REHOVOT, Israel, Sept. 24, 2026 — CollPlant Biotechnologies Ltd. today announced the publication of joint research conducted by its subsidiary LightSolver Ltd., a pioneer of pure photonic computing and developer of the Laser Processing Unit (LPU), and the High-Performance Computing Center Stuttgart (HLRS), one of Europe’s premier supercomputing institutions.

Published in the proceedings of the 23rd ACM International Conference on Computing Frontiers, the research demonstrates the potential of LightSolver’s all-optical computing architecture to dramatically accelerate one of the most fundamental and computationally intensive workloads in high-performance computing (HPC): solving large-scale sparse systems of linear equations.

In the joint study, researchers from LightSolver and HLRS benchmarked projected LPU performance, using an emulator of LightSolver’s LPU architecture, against established state-of-the-art iterative algorithms running on a GPU. Across the benchmark problems evaluated, the study projected that the LPU could achieve time-to-solution acceleration ranging from approximately 40× to more than 80,000×, depending on the algorithm and problem evaluated.

Large-scale linear systems sit at the mathematical core of some of the world’s most demanding scientific and engineering simulations, including computational fluid dynamics, structural mechanics, electromagnetics, molecular dynamics and materials science. Solving these systems can dominate runtime and energy consumption in HPC applications, making them a critical computational bottleneck and an important target for next-generation computing architectures.

The findings highlight the potential for photonic processors to serve as specialized accelerators for large-scale linear systems, particularly for structured problems and computational workloads in which these mathematical operations must be solved repeatedly.

Executive Commentary

“This joint work demonstrates how an emerging photonic computing architecture can be evaluated against established numerical methods and modern GPU platforms,” said Prof. Michael Resch, Director of the High-Performance Computing Center Stuttgart (HLRS). “The results provide valuable insight into where photonic computing may offer advantages for high-performance computing workloads and how it could be integrated into future hybrid computing systems.”

“This research marks an essential milestone in validating the LPU as a transformative computing layer alongside traditional CPUs and GPUs,” said Dr. Ruti Ben-Shlomi, CEO and Co-Founder of LightSolver. “Working with HLRS and its world-class researchers gave us an exceptional opportunity to benchmark and validate our technology against established computing platforms and test it on some of the most demanding computational problems. Collaborations like this are invaluable as we continue to expand the range of problems the LPU can address and demonstrate its potential across science, engineering and high-performance computing. We look forward to continuing to work with leading research institutions around the world to validate, challenge and advance this new computing architecture.”

The Future of HPC: Heterogeneous Hybrid Architectures

The published research reinforces LightSolver’s vision for the future of enterprise and scientific computing: a heterogeneous hybrid architecture in which CPUs, GPUs and LPUs work together, each executing the workloads best suited to its underlying architecture:

• CPUs provide general-purpose processing, orchestration and system control.
• GPUs accelerate massively parallel digital computing and tensor-based workloads.
• LPUs serve as specialized photonic accelerators, offloading computationally intensive mathematical workloads, including linear and differential equations and optimization-where LightSolver’s optical architecture can provide significant acceleration.

This hybrid approach is designed to enable supercomputing centers, data centers and cloud providers to integrate photonic acceleration into existing computing environments, potentially delivering dramatic reductions in time-to-solution, energy consumption, computing costs and carbon footprint for targeted workloads.