturers, so commercial success depends on qualification by a handful of firms rather than broad market adoption. That concentrates both the opportunity and the risk: a single design win can transform a company's prospects, and a single delayed qualification can consume a year.
The wider context is an infrastructure build-out proceeding faster than the grid connections supporting it. Where power availability, rather than capital, limits how much computing capacity can be installed, efficiency gains have a value beyond the electricity they save — they determine how much can be deployed at a given site at all.
The size of any future funding requirement, and the timetable for commercial deployment, have not been disclosed.
UKFOS editorial · published 23 August 2026