Taiwanese laptop maker Asus' subsidiary builds supercomputer to boost island's capacity

A unit of Taiwanese consumer electronics giant Asus is developing a supercomputer, powered by Nvidia chips, that is expected to boost the island's computing capacity by at least 50 per cent, according to a senior executive.

That infrastructure, a collaboration between Taiwan AI Cloud and the island's National Centre for High-Performance Computing, is expected to launch its initial phase with a capacity of 80 petaflops in December, said Peter Wu, CEO of Asus Cloud and Taiwan AI Cloud, in an interview with the South China Morning Post.

Located in the southern city of Tainan, the supercomputer would eventually have a total capacity of 250 petaflops. One petaflop is equal to 1 trillion calculations per second.

Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.

By comparison, Hewlett Packard Enterprises' El Capitan ranked No 1 in the June 2025 list of the world's most powerful supercomputers, with a capacity of 1,742 petaflops. This supercomputer, which started operations last year, takes up a 7,500 sq ft of floor space at the Lawrence Livermore National Laboratory in California.

Wu said Taiwan's latest supercomputer project would boost the island's total computing capacity by at least 50 per cent. It forms part of the island's efforts to become a "top three computing power in Asia".

In February, the Taiwanese government's technology council said it plans to expand the island's overall computing capacity to around 1,200 petaflops by 2029, from 160 petaflops at present.

The new supercomputer would eventually be powered by 1,700 Nvidia H200 graphics processing units (GPUs), according to Wu. He pointed out that parent firm Asus had "good relations with Nvidia" and started talks one or two years before the project commenced to ensure chip supply.

Peter Wu, CEO at Asus Cloud and Taiwan AI Cloud. Photo: Handout alt=Peter Wu, CEO at Asus Cloud and Taiwan AI Cloud. Photo: Handout>

The supercomputer project of Taiwan AI Cloud, previously known as Taiwan Web Service Corp, shows the intense pace of infrastructure development in the Asia-Pacific in response to increased demand for AI, which calls for more data centre and computing capacity.

Wu said Taiwan AI Cloud had done similar AI infrastructure projects elsewhere in the region. These included a data centre in Singapore and another facility under construction in Vietnam for state-owned telecommunications network operator Viettel.

The Vietnam project started earlier this year after the US government approved the shipment of Nvidia chips. The Viettel infrastructure will run on 200 Nvidia GPUs.

Wu said the momentum was rising for agentic AI, a technology that autonomously performs tasks on behalf of a user or another system.

For Taiwan's new supercomputer, the island's Ministry of Digital Affairs "encouraged us to provide an open architecture with an agentic AI framework" to help local companies use or upgrade their existing applications, Wu said.

On prospects in mainland China, Wu said it would "not be easy" for the company to do similar projects there "because of the GPU supply" situation.

China's approach of "stacking and clustering" lower-performance chips to achieve similar processing as systems using advanced AI chips could be feasible in terms of inference, the process that a trained AI model uses to generate conclusions from new data.

Wu indicated that DeepSeek's launch of its R1 reasoning model in January saw a rise in demand for inference because this model excelled in tasks that required logical inference.

"If the workload is similar to [inference], it will be easier to adopt an alternative technology scheme with existing [chips]," Wu said. He added, however, that developers "may be concerned about GPU selection" if the project involved training or fine-tuning AI systems.

He expected three AI segments to further develop in the future. These are computational genomics, quantum computing and so-called digital twin, involving the creation of a virtual replica of an intended or actual real world.

"The killer app [for digital twin] may be in elderly care, which can help them get drugs, food, or take a shower," he said.

This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice reporting on China and Asia for more than a century. For more SCMP stories, please explore the SCMP app or visit the SCMP's Facebook and Twitter pages. Copyright © 2025 South China Morning Post Publishers Ltd. All rights reserved.

Copyright (c) 2025. South China Morning Post Publishers Ltd. All rights reserved.

Scroll to Top