UTM Launches Blackwell GPU Supercomputer at Its Faculty of AI

Universiti Teknologi Malaysia (UTM) has switched on what it calls the fastest GPU supercomputer at any Malaysian public university, making it the first public university in the country with advanced high-performance computing infrastructure built around graphics processors. The system was launched on October 2 at the Vice-Chancellor’s Gathering at UTM Kuala Lumpur, Bernama reported.

Housed at UTM’s Faculty of Artificial Intelligence (FAI), the machine is based on NVIDIA’s Blackwell architecture and can train and run models with up to one trillion parameters over a 400 Gbps InfiniBand network. UTM quotes processing capacity of up to 72 PFLOPS for training and 144 PFLOPS for inference. Alongside it sits an AI Workstation Laboratory with 120 AI desktop computers, so up to 120 students and researchers can build and train models at the same time.

Vice-chancellor Prof Dr Mohd Shafry Mohd Rahim said the infrastructure puts UTM in a better position to provide the research, learning and innovation environment needed to produce AI talent for industry and the country. The facility is part of a RM20 million Ministry of Higher Education investment under Budget 2024 to establish FAI, Malaysia’s first faculty dedicated to artificial intelligence.

What it means in practice

For researchers, the main change is access. University teams working on large language models or computer vision have often depended on scarce cloud credits or overseas partners. UTM says the cluster will support work on healthcare, smart manufacturing and local-language systems, areas where Malaysian data and Malay-language performance matter, as projects like YTL AI Labs and NVIDIA’s Nemotron Personas Malaysia dataset have shown.

For employers, a campus cluster means graduates who have handled industry-grade hardware before their first job. That complements the student-focused ILMUcode launch at Universiti Malaya and the government’s plan for a national AI compute exchange in early 2027, which aims to make GPU capacity easier to share across institutions.

Some caution is warranted. A single faculty cluster is small next to the commercial AI factories going up in Johor and Selangor, and peak PFLOPS figures say little about how many projects can run at once or how long the queues will be. Steady funding for power, maintenance and support staff will decide whether the system stays useful beyond its launch year.

What to watch: whether UTM opens time on the cluster to other universities and industry partners, the first research results trained on it, and whether other public universities secure similar budgets. For now, UTM has set a new baseline for what AI compute on a Malaysian public campus looks like.