Company Overview
Neurophos addresses the growing demand for computational power in AI by pioneering a revolutionary Optical Processing Unit (OPU) using metamaterials, silicon photonics, and analog computing. The OPU’s Compute-In-Memory (CIM) architecture integrates memory and processing into a single unit, reflecting a design mimicking the brain’s efficiency, enabling rapid, energy-efficient processing directly within the memory array. This delivers unprecedented gains in processing speed, energy efficiency, and computational density. Simulations showed that their existing optical ‘transistors’—which are already 10,000× smaller than traditional ones— enable a chip delivering over 1 ExaOPS (one million TOPS), equivalent to a thousand GPUs. This is achievable using current technology, with no need for further miniaturisation or next-gen nodes.
Thesis Fit
GEOINT relies heavily on processing large amounts of data, particularly imagery and sensor data, to derive insights about the physical and environmental conditions of specific geographic areas. This often requires substantial computational power, especially for tasks like image recognition, data fusion, and real-time analytics. Neurophos’ focus on AI inference accelerators using optical computing and metamaterials to improve the computational efficiency of AI hardware has a direct impact as it can significantly optimize the compute capabilities required for large scale geospatial processing to achieve higher speed and energy efficiency. In addition, Neurophos’s technology can also potentially support the high-demand computational tasks in orbit, crucial for real-time satellite data processing, as well as communication systems.
