NVIDIA and Foxconn have joined forces to build the fastest AI supercomputer in Taiwan, powered by Blackwell GB200 NVL72 servers.
The Supercomputer at Kaoshiung will be Taiwan's fastest based on the NVIDIA GB200 NVL72 platform
Taiwan's electronics manufacturer, Foxconn has announced the commencement of building the most powerful AI supercomputer in the country through collaboration with NVIDIA. The new supercomputer will be based on NVIDIA's upcoming and the most powerful Blackwell architecture for powering the system.
In a blog post, NVIDIA announced that Foxconn has collaborated with it and plans to build the supercomputer at Hon Hai Tech Day. The supercomputer will use the latest GB200 NVL72 platform at its core, which will combine the power of Grace CPUs and Blackwell data center GPUs.
As per the report, the new supercomputer will deliver over 90 exaflops of AI performance, which is the fastest Taiwan has ever seen. This will position it as a global leader in AI-driven industries and will equip the country to perform advanced operations such as cancer research, large language model development, and smart city innovations.
The VP and Spokesperson at Foxconn, James Wu said,
Powered by NVIDIA’s Blackwell platform, Foxconn’s new AI supercomputer is one of the most powerful in the world, representing a significant leap forward in AI computing and efficiency,
Foxconn is working on a three-platform strategy that revolves around manufacturing, cities, and electric vehicles. The new supercomputer will help it develop AI-assisted services for urban areas. With the power of the GB200 NVL72 platform, it will be able to utilize up to 36 Grace CPUs and 72 Blackwell GB200 GPUs on a single rack through a 72-GPU NVLink domain.
As per NVIDIA, each rack can deliver up to an incredible 3240 TFLOPS of FP64 and FP64 Tensor Core performance and will deliver up to 13.5 TB of HBM3e memory for a massive 576 TB/s of memory bandwidth. Compared to the NVIDIA H100 Tensor Core GPU, the GB200 NVL72 can deliver 30x higher LLM Inferences, 4x better LLM training, and 18x better data processing through the CPU, all of that while being 25x more powerful and efficient than H100.
Since the GB200 NVL72 will bring up to 64 racks, there will be a total of 4608 Blackwell GB200 GPUs for high-performance computing needs. This will make it ideal for training large AI models and performing complex inference operations on trillion-parameter models.
As per the report, Foxconn is set to launch the supercomputer in the first phase by mid-2025 and will fully deploy it by 2026. The supercomputer will use several NVIDIA technologies such as NVIDIA Omniverse, Isaac robotics platforms, and Digital Twins, which will offer various NVIDIA tools and libraries for the better manufacturing process.
News Source: NVIDIA
About the author: Sarfraz Khan is a hardware reporter with a focus on PC components and the builder community. With years of experience writing about PC hardware and laptops, his work has been featured on several reputable technology publications. Sarfraz's hands-on experience is demonstrated through his first-person accounts of using and comparing different hardware configurations, providing practical and relatable insights for everyday users. His technical analysis is respected by peers in the enthusiast community and has been cited by specialized hardware sites such as Germany's Igor's Lab.
Follow Wccftech on Google to get more of our news coverage in your feeds.
