Summary
The collaboration between Nvidia, Foxconn, and the Taiwan government to create a cutting-edge AI supercomputer represents a landmark initiative aimed at advancing artificial intelligence research, high-performance computing (HPC), and industrial innovation in Taiwan. Centered around the Hon Hai Kaohsiung Supercomputing Center, this project leverages Nvidia’s latest Blackwell GPU architecture and Foxconn’s manufacturing expertise to build a system delivering over 90 exaFLOPS of AI computing power, positioning it among the world’s most powerful AI supercomputers. Supported by substantial government investment and policy backing, the collaboration reflects Taiwan’s strategic ambition to become a global hub for AI-driven technologies across sectors such as healthcare, smart manufacturing, robotics, and smart city infrastructure.
This partnership builds on Nvidia’s broader efforts to decentralize AI hardware manufacturing and strengthen supply chain resilience, incorporating advanced platforms like the GB200 NVL72 and integrating digital twin and robotics technologies through Nvidia Omniverse and Isaac platforms. Foxconn, traditionally known for electronics manufacturing, is expanding into semiconductor packaging, electric vehicles, and AI-enabled smart city projects, using the supercomputer to accelerate innovation in these areas. The facility is expected to enable breakthroughs in large language models, cancer research, and autonomous technologies, while also enhancing Taiwan’s competitiveness in the global AI and HPC landscape.
Despite its promising technological and economic impact, the collaboration raises complex challenges including the intricacies of the AI hardware supply chain, ethical concerns around sovereign AI development, and geopolitical tensions. Critics have warned that the pursuit of national AI infrastructure might exacerbate risks related to information control and global cooperation, potentially empowering authoritarian surveillance practices and fostering isolationism. These issues underscore the broader implications of the project beyond pure technological advancement.
Construction of the supercomputer began in Kaohsiung, with phased operation expected to start by mid-2025 and full deployment targeted for 2026. The project exemplifies a strategic alignment of government, industry, and technology leaders aimed at catalyzing a new industrial revolution driven by AI, positioning Taiwan at the forefront of next-generation computing innovation.
Background
The collaboration between Nvidia, Foxconn, and the Taiwan government to create a cutting-edge AI supercomputer stems from the convergence of Taiwan’s established leadership in semiconductor manufacturing, high-performance computing (HPC), and emerging AI-driven industries. Foxconn, formally known as Hon Hai Precision Industry, has been diversifying its business beyond contract electronics manufacturing into areas such as semiconductor packaging, electric vehicles (EVs), robotics, and smart cities, integrating artificial intelligence into its operations to enhance productivity and innovation.
Nvidia, a global leader in AI hardware and software, contributes its latest technologies including the Grace Blackwell superchip architecture and GB200 NVL72 platform, designed specifically for high-performance AI processing. This partnership aligns with Nvidia’s broader manufacturing decentralization strategy and supply chain resilience efforts, which include collaborations with leading Asian and North American companies such as TSMC, Foxconn, and others. By leveraging these advanced components and Foxconn’s manufacturing expertise, the project aims to develop one of the world’s most powerful AI supercomputers.
Taiwan’s government, recognizing the transformative potential of AI across industries, is actively supporting this initiative through investments and policy support to accelerate AI development and adoption nationwide. This public-private collaboration is intended to position Taiwan as a global hub for AI research and innovation, enhancing its competitive edge in fields ranging from smart manufacturing and digital twins to healthcare and smart urban infrastructure.
The planned AI supercomputer, based at the Hon Hai Kaohsiung Supercomputing Center, will harness Nvidia’s Blackwell architecture to deliver over 90 exaFLOPS of AI computing power—an unprecedented scale that enables breakthroughs in large language model development, cancer research, and autonomous technologies. The facility also exemplifies the integration of digital twin and robotic automation technologies to support Foxconn’s smart factories and smart city projects, reflecting broader industry trends towards AI-assisted services in urban environments like Kaohsiung.
Collectively, this collaboration not only showcases Taiwan’s expanding role in the global AI and HPC landscape but also highlights the strategic alignment of government initiatives, semiconductor manufacturing excellence, and AI hardware innovation to drive the next industrial revolution.
Formation of the Collaboration
The collaboration between NVIDIA, Foxconn, and the Taiwan government was initiated to establish a cutting-edge AI supercomputer aimed at advancing AI research and innovation across multiple industries in Taiwan. This partnership builds upon a longstanding relationship between NVIDIA and Foxconn, first publicly showcased at Hon Hai Tech Day 2023 (HHTD23), where Foxconn demonstrated its use of the NVIDIA Omniverse platform to create 3D digital twins for automated production line simulation at its Hsinchu factory.
Recognizing Taiwan’s strategic position as a global technology hub, the three parties came together to design a supercomputer equipped with 10,000 NVIDIA Blackwell GPUs, significantly expanding the country’s AI computing capabilities. The Taiwan National Science and Technology Council played a pivotal role in supporting this initiative, aiming to accelerate AI development and adoption across industries, including semiconductor research facilitated by Taiwan Semiconductor Manufacturing Corporation (TSMC).
Construction of the AI supercomputer began with the goal of achieving over 90 exaflops of AI performance upon full deployment, expected by 2026. This infrastructure is designed not only to boost Taiwan’s innovation capacity in AI and robotics but also to strengthen its position as a global leader in AI-driven industries, fostering breakthroughs in areas such as cancer research, large language model (LLM) development, and smart city applications.
Moreover, the partnership includes collaborative efforts to apply generative AI within smart city platforms, notably in Kaohsiung, where Foxconn and NVIDIA are working alongside ecosystem partners to implement AI-driven public transportation management and other advanced computing initiatives. This collaboration also aligns with broader industry trends toward integrating AI and high-performance computing (HPC) within manufacturing and supply chain processes, leveraging technologies like NVIDIA Isaac GR00T robotics and digital twin simulations to enhance operational resilience and efficiency worldwide.
Development of the AI Supercomputer
The development of the AI supercomputer, known as the Hon Hai Kaohsiung Super Computing Center, is a collaborative effort between Nvidia, Foxconn (Hon Hai Precision Industry Co.), and Taiwan’s National Science and Technology Council. This initiative aims to establish one of the world’s most powerful AI-optimized supercomputers in Kaohsiung, Taiwan, to advance AI capabilities across various sectors including healthcare, manufacturing, robotics, and smart city infrastructure.
Construction of the supercomputer has already begun, with the first phase scheduled to be operational by mid-2025 and full deployment expected by 2026. The system will be built around Nvidia’s Blackwell GPU architecture and leverage the GB200 NVL72 platform, which consists of 64 racks housing a total of 4,608 Tensor Core GPUs. Each rack features 36 Nvidia Grace CPUs and 72 Nvidia Blackwell GPUs interconnected via Nvidia NVLink technology, delivering an unprecedented 130 terabytes per second (TB/s) of bandwidth. This high-bandwidth interconnect allows the 72-GPU system to function as a unified entity, ideal for training large-scale AI models and executing complex inference tasks on trillion-parameter models in real time.
Once fully operational, the supercomputer is expected to deliver over 90 exaflops of AI performance, positioning it among the fastest AI supercomputers globally and making it the most powerful system in Taiwan. This computational power is anticipated to rival other leading systems such as Germany’s Jupiter supercomputer. The project’s advanced architecture will support Nvidia’s DGX Cloud Infrastructure and Spectrum-X networking frameworks, further enabling scalable and efficient AI model training.
Foxconn plans to leverage the supercomputer to drive innovation in cancer research, large language model (LLM) development, and smart city initiatives, particularly supporting its “three-platform strategy” focused on smart manufacturing, smart cities, and electric vehicles (EVs). The supercomputer will also integrate with Nvidia’s Omniverse and Isaac robotics platforms, enhancing digital twin technologies and robotic automation that are critical to transforming manufacturing processes and urban infrastructure in Taiwan.
This collaboration highlights Taiwan’s growing role in the global AI and high-performance computing (HPC) landscape. The supercomputer is expected to bolster AI adoption across private and public institutions in Taiwan, facilitating breakthroughs in diverse industries such as healthcare, autonomous driving, and smart urban development. Furthermore, Nvidia’s partnership with Foxconn aligns with a broader strategy to strengthen supply chains and manufacturing capabilities for AI infrastructure, including packaging, assembly, and testing operations conducted in partnership with companies like Amkor and SPIL in the United States.
Applications and Impact
The AI supercomputer developed through the collaboration of Nvidia, Foxconn, and Taiwan’s government is poised to significantly influence multiple industries by providing unprecedented computational power exceeding 90 exaFLOPS, enabled by Nvidia’s Blackwell architecture and GB200 NVL72 platform. The system is designed to support a wide array of AI-driven applications, including smart manufacturing, robotics, healthcare, autonomous driving, and smart city initiatives.
One of the major applications of the supercomputer is the advancement of digital twin technology. Utilizing Nvidia Omniverse and Isaac robotics platforms, the project aims to revolutionize manufacturing processes by enabling the creation and simulation of highly detailed 3D digital twins for automated production lines, as demonstrated in Foxconn’s Taiwan Hsinchu factory. This integration of AI and digital twins is expected to enhance operational efficiency, streamline supply chain management, and improve corporate resilience across global facilities.
In healthcare, the supercomputer will support cancer research and other biomedical applications, leveraging AI to accelerate breakthroughs and innovation. Furthermore, large language model development and smart city solutions are key focus areas, reflecting the system’s role in driving AI research that could transform urban infrastructure and services, especially in Kaohsiung and other urban centers in Taiwan.
Beyond specific industry impacts, the partnership strategically positions Taiwan as a leading global hub for AI research, innovation, and high-performance computing (HPC). By deploying one of the world’s most powerful AI supercomputers, Foxconn and its partners aim to bolster Taiwan’s influence in large-scale AI development, attracting interest from semiconductor giants like TSMC and reinforcing the country’s role in the evolving AI and robotics landscape.
The supercomputer also underpins Foxconn’s broader growth strategy, which includes next-generation communications technology, AI, semiconductors, electric vehicles, digital health, and robotics. This aligns with Foxconn’s efforts to integrate AI more deeply into its business operations and to foster innovation across multiple sectors, including smart vehicles and autonomous systems.
Challenges and Considerations
The collaboration between Nvidia, Foxconn, and Taiwan’s government to develop a cutting-edge AI supercomputer presents several challenges and considerations spanning technological, ethical, and geopolitical domains.
Technological Complexity and Supply Chain Demands
The creation of advanced AI chips and supercomputers requires a highly complex supply chain involving sophisticated manufacturing, packaging, assembly, and testing processes. Nvidia has partnered with firms such as Amkor and SPIL in Arizona, and plans significant production ramp-ups through collaborations with TSMC, Foxconn, and Wistron in the United States. The company aims to produce up to half a trillion dollars worth of AI infrastructure within the next four years, leveraging its AI, robotics, and digital twin technologies to optimize factory operations and automate manufacturing processes. This includes employing Nvidia Omniverse for digital twin creation and Nvidia Isaac GR00T for robotics automation. Additionally, Nvidia is building manufacturing plants in Texas alongside Foxconn in Houston and Wistron in Dallas, with mass production expected to scale within 12 to 15 months.
