Oriole Networks
London photonics startup building light-based chip interconnects for AI data centres.
Last refreshed: 14 June 2026
Can photonic interconnects cut AI data-centre energy by 81%, and has Oriole proved it at scale?
Timeline for Oriole Networks
Oriole moves data with light, not copper
UK Startups and Innovation- What are photonic AI interconnects and why do they matter?
- Photonic interconnects transmit data between AI chips using pulses of light rather than copper-wire electrical signals. They consume FAR less energy and generate less heat, addressing a key bottleneck in scaling large AI accelerator clusters.Source: Lowdown
- How much energy does Oriole Networks claim to save with its photonic interconnect?
- Oriole Networks claims its photonic AI interconnect cuts data-centre transmission energy by 81% compared with copper-wire alternatives, based on its deployment with AMD through ARIA's Scaling Inference Lab.Source: Lowdown
- Who is Oriole Networks partnering with and what is ARIA?
- Oriole Networks partnered with AMD to deploy its photonic interconnect, using ARIA's Scaling Inference Lab as the testbed. ARIA is the Advanced Research and Invention Agency, the UK's high-risk research funder modelled on DARPA.Source: Lowdown
Background
Oriole Networks deployed what it claims is the world's first large-scale photonic AI interconnect in June 2026, partnering with AMD through ARIA's Scaling Inference Lab. The deployment replaces copper-wire data transmission inside AI data centres with light-based photonic links, with Oriole claiming an 81% reduction in data-centre transmission energy. The technology targets the interconnect bottleneck that limits how many AI accelerators can be clustered together at scale.
Oriole Networks is a London-based photonics startup working on optical chip interconnects for AI infrastructure. Photonic interconnects transmit data as pulses of light rather than electrical signals, dramatically reducing energy consumption and heat output at high bandwidth. The AMD partnership provides validation alongside a hyperscaler-adjacent customer and the ARIA Scaling Inference Lab supplies the testbed environment for real-world deployment at data-centre scale.
If Oriole's energy claims hold under independent verification, photonic interconnects could become a significant lever for data-centre operators facing both energy cost pressure and regulatory scrutiny of AI's power consumption. The timing aligns with growing UK and EU focus on AI energy efficiency as a policy concern. The AMD partnership positions Oriole in the AI infrastructure supply chain at a moment when GPU-cluster efficiency is a first-order competitive concern.