KISTI Assists BRIN in Building High-Performance Supercomputer in Indonesia
Cutting-Edge Supercomputing to Support Research Across 11 ASEAN Countries
A New Model of Official Development Assistance (ODA) Using the ASEAN Cooperation Fund
Korea

Seunghae Kim, Team Leader at KISTI (left), and Rifki Sadikin, Director of BRIN Computing Research Center, shake hands in front of a supercomputer installed with the help of Korea. Photo by the ASEAN-Korea Centre

Seunghae Kim, Team Leader at KISTI (left), and Rifki Sadikin, Director of BRIN Computing Research Center, shake hands in front of a supercomputer installed with the help of Korea. Photo by the ASEAN-Korea Centre

View original image

As soon as the server-room door opened, a deafening roar hit our ears. It was the sound of cooling fans spinning nonstop to dissipate the heat generated by NVIDIA’s high-performance H200 graphics processing units (GPUs) and neural processing units (NPUs) made by South Korea’s FuriosaAI as they performed computations simultaneously. The noise was so loud that it was difficult to talk to someone standing right next to you. Yet for Indonesia and the Association of Southeast Asian Nations (ASEAN), it was a “blessing,” promising to help predict typhoons and floods, develop new AI, and address the region’s social and economic challenges.


The High-Performance Computing (HPC) Center at Indonesia’s National Research and Innovation Agency (BRIN), which we visited on Aug. 7, is located at the BRIN Soekarno Science and Technology Park in Cibinong, Bogor Regency, West Java, south of Jakarta. The traffic-clogged roads, motorcycles, and surrounding landscape of ongoing development along the way were typical of a developing ASEAN country. Inside the data center, however, the scene was entirely different. A state-of-the-art supercomputer was ready to support scientific research across all 11 ASEAN member states.


The supercomputer is a joint research infrastructure established by the South Korean government to strengthen ASEAN’s AI and digital research capabilities. The Korea Institute of Science and Technology Information (KISTI), a government-funded research institute under the Ministry of Science and ICT, is carrying out the project with BRIN. The project is funded by $10 million from the ASEAN-Korea Cooperation Fund, established by South Korea’s Ministry of Foreign Affairs, for the period from 2024 to 2028.


In a sense, this is an updated version of official development assistance (ODA) projects that once focused on building bridges and roads. It is a foreign aid program suited to the AI era. A South Korean flag, clearly identifying the ASEAN-Korea Cooperation Fund as the project’s supporter, stood atop the supercomputer alongside the Indonesian flag and the ASEAN logo, a visible sign of this shift.


It is no exaggeration to say that supercomputers are a measure of national power. They are essential for a wide range of research and industrial development. But computing disparities are severe within ASEAN, whose members vary widely in their levels of economic and technological development. Many ASEAN member states, with the exceptions of Singapore and Thailand, lack the HPC systems needed to conduct large-scale AI research. They also lack not only equipment but skilled personnel and training programs for using GPUs and NPUs. Developing generative AI was out of the question, and even when it came to forecasting climate and natural disasters—an immediate practical need—the necessary “brains” were missing. The South Korean government stepped in to address this problem.

The supercomputer installed at BRIN is equipped with 'Renegade,' a domestically produced NPU developed by Furiosa AI, alongside NVIDIA's GPU. Photo by KISTI

The supercomputer installed at BRIN is equipped with 'Renegade,' a domestically produced NPU developed by Furiosa AI, alongside NVIDIA's GPU. Photo by KISTI

View original image

The system built at BRIN is equipped with 64 NVIDIA H200 GPUs across 16 nodes. It has a total of 32 central processing units (CPUs). The supercomputer’s overall computing performance is rated at 4.28 petaflops (PF). One PF represents the ability to perform 1 quadrillion floating-point operations per second.


The system’s crowning feature is a homegrown semiconductor. FuriosaAI’s second-generation NPU, “Renegade (RNGD),” is installed in seven nodes, with four cards per node. The sight of Renegade, developed by a South Korean startup, operating in the same data center as NVIDIA’s H200, a chip from the company that dominates the global AI semiconductor market, stood out.


While the H200 supports a broad range of AI training and inference tasks, Renegade is designed to improve power efficiency in generative AI inference. If NVIDIA’s GPU, which handles powerful general-purpose computing, is the “big brother,” FuriosaAI’s NPU takes on the specialized role of improving inference efficiency. The “Red Team” of FuriosaAI will also be making a strong contribution with its small but powerful chip, taking on NVIDIA, known as the “Green Team.”


BRIN is also an important overseas testing ground for FuriosaAI. AI semiconductors cannot easily enter the market on design performance alone. To win over global customers, they must be installed in real data centers and used to run AI models, demonstrating software compatibility, power efficiency, and long-term stability. The South Korean development cooperation project has therefore become a showcase: it strengthens ASEAN’s research infrastructure while establishing an overseas operational track record for a homegrown AI semiconductor and laying the groundwork for its expansion abroad.


Timing also worked in the project’s favor. Kim Seunghae, head of KISTI’s ASEAN-Korea Digital Innovation Project Team, explained, “We purchased the equipment before memory prices began to surge in earnest due to AI demand, which allowed us to increase memory capacity beyond the original plan.” He added, “By upgrading the specifications of the entire system, we also earned a place in the TOP500, the global ranking of supercomputer performance.” Procuring the equipment just before memory prices soared enabled the team to significantly improve the system’s specifications within the same budget. The supercomputer is ranked 435th on the TOP500 list. Lim Gyeongbin, CEO of eCore, which handled the supercomputer installation, said the team also faced “many difficulties during customs clearance,” hinting at the many obstacles they had to overcome to bring the project to fruition.


