Toronto, Ontario -- Subaru has partnered with Hewlett Packard Enterprise to develop its next-generation EyeSight driver‑assist system, the company announced.
The collaboration will leverage HPE’s Cray XD670 servers to enhance AI model training, algorithm refinement, and image recognition performance.
The upgraded system is designed to improve preventive safety by measuring distance to objects precisely and classifying them using AI, based on images from Subaru’s stereo camera technology. The rollout aligns with Subaru’s goal of achieving zero fatal traffic accidents by 2030HPE Cray XD670 servers feature energy-efficient direct liquid cooling and high‑performance GPU acceleration. That infrastructure is expected to significantly speed development of AI algorithms for the next‑gen EyeSight system.
Subaru Lab, established in December 2020, serves as the automaker’s in‑house hub for ADAS and AI‑based image recognition development. Engineers collect driving video globally to optimize inference models, which will now run on HPE’s NVIDIA H200 GPU‑powered Cray servers with NVLink support across up to four GPUs.
Hirokazu Mochizuki, Managing Director for HPE Japan, highlighted Subaru’s “unwavering dedication” to high-performance safety and noted the “remarkably low rear‑end collision rate” of EyeSight-equipped vehicles. He said HPE is honored that Subaru selected Cray XD670 to bolster its next-generation safety capabilities.
Takashi Kanai, Deputy Chief of Subaru Lab, emphasized Subaru’s All‑Around Safety philosophy—spanning Preventive, Primary, Active, Passive and Connected Safety—and stated that the new setup delivers “twice the performance” of its prior systems, enabling faster iteration and more efficient use of GPU resources
