The new supercomputer is scheduled to come online in Q1 of 2025. The project, named "PanaAI AUS AISF", will offer cloud services to public and private organisations, enabling the development of large foundational models for generative AI and accelerating advancements in research, innovation, and societal applications.
The PanaAI AUS AISF system will host up to 4,088 NVIDIA H200 Tensor Core GPUs, interconnected through the NVIDIA Quantum-2 InfiniBand platform, to meet the increasing computational demands of AI workloads. The theoretical peak performance is projected to reach approximately 16.4 exaflops in half-precision (FP8). PanaAI AUS AISF is anticipated to become the largest AI supercomputing cluster in Australia and the fastest AI supercomputer in the Southern Hemisphere. It will play a crucial role in addressing Australia's growing demand for computing resources to support advanced AI workloads across both the public and private sectors.
PanaAI AUS AISF utilises NVIDIA's Hopper architecture-based H200 Tensor Core GPUs. Compared to the H100, the H200 offers 1.8x larger memory (141 GB) and 1.4x higher bandwidth (4.8 TB/s), enabling it to accommodate larger model sizes and datasets. This reduces the need to continuously retrieve data from slower external memory, without relying on complex techniques such as tensor parallelism (data splitting) or pipeline parallelism (staged processing). Additionally, the H200 delivers up to a 45% performance improvement over the H100 in Llama and GPT workloads.
The establishment of PanaAI AUS AISF aims to meet the demands of large-scale projects leveraging artificial intelligence, with a priority on the highest level of security. PanaAI AUS AISF will provide users with accelerated computing cloud services, including public cloud services, AI compute power rentals, and subscriptions for large language model training. With its powerful computing performance, it is expected to rank among the top 10 most powerful AI supercomputers in the world (based on the TOP500 rankings as of June 2024). Its goal is to accelerate research and innovation across various fields, including healthcare, energy development, financial services, autonomous driving, life sciences, and public services.
PanaAI stated: "To meet the growing demand for generative AI and computing resources, PanaAI is planning to establish multiple world-class, powerful supercomputing systems to provide unparalleled accelerated services to users worldwide. We highly value our collaboration with NVIDIA to deliver AI cloud services and rapidly train large foundational models built on ever-expanding datasets of images, acoustics, and sensor information. We look forward to closer collaboration to support users around the globe."
About PanaAI
PanaAI is an AI computing service and operations company specializing in providing GPU capabilities for high-performance computing, machine learning, image rendering, and more. Additionally, it offers full-stack services for users, including the design, implementation, maintenance, and operation of AI compute factories. As an NVIDIA NCP partner and a member of the NVIDIA ecosystem, PanaAI is dedicated to driving the AI revolution by ensuring innovators and developers have the computational resources needed to shape the future. Leveraging world-class, powerful supercomputing systems, PANAAI delivers unparalleled accelerated computing services to users worldwide.
About Junee Limited
Junee Limited (Nasdaq: JUNE) and its wholly owned subsidiary, ASPAC AI Computing Pty Ltd (ASPAC), are innovative technology companies specializing in supercomputing infrastructure. They are dedicated to advancing artificial intelligence and high-performance computing. With the surging global demand for GPU-accelerated cloud computing and large-scale computation, they continue to invest in cutting-edge infrastructure, providing robust computational power to industry users, startups, government agencies, and research institutions.
