KUALA LUMPUR: YTL Power International Bhd and Nvidia have brought an artificial intelligence data centre online in Kulai, Johor, a development the partners describe as a major step forward for Malaysia's AI ecosystem. The facility was announced during the 2025 APEC Leaