AMD plans to invest up to £2 billion (US$2.7 billion) over the next five years to strengthen the United Kingdom’s AI ecosystem through new infrastructure, research collaborations, and workforce development initiatives. Speaking at the London Tech Week conference, AMD Chair and CEO Lisa Su outlined a strategy aligned with the UK government’s AI Opportunities Action Plan and AI Hardware Strategy, focusing on sovereign AI capabilities, scientific computing, and advanced research.
A major component of the initiative centers on expanding national AI computing infrastructure. AMD and Dell Technologies are supporting two new UK supercomputing systems at the University of Cambridge. The Zenith AI supercomputer, funded by the UK Department for Science, Innovation and Technology (DSIT) and UK Research and Innovation (UKRI), is designed as a national AI-for-science platform. A second system, Sunrise, is being developed in partnership with the United Kingdom Atomic Energy Authority to support fusion energy research. Both systems will use AMD Instinct GPUs, AMD EPYC CPUs, and AMD software to support applications spanning healthcare, climate science, materials research, engineering simulation, fusion research, and scientific AI model development.
AMD also announced new research partnerships with Imperial College London and Oriole Networks. The Imperial collaboration will focus on computational science, healthcare innovation, climate modeling, AI optimization, and data-intensive scientific workflows. Meanwhile, AMD and Oriole Networks are collaborating on the UK’s Advanced Research and Invention Agency (ARIA) Scaling Inference Lab, a £50 million initiative aimed at addressing AI infrastructure bottlenecks. The project combines Oriole’s PRISM photonic networking architecture with AMD Instinct GPUs and EPYC processors to evaluate new approaches for scaling AI inference workloads with lower latency and improved energy efficiency.
- AMD plans to invest up to £2 billion in the UK over five years.
- New partnerships include Imperial College London and Oriole Networks.
- Zenith AI supercomputer will support national AI-for-science initiatives.
- Sunrise AI supercomputer will support UK fusion-energy research.
- AMD hardware will underpin sovereign AI infrastructure projects across research, healthcare, climate science, and engineering.
- Oriole Networks and AMD will deploy what they describe as the world’s first large-scale AI system powered by a pure photonic network.
- The ARIA Scaling Inference Lab is backed by £50 million ($68 million) in UK government funding.
- Oriole’s PRISM architecture replaces electronic switching in the network core with nanosecond-scale optical circuit switching.
- Oriole says its photonic network can reduce core network power consumption by up to 81% while increasing AI inference throughput.
“The United Kingdom has the talent, research excellence and ambition to help lead the next era of AI,” said Lisa Su, chair and CEO of AMD. “AMD is proud to deepen our commitment to the UK and work with partners across government, academia and industry to expand access to the compute infrastructure needed to advance sovereign AI, accelerate discovery and drive long-term economic growth.”
🌐 Analysis: Oriole Networks, a start-up based in London, aims to bypass traditional AI cluster design limitations by eliminating from the network core. In conventional topologies using InfiniBand or Ethernet, repeated optical-to-electrical-to-optical (OEO) conversions introduce serialization delays and tail-latency spikes. The company’s PRISM (Photonic Routing Infrastructure for Scalable Models) architecture replaces active electronic switching with a passive optical router core. By establishing direct optical paths between nodes, Oriole says it can dramatically reduce communication-dependent GPU idle times, which can otherwise severely bottleneck large-scale workloads.
To handle dynamic AI traffic without relying on electrical packet buffers, the PRISM architecture combines nanosecond-scale optical circuit switching across three dimensions: time, wavelength, and space. According to the company, this multi-dimensional switching allows for rapid, on-the-fly circuit reconfiguration optimized for the heavy all-to-all communication patterns characteristic of large language models. Furthermore, Oriole claims its approach can facilitate a 1-hop network diameter capable of interconnecting up to one million endpoints. By shifting control logic to the edge via integrated photonic switches and tunable transceivers, the company projects core network power consumption can be reduced by up to 81%.
A key strategic element of the PRISM architecture is its xPU-agnostic design. Rather than relying on proprietary interconnects that lock operators into a specific silicon ecosystem, Oriole decouples the transport plane from the compute plane. The company states that its system integrates into existing host software stacks via native PCIe drivers and custom acceleration libraries (such as NCCL for NVIDIA or RCCL for AMD), allowing it to support multiple hardware platforms without altering underlying AI frameworks. The upcoming deployment at the UK’s ARIA Scaling Inference Lab will serve as a critical industry validation point, testing whether pure photonic networks can deliver deterministic performance and open up proprietary compute silos at production scale.
| Profile: Oriole Networks | |
|---|---|
| Headquarters | London, United Kingdom |
| CEO | James Regan |
| Focus | Photonic networking for AI infrastructure |
| Flagship Technology | PRISM Photonic Networking Platform |
| Architecture | Nanosecond optical circuit switching replacing electronic network core switches |
| Key Claim | Up to 81% reduction in network core power consumption |
| Deployment | ARIA Scaling Inference Lab |
| Commercial Milestone | First commercial deployment announced in 2026 |
| Industry Rollout | Broader deployment planned for 2027 |
| Technology Positioning | xPU-agnostic AI networking infrastructure |