HyperAccel, an AI semiconductor company founded in 2023, is a startup developing the 'LPU (LLM Processing Unit)', a semiconductor specialized for large-scale Gen-AI. The LPU is a semiconductor that optimizes memory bandwidth, crucial for large-scale generative AI computations, and integrates AI-specific logic. HyperAccel supports the advancement of generative AI technology and infrastructure development by accelerating inference in Large Language Models (LLMs). The company is challenging innovation in the AI semiconductor field across semiconductor chips, servers, and IP licensing.

The LPU (LLM Processing Unit) is an AI semiconductor chip capable of accelerating Large Language Model (LLM) inference. It can accelerate the entire inference process of transformer-based generative large language models (such as GPT, Llama, etc.).
We are currently selling Orion, a server appliance product that utilizes this LPU chip. Orion server uses eight LPU semiconductor chips (FPGA) to support AI inference with high efficiency and low power consumption.