2025 GLOBAL FINALIST
HyperAccel Co., Ltd.
Chips & Fab
LLM Processing unit providing batched inference at 1/10 coast and 1/3 power Our LLM Processing Unit (LPU) is a next-gen AI processor designed for the LLM service era, delivering magnitude gains in cost and energy efficiency through unique design from deep LLM understanding. Our Streamlined Dataflow Architecture achieves unprecedented tokens-per-TFLOPS maximizing bandwidth and logic utilization across every stage of inference. Our SDK intelligently orchestrates multi-LPU systems for performance and developer usability, fully compatible with frameworks like vLLM and PyTorch.