ARIA’s Scaling Inference Lab Deploys Lumai’s Optical Compute Tech

UK research agency selects light-based processor for inference lab

By GPU Data Hub Desk · GPU Data Hub·Editorial standards·3 Nov 2026, 20:22·United Kingdom

ARIA’s Scaling Inference Lab Deploys Lumai’s Optical Compute Tech

The UK's Advanced Research and Invention Agency will install Lumai's three-dimensional optical compute technology at its Scaling Inference Lab. Lumai uses light-based processing to conduct artificial intelligence inference, claiming up to a 90 per cent drop in data centre electricity consumption alongside 50-fold performance gains. The initiative forms part of state-backed research into next-generation processing platforms designed to curb infrastructure energy demands.

Original source

This summary was written by the GPU Data Hub desk from reporting published by HPCwire on 3 Nov 2026, 20:22. Read the full article at the original publisher.

Read Original Source →

Companies

ARIA Lumai

Topics

optical compute AI inference energy efficiency

LinkedIn X

More from GPU Data Hub

Latest →

AMD rides data center riches to ‘largest and most profitable’ quarter in history

AMD generated a record $13.4 billion in quarterly revenue, marking the most profitable period in the semiconductor firm's history behind surging data centre demand. Chief executive Lisa Su indicated that momentum remains robust, with fourth-quarter sales projected to reach $17 billion.

Semiconductors·4h ago·The Verge

AMD stock jumps 8% on earnings as data center revenue nearly doubles

Shares in AMD climbed eight per cent following the release of third-quarter financial figures that exceeded analyst expectations. The chipmaker's stronger performance was driven by accelerating demand for artificial intelligence hardware, which saw data centre division revenue almost double.

Semiconductors·6h ago·CNBC Technology

Randomized Caches in HW: Overheads And Limitations (RUB, Max Planck, Georgia Tech)

Researchers from Ruhr University Bochum, the Max Planck Institute, and Georgia Tech have assessed the practical trade-offs involved in implementing randomised caches in hardware. The technique mitigates side-channel security threats by masking the correlation between memory addresses and physical cache positions. The authors note that while previous proposals tackled architectural behaviour, physical hardware implementation obstacles have remained largely unaddressed.

Semiconductors·11h ago·Semiconductor Engineering

Fast, Sub-Volt HZO Ferroelectric Capacitors For DRAM-Like Memory (SNU)

Researchers led by Seoul National University have developed sub-volt ferroelectric capacitors designed for DRAM-style memory applications using hafnium zirconium oxide. The development resolves long-standing trade-offs between switching speeds and cycling endurance at low voltage levels.

Semiconductors·12h ago·Semiconductor Engineering

Foxconn details quantum ambitions, wants to become world’s “largest quantum computer manufacturer”

Electronics assembly specialist Foxconn has announced a long-term goal to emerge as the pre-eminent manufacturer of quantum computing hardware globally. The corporation cautioned, however, that commercial viability for these systems is not anticipated for at least two years.

Semiconductors·13h ago·Data Center Dynamics

Intel 14A moves closer to production as PDK 0.9 becomes available — plans to release PDK 1.0 in the first quarter of 2027

Intel has achieved a key technical milestone for its upcoming 14A manufacturing node with the availability of process design kit 0.9, ahead of a planned 1.0 release in early 2027. The company stated that the next-generation process is tracking in line with yield and performance projections. The roadmap highlights the foundry division's progress towards restoring its competitive edge in cutting-edge fabrication.

Semiconductors·14h ago·Tom's Hardware