We're building a new kind of computing architecture designed specifically for the AI inference bottleneck. Our approach integrates memory and compute in a way that eliminates the data movement overhead that limits traditional GPUs. The team spans from transistor-level circuit design up to cloud inference server logic - we work across the full stack because that's the only way to solve this problem properly.
Founded in 2022 and operating out of London and Bristol, we emerged from stealth in 2024 with backing from NATO's Innovation Fund, Oxford Science Enterprises, and Kindred Capital. Our goal is straightforward: run frontier AI models 25x faster at 1/10th the cost. We're taking a hardware-software co-design approach, building our own software stack alongside custom silicon rather than trying to patch over the limitations of general-purpose hardware.