Defining moment for Malaysia's AI, data centre ambitions

The Sun

10 hours ago

PETALING JAYA: As US-led semiconductor chip restrictions tighten and regional energy costs rise, Malaysia faces a defining moment in its artificial intelligence (AI) and data centre ambitions.
Industry leaders say these global disruptions could either stall the nation's progress or become the very trigger that accelerates its push for AI self-reliance and digital sovereignty.
National Tech Association of Malaysia research committee chairman Woon Tai Hai said the recent US export curbs on high-end AI chips, such as Nvidia's H100 and A100, are already impacting AI-focused startups, research institutions and data centre operators in Malaysia.
'These chips are crucial for training large AI models and powering generative AI applications. Without them, we're seeing delays in deployment and increased costs for local developers,' he told SunBiz.
While some companies are pivoting to older graphic processing unit (GPU) models or exploring Chinese-made alternatives, such as Huawei's Ascend chips, the transition is not seamless.
Compatibility issues, software support gaps and geopolitical uncertainty make it a complex adjustment.
Meanwhile, electricity and cooling costs have surged, especially with Malaysia's high ambient temperatures pushing the limits of energy efficiency in data centres.
Coupled with US tariffs on Malaysian exports and the weakening ringgit, Woon said, the environment is increasingly hostile for small players. 'This triple hit of chip shortages, energy inflation and trade pressure could force some AI projects to downscale or pause altogether.'
However, Woon believes this challenge presents a rare opportunity for Malaysia to reposition itself.
'We are still an attractive alternative to Singapore for hyperscalers, especially with land and energy constraints over there,' he said, pointing to recent investments such as Google's RM9.4 billion data centre in Selangor.
He added that Malaysia could leverage this disruption to double down on home-grown capabilities, forge new global partnerships beyond the US-China binary and evolve from being just a digital consumer to a true AI contributor.
'This is not just a supply chain issue; it's a wake-up call. If we want to lead in AI, we can't import our way to success,' Woon stressed.
Universiti Malaysia Kelantan Institute for Artificial Intelligence and Big Data director Dr Muhammad Akmal Remli warned that Malaysia's long-standing dependency on foreign hardware, cloud platforms and proprietary models has become a strategic liability.
'The AI ecosystem doesn't just rely on talent and data. It runs on compute power and right now, Malaysia doesn't own its compute destiny,' he said.
Akmal called for a comprehensive localisation strategy, beginning with substantial investment in foundational AI research and development through universities, public research agencies and long-term national programmes.
'We can't just be training people to use ChatGPT. We need to train them to build the next generation of language models,' he explained.
Beyond research and development, Akmal proposed the creation of strategic hardware stockpiles and investment in alternative chip architectures, such as RISC-V, Graphcore and Tenstorrent, to diversify away from US-made GPUs.
'Waiting for supply to return to normal is naive. This is structural, not cyclical,' he warned.
Akmal also urged the government to champion a sovereign compute initiative, a state-supported push to establish Malaysia's own high-performance computing infrastructure.
'This is no longer a luxury. We need our own compute backbone to support AI research, secure data hosting and digital services that cannot be outsourced,' he said.
While some pilot efforts exist, Akmal noted that they remain fragmented and underfunded.
What's missing, he said, is a unified national AI policy that aligns research, compute infrastructure, industry application and talent development under one coordinated strategy.
'Right now, we have isolated efforts by Mimos, Mosti, universities and agencies like MRANTI, but they aren't talking to each other. We need a central AI authority or framework to synchronise this,' he said.
Akmal also cautioned that Malaysia's delay in building domestic capacity will ultimately result in higher costs.
'We are not just competing for tech, we're competing for independence. The AI race is about who owns the tools of the future, and right now we're still borrowing them,' he said.
Akmal believes that Malaysia still has the talent, infrastructure and investor interest to build a competitive, ethical and independent AI ecosystem, but only if it takes bold steps now. 'AI is not just about innovation anymore. It's about sovereignty, resilience and relevance in a fractured world.'