The AI supercomputer project itself, centered on the Hon Hai Kaohsiung Supercomputing Center, aims to deliver over 90 exaFLOPS of AI computing power utilizing Nvidia’s Blackwell architecture. The system will integrate over 2,000 GB200 NVL graphics cards and is designed to support bandwidths up to 130 TBps, incorporating Nvidia’s DGX Cloud Infrastructure and Spectrum-X networking for scalable AI model training. These technological ambitions demand continuous innovation and coordination across the supply chain to meet performance and efficiency goals.
Ethical and Societal Concerns
While the technological advancements promise transformative impacts across industries such as smart manufacturing, healthcare, robotics, and autonomous driving, ethical considerations remain critical. The concept of “sovereign AI,” which involves nations developing AI capabilities using their own infrastructure and data, carries risks related to information control and potential misuse. Nvidia’s sales of AI systems to multiple governments, including some with controversial human rights records, have raised concerns about legitimizing authoritarian surveillance and information control practices. As one critic noted, framing these technologies without clear reflection can inadvertently support authoritarian agendas, leading to risks of isolationism and diminished societal benefits.
Geopolitical and Strategic Implications
The project highlights Taiwan’s growing significance in the global AI and high-performance computing (HPC) landscape, reinforcing its position as a hub for advanced computing. Foxconn’s diversification into semiconductor packaging and electric vehicles, combined with its partnership with Nvidia, exemplifies Taiwan’s ambitions to integrate AI deeply into various sectors. Furthermore, Taiwan Semiconductor Manufacturing Corporation (TSMC), Nvidia’s primary chip supplier and the world’s largest chip manufacturer, is involved in leveraging the new supercomputer for research and development, underscoring the strategic importance of the collaboration in the ongoing industrial revolution driven by AI.
However, this development occurs within a complex geopolitical environment, where advanced AI capabilities can become points of national competitiveness and strategic vulnerability. The movement toward sovereign AI also reflects broader tensions around technological autonomy versus global collaboration, which may impact innovation and international relations.
Timeline and Project Status
Construction of the Hon Hai Kaohsiung Super Computing Center, a collaborative project between Foxconn and NVIDIA, has already begun in Kaohsiung, Taiwan. The project is planned to be executed in multiple phases, with the first phase expected to be operational by mid-2025. Full deployment of the supercomputer is targeted for completion sometime in 2026.
The supercomputer will be built around NVIDIA’s advanced Blackwell GPU architecture, featuring the GB200 NVL72 platform that includes 64 racks and a total of 4,608 Tensor Core GPUs. This setup positions the machine to become one of the most powerful AI supercomputers globally and the most powerful in Taiwan.
Integration with NVIDIA’s Omniverse and Isaac robotics platforms will enable advanced AI capabilities and digital twin technologies aimed at transforming manufacturing processes. Foxconn Vice President James Wu highlighted that the new AI supercomputer represents a significant leap forward in AI computing and efficiency.
The project also aligns with broader collaborations between Foxconn and NVIDIA, including the development of EV ADAS platforms and smart travel solutions, as well as support for smart city initiatives in Kaohsiung. This strategic timeline and phased deployment reflect an ongoing commitment to advancing Taiwan’s AI infrastructure and technological ecosystem through this landmark supercomputing center.
Future Prospects and Developments
The collaboration between NVIDIA, Foxconn, and Taiwan’s government marks a significant advancement in AI technology and infrastructure, positioning Taiwan as a global hub for AI research and innovation. The new supercomputer, currently under construction in Kaohsiung, Taiwan, is expected to be operational in phases, with initial deployment by mid-2025 and full completion targeted for 2026.
Powered by NVIDIA’s Blackwell architecture and the GB200 NVL72 platform, the system will consist of 64 racks and 4,608 Tensor Core GPUs, delivering AI performance exceeding 90 exaflops. This performance level is comparable to leading global supercomputers, such as Germany’s Jupiter, and will make it the fastest AI system in Taiwan. The supercomputer’s capabilities are anticipated to drive breakthroughs in critical fields including cancer research, large language model development, and smart city technologies
The content is provided by Blake Sterling, Clear Reporters
