A data centre may not look glamorous. To some it is just extracting the nation’s water and electricity to stream cat videos and train chatbots for foreign companies. That looks like a powerful image but to a large extent, not correct. (Even for a blog writer like myself)
Between 2021 and mid-2025, the Malaysian Investment Development Authority (Mida) approved 143 data centre projects worth RM144.4 billion. Amazon, Microsoft, Google and Oracle alone have committed around RM69 billion. Yet, they did not have to choose Malaysia.
Industry frameworks
such as NVIDIA's "five-layer cake" describe: energy, data centres,
cloud, foundation models, and applications. In Malaysia, one group, YTL, has
built assets across every layer, deliberately. "Artificial intelligence
will become one of the defining technologies of this generation, much like
electricity and the internet before it. Building sovereign AI is not about one
model or one application. It is about developing the full stack of
capabilities, from digital infrastructure and cloud to foundation models and
real-world applications, that allows Malaysia to innovate, compete, and chart
its own AI future," says Yeoh Keong Hann, Executive Director of YTL
Power International Berhad.
Source: https://ytldatacenters.com
Energy is the foundation layer. Training and running AI models takes enormous, continuous power, and an interruption can set a training run back by days. A country that relies on external energy for its AI ambitions inherits that dependency before a single chip or model is even involved. YTL Power has spent more than three decades as an independent power producer, powering major industries, and AI is now the most demanding of them.
Data centres are the physical facilities where that energy becomes usable compute. Where they sit determines whose laws apply to the hardware inside, and how quickly capacity can grow. YTL has built that scale at its data centre campus in Johor, which spans roughly 1,600 acres near Singapore and is scaling to 1,200MW of capacity within the same site.
Cloud is the software layer that turns data centre hardware into compute that businesses and researchers can rent, without owning or managing the machines themselves. Most countries and businesses rent this layer from a small number of foreign hyperscalers, since building it takes capital and technical validation few companies achieve. YTL AI Cloud closes that gap for Malaysia. In May, YTL AI Cloud became among the first in Asia to reach NVIDIA Exemplar Cloud status, a designation NVIDIA reserves for providers meeting its highest technical bar for AI workloads. It is one of only fourteen cloud providers worldwide to achieve these standards.
The foundation model is the underlying AI system supplying the intelligence. It interprets the user's request and generates a response. If built and governed by a foreign company, its behaviour and limitations are set by that company's priorities, usually a broad global audience, with local markets adapted in after the fact. ILMU, developed by YTL AI Labs, was built for Malaysia, with a deep understanding of Malaysian language, culture and context. ILMU aims to be world-class with a Malaysian advantage, ranking first in the world on the MalayMMLU benchmark, which tests academic and general knowledge across dozens of subjects in Bahasa Melayu, ahead of global models.
Applications are where value is created for the end user, and where intelligence, scale, and security all have to come together. In Malaysia, that convergence already exists: the capability to build autonomous AI agents, the country's first AI-powered bank, and an AI-first telco putting AI directly into millions of pockets.
Agents are systems built around a model that let it act rather than simply answer: filing a form, processing a document, completing a transaction, without a person driving every step. YTL AI Labs has already moved into this space with ILMU Claw, built on the open-source OpenClaw framework, letting developers and enterprises build and host autonomous AI agents on Malaysian infrastructure.
Ryt Bank, Malaysia's
first AI-powered digital bank, runs on Ryt AI, a multi-agent system built on
ILMU that transfers money and pays bills from a typed request or a snapped
photo of a bill. It has scaled to more than a million users in its first year,
on infrastructure built in the country.
Yes began as Malaysia's first nationwide 4G and, later, 5G operator, and is now an AI-first telco, giving every subscriber complimentary access to ILMUchat Pro and bringing ILMUchat to more than three million users. Yes also has its own AI assistant, Sofea, built in directly into the MyYes app, enabling customers check their usage and pay bills seamlessly.
The conversation is now shifting from adopting AI to building the capability behind it. The countries that succeed in AI will not simply be those that use the technology most. They will be the ones that are able to independently build the infrastructure, talent and technological capabilities to innovate continuously. And the Government has blocked any data centre that is “merely taking advantage of cheaper water facilities and lower energy prices” (PM Anwar Ibrahim).
None of this shows up as mass employment. Critics are right that a finished data centre needs few staff. But a KPMG study in 2025 put the sector’s likely output at around RM139 billion a year by 2030, close to 4% of the economy, and roughly 30,900 jobs once the supply chain is counted.
The opportunity now is for every Malaysian to take hold of this and make full use of it: to work together, participate, and collaborate, across industry, academia, and government, so that Malaysia becomes an AI nation that leads the future.
References:
Malaysia’s data centre backlash is loud, popular and wrong, K. Kathirgugan, FMT, 2 July 2026
Why Malaysia's AI
Future Depends on Building Capability, theedgemalaysia.com,
16 July 2026
No comments:
Post a Comment