Taiwan Boosts Computing Power with New Nvidia-Powered Supercomputer
- tech360.tv
- Aug 26
- 2 min read
A subsidiary of Taiwanese consumer electronics giant Asus is developing a supercomputer, powered by Nvidia chips, expected to increase the island's computing capacity by at least 50%. The project is a collaboration between Taiwan AI Cloud and the National Centre for High-Performance Computing.
Chief Executive Officer Peter Wu of Asus Cloud and Taiwan AI Cloud stated the initial phase, with a capacity of 80 petaflops, is expected to launch in Dec. One petaflop equals 1 trillion calculations per second. This infrastructure is located in the southern city of Tainan.
The supercomputer will eventually have a total capacity of 250 petaflops. Wu noted this project contributes to Taiwan's goal of becoming a "top three computing power in Asia."
In Feb., the Taiwanese government's technology council announced plans to expand the island's overall computing capacity to approximately 1,200 petaflops by 2029, up from its current 160 petaflops. The new supercomputer will eventually be powered by 1,700 Nvidia H200 graphics processing units (GPUs).
Wu indicated that Asus, the parent organisation, has established good relations with Nvidia to secure chip supply for the project. This supercomputer effort highlights the rapid infrastructure development in the Asia-Pacific region, driven by rising demand for AI and increased data centre needs.
By comparison, Hewlett Packard Enterprises’ El Capitan, ranked as the world’s most powerful supercomputer in June 2025, has a capacity of 1,742 petaflops. El Capitan, which began operations last year, occupies 7,500 sq ft of floor space at the Lawrence Livermore National Laboratory in California.
Taiwan AI Cloud has undertaken similar AI infrastructure projects regionally, including a data centre in Singapore and another facility under construction in Vietnam. The Vietnamese project, for state-owned telecommunications network operator Viettel, started earlier this year.
The US government approved the shipment of Nvidia chips for the Vietnam project, which will run on 200 Nvidia GPUs. Wu observed a growing momentum for agentic AI, a technology designed to autonomously perform tasks for users or other systems.
Taiwan’s Ministry of Digital Affairs encouraged providing an open architecture with an agentic AI framework for the new supercomputer. This aims to help local companies utilise or upgrade their existing applications.
Regarding prospects in mainland China, Wu believes similar projects would be challenging due to the GPU supply situation. He noted that "stacking and clustering" lower-performance chips for inference could be feasible there.
In Jan., DeepSeek’s launch of its R1 reasoning model increased demand for inference, which involves an AI model generating conclusions from new data. Wu added that developers may be concerned about GPU selection if a project involves training or fine-tuning AI systems.
Wu anticipates three AI segments will see further development: computational genomics, quantum computing, and digital twin technology. He suggested that a "killer app" for digital twin could be in elderly care, assisting with tasks such as obtaining drugs, food, or taking a shower.
Asus subsidiary, Taiwan AI Cloud, is building a supercomputer in Tainan to boost Taiwan's computing capacity.
The supercomputer will be powered by Nvidia chips, with an initial 80 petaflops capacity launching in Dec., eventually reaching 250 petaflops.
Taiwan aims to increase its overall computing capacity to 1,200 petaflops by 2029 and become a "top three computing power in Asia."
Source: SCMP