Tech news in 3 minutes

The Yield Partnership: Intel and PDF Solutions Tackle Advanced Nodes

29 d ago

James Regan, Co-Founder and CEO of Oriole, is a seasoned technology executive and physicist with over 30 years in optical communications. Oriole, founded from University College London after 20+ years of research, is building a next-generation data center network to scale AI. The company starts with the problem—how to scale AI—and works backwards, creating a pure photonic network that keeps data as light moving through the data center. This full-stack system delivers higher performance at lower power, enabling AI to scale without physical or economic limits. Oriole solves the core problem of AI hitting infrastructure limits. Existing electrical switching architectures are bottlenecks, introducing latency that reduces tokens per second and revenue. The company reimagines the network from first principles, allowing data to move as light directly between processors with extremely low, deterministic latency. This drives higher tokens per second and more value from the same compute. The primary focus is large-scale AI training and inference, where data movement is the limiting factor. Beyond hyperscale AI, there is interest from financial services. Customers worry about scale and efficiency—whether they can afford to scale AI and if infrastructure can keep up. The competitive landscape features incremental improvements like faster electrical switches or hybrid approaches. Oriole differentiates architecturally with the world’s first pure photonic AI network: data stays as light end-to-end, every node connects directly to every other in a single-hop, fully connected, contention-free topology, removing constraints entirely. Key technology includes PRISM (Photonic Routing Infrastructure for Scalable Models), already in real-world deployments, and PRISM Ultra, announced in March this year, which pushes light closer to the processor for greater scale and performance. Innovation continues across software, control, and photonic hardware. Customers typically engage through initial conversations to understand the novel solution. For PRISM, it's a transactional model with data center builders. For PRISM Ultra, it's strategic partnerships with hyperscalers. Engagement is deep, involving a "staircase" from demonstration to scaling to full production deployments.