New Photonic Engine Accelerates AI Matrix Calculations
The device uses phase-change materials and light to perform matrix multiplication, bypassing traditional electronic limits.
Updated on Oct. 24, 2026 in Semiconductors
Researchers have developed a non-volatile photonic matrix-vector multiplication engine that utilizes phase-change materials for optical computing. The research-stage device offers a compact solution for accelerating AI workloads by bypassing traditional electronic bottlenecks.
Why it matters
Electronic systems currently struggle with energy efficiency and speed limits in matrix-vector multiplication, the core operation of modern AI models. This photonic approach promises to overcome these bottlenecks by using light-based processing to handle massive computational loads.
The device measures 185 μm by 21 μm and delivers an extinction ratio of 17.3 dB for the TM1 mode at 1550 nm. It supports 1296 distinct configurations using a 40 nm bandwidth.
The details
The engine achieves high-speed processing through mode division multiplexing, where fundamental and first-order transverse magnetic modes (light waves polarized perpendicular to the direction of propagation) carry data. It utilizes twenty Ge2Sb2Se4Te1 phase-change cells—materials that switch between crystalline and amorphous states—to store and modulate weights. Crystallization is triggered by a 5 V, 33 ns pulse, while amorphization requires a 12 V, 5 ns pulse.
Timeline
The research findings were published on October 24, 2026.
The Tech Race
This development addresses the fundamental compute bottleneck in modern AI by shifting the load from standard electricity-based silicon transistors to photonic circuitry. It follows a growing field-wide trend of replacing electronic matrix multiplication with hardware-optimized light-based architectures.
This device remains in the research stage and is not currently available for commercial AI infrastructure. It serves as a proof-of-concept for future photonic chip architectures that could eventually reduce the energy footprint of large-scale data centers.
The takeaway
This engine demonstrates that photonic hardware can replace standard electronic matrix multiplication for high-efficiency AI acceleration. Observers should watch for upcoming benchmarks involving high-speed switching stability in phase-change photonic circuits.
Further reading
Explore deeper into hardware innovations within the Semiconductors sector.
