HexSeed raises over €700k to develop diamond-coated GaN cooling for AI data centers
The AMW Read
A new UK deep-tech entrant applies a low-temperature diamond-coating process to finished GaN power devices, addressing a real thermal bottleneck in AI data center power conversion, but the round is small and pre-commercial.
HexSeed raises over €700k to develop diamond-coated GaN cooling for AI data centers
UK-based HexSeed Technology, founded in late 2025, has raised over €700k (£600k) in early-stage funding led by Carbon13, with participation from Net Zero Technology Centre and Vento Ventures. The round also unlocks a previously provisional Innovate UK Partnership Grant. The three-person founding team — CEO Mark Tandy, CTO Dr Leonardo Santoni and COO Dr Michael Glerum — is working with University of Bristol researchers Paul May (diamond materials) and Martin Kuball (semiconductor thermal management) to develop a low-temperature microwave plasma process that grows diamond coatings directly onto finished gallium nitride (GaN) power devices, pulling heat away from their active regions.
The problem HexSeed is targeting sits at the power-conversion layer feeding AI infrastructure rather than the AI chips themselves. GaN transistors convert power more efficiently than the silicon devices they are replacing, but they run hot, and heat is the main constraint on how much of that efficiency gain reaches production. Diamond is the best-known thermal conductor, but growing it conventionally requires temperatures that would destroy a completed semiconductor device — which is why a low-temperature process able to coat finished GaN parts is the specific gap HexSeed says it is closing. The company cites IEA projections of data center electricity demand reaching 945 TWh by 2030 as AI-optimized facilities scale, with a large share of that power currently lost as heat.
HexSeed plans to use the funding to demonstrate the coating process on commercial GaN power devices and move toward pilot customer engagements, naming data center power conversion hardware as its initial target market. This is a pre-commercial, lab-to-device bet: the round is small and no pilot customers are named yet, so the near-term test for builders and investors watching the AI power-infrastructure stack is whether HexSeed can prove device-level reliability and manufacturable cost at the plasma-deposition step before GaN device makers or data center operators redesign power-delivery supply chains around it.