Microsoft researchers at Cambridge have developed a prototype analog optical computer (AOC)—a computing device that leverages light (via micro‑LEDs, optical lenses, and camera sensors) rather than traditional binary electronics to perform computations. Their goal was to utilize commercially available components to build a system that could operate at room temperature, be energy-efficient, and scalable. The AOC has already demonstrated its capabilities by solving two real‑world optimization problems: one involving complex interbank transaction settlement with Barclays, and another in reconstructing MRI scans with the potential to reduce scan times dramatically—from about 30 minutes down to five—using a “digital twin” model to simulate runs at scale. Additionally, the architecture shows promise for AI inference, potentially enabling energy‑efficient reasoning tasks that current GPU-based systems struggle with, thanks to its unique design and fixed-point computation approach. The work, published on September 3, 2025, includes sharing of the optimization algorithm and digital twin to foster broader research collaboration.
Share This Article