KISTI’s support went beyond hardware, including GPUs and NPUs. It built the data-center infrastructure needed to operate the servers, including temperature and humidity controls, power, and rack space, and applied South Korean technologies to security measures such as access control and data protection. The project budget also covered electricity costs during the initial operating period. This was intended to avoid a limitation of previous aid projects, in which utilization rates fell because recipients could not afford the cost of electricity and cooling despite receiving expensive equipment. The project will also train local personnel and transfer NTIS, South Korea’s research project management system. In this way, South Korea’s experience in scientific development is being transferred to ASEAN through a supercomputer.


The supercomputer is not exclusively for Indonesia. Although Indonesia has made a significant contribution, the system will serve all ASEAN member states. A management committee involving ASEAN members will assess requests from research institutions in each country and evaluate research projects before allocating computing resources.


“By using the shared infrastructure, researchers in each member state will be able to help narrow the computing gap between countries,” said Rifki Sadikin, director of BRIN’s Computing Research Center.


Interest is already high. Lee Jongseong, a KISTI executive who has worked with ASEAN countries, said, “We are receiving a steady stream of inquiries from ASEAN countries asking how they can use the supercomputer and how researchers in their countries can participate.” He expects ASEAN-level AI initiatives to take concrete shape soon and believes the facility could become key infrastructure for related research. Shin Hyunseong, science and technology attaché at the South Korean Embassy in Indonesia, also stressed, “It has not yet begun full-scale operations, but inquiries are pouring in.”


One of the first applications drawing ASEAN’s attention is natural disaster management. In ASEAN, where major disasters such as typhoons, torrential rain, floods, and wildfires occur frequently, the ability to quickly process satellite, weather, and terrain data determines how fast and accurately forecasts can be made.


The need became clear during our reporting trip as well. Our plans to travel from Indonesia to the Philippines after completing our reporting in Indonesia were disrupted by local downpours and a typhoon. A supercomputer can run weather models more quickly to calculate a typhoon’s path, rainfall, and the risk of flooding in different areas. It will also make it possible to research how to predict in advance which areas are likely to be hardest hit, rather than responding only after a disaster strikes. If weather-related disasters can be forecast and warnings issued even a few minutes earlier, lives and significant economic losses can be spared.


The system will also be used to develop AI models and study social and economic issues. Large datasets can be analyzed to examine complex challenges facing ASEAN, including healthcare, agriculture, the environment, transportation, urban planning, population, poverty, and industrial structure. Supercomputers can also simulate a range of scenarios and compare their effects before economic and social policies are implemented.

Seunghae Kim, Team Leader at KISTI, is explaining the operating status of the supercomputer installed at Indonesia's BRIN with support from Korea. Photo by Paek Jongmin, Tech Specialist

Seunghae Kim, Team Leader at KISTI, is explaining the operating status of the supercomputer installed at Indonesia's BRIN with support from Korea. Photo by Paek Jongmin, Tech Specialist

View original image

South Korea’s system for managing research data will also be transferred, in addition to computing capacity. KISTI will establish its National Science and Technology Information Service (NTIS) in BRIN’s HPC environment and train local researchers to use it. NTIS systematically collects and enables searches of information on national research and development projects, researchers, and research outcomes.


Once NTIS is established, research projects and results across ASEAN can be managed systematically. It can help identify what research is being conducted in each country, reduce duplicated investment, and create a foundation for joint research among member states. South Korea is thus providing not only a supercomputer but also its experience in research administration.


South Korea is also helping train the people who will operate the facility. KISTI will train 40 ASEAN researchers each year for four years, for a total of 160. The curriculum includes AI training using actual GPUs, HPC operations, and NTIS use. The project’s core aim is not simply to install equipment and leave, but to build local researchers’ capacity to operate and use it independently.


ASEAN Secretary-General Kao Kim Hourn also expressed high expectations. Citing the assessment of the ASEAN Ministerial Meeting on Science, Technology and Innovation, he said, “We attach great importance to this kind of cooperation,” adding that it is “highly timely and meaningful in the current era.”


Secretary-General Kao stressed, “I hope the project we are pursuing together will benefit ASEAN as a whole, have a significant long-term impact, and contribute to the development of relations between ASEAN and South Korea.” He also identified disaster management and cybersecurity as areas in which ASEAN and South Korea could expand their cooperation.


The BRIN project also demonstrates how South Korea’s official development assistance (ODA) is changing. While past development cooperation focused on building roads and buildings or providing equipment, this project brings together HPC, homegrown AI semiconductors, security, a research information system, operating costs, and personnel training in a single package.



Projects like this cannot end with a one-time contribution. After a certain period of use, a supercomputer needs to be upgraded to improve its performance. KISTI has also gained experience replacing and upgrading national supercomputers by generation at its headquarters in Daejeon. The latest version is the sixth-generation system, “Hangang.” Lee Sik, president of KISTI, stressed, “We need to contribute actively to the development of research in ASEAN through sustained support, rather than one-off assistance.”


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing