AI Processing Unit (AIPU)

AI Processing Unit Inference Accelerator

Designed for AI inference acceleration in HPC applications


Objectives

  • Development of a high-performance, high-energy-efficiency and scalable AI Inference chiplet ​

    • AI accelerated HPC (weather models, materials discovery, ​
      search requests, PINNs etc.)​
    • Large-scale neural networks, i.e., Generative and ​
      Foundation models​

  • AI TA and the GPP TA will build a complete system-in-a-package (SiP) that includes:​

    • AIPU chiplet(s) for AI inferencing​
    • GPP chiplet(s) for classical HPC workloads​

  • Develop a complete compiler and software stack based ​on the RISC-V architecture and Axelera AI’s infrastructure​

    ​

Explore more innovations