Building the future: AI supercomputing in Taiwan
Foxconn Teams Up with Nvidia for Taiwan's Fastest AI Supercomputer
In an exciting development for AI enthusiasts, Foxconn and Nvidia have announced the construction of Taiwan's largest and fastest supercomputer, utilizing cutting‑edge Nvidia Blackwell GPUs. The Hon Hai Kaohsiung Super Computing Center promises over 90 exaflops of AI performance, setting new benchmarks in speed and efficiency. This supercomputer will accelerate advancements in cancer research, language model development, and smart city technology, reinforcing Taiwan's position as a leader in AI innovation. Set for completion by 2026, the project underscores Foxconn's commitment to smart manufacturing, cities, and electric vehicles.
Nvidia and Foxconn have announced a significant partnership to build Taiwan's largest and fastest supercomputer, a move set to bolster the nation's position in the global AI industry. The supercomputer, called Hon Hai Kaohsiung Super Computing Center, is set to be the most powerful in Taiwan, leveraging the advanced capabilities of Nvidia's Blackwell GPU architecture. Once completed, this high‑performance computing facility is expected to exceed 90 exaflops of AI performance, establishing a new benchmark within the region.
Foxconn’s strategy for the new supercomputer centers around advancing AI‑driven industries in Taiwan, with major applications planned in cancer research, large language model development, and smart city innovations. The supercomputer will be crucial for achieving breakthroughs in these areas, positioning Taiwan as a leader in tech innovation on a global scale.
Central to this development is Foxconn's three-platform strategy, which focuses on integrating smart manufacturing, smart cities, and electric vehicles. Through the supercomputer, Foxconn aims to optimize digital twins, robotic automation, and smart urban infrastructure, thereby enhancing AI‑assisted services in urban regions like Kaohsiung.
Construction of the supercomputer has already begun in Kaohsiung, Taiwan, with the first phase expected to be operational by mid‑2025 and full deployment targeted for 2026. This timeline aligns with Foxconn's vision to integrate advanced technologies such as Nvidia Omniverse and Isaac robotics platforms into their projects, transforming manufacturing and urban services through sophisticated AI and digital twin technologies.
Foxconn's new AI supercomputer will feature Nvidia's GB200 NVL72 platform, comprising 64 racks and 4,608 Tensor Core GPUs. The configuration utilizes 36 Nvidia Grace CPUs and 72 Nvidia Blackwell GPUs per rack, interconnected via Nvidia's NVLink, delivering immense bandwidth for efficient AI model training and complex inference tasks. This setup allows the supercomputer to handle trillion‑parameter models, making it one of the most powerful AI systems globally.
The choice of Taiwan for this massive project underscores the country's strategic importance in the technology and electronics sectors. Foxconn, known for its expansive manufacturing capabilities, is leveraging this opportunity to further establish Taiwan's footprint in AI innovation, integrating AI seamlessly into various sectors such as manufacturing and urban development.
Foxconn's collaboration with Nvidia also reflects a broader trend of major electronics manufacturers embracing AI to transform traditional industries. This move is part of a broader initiative to develop digital twins of Foxconn's factories using Nvidia Omniverse, facilitating enhanced efficiency and innovation in production lines for smart manufacturing, intelligently networked electric vehicles, and smart city projects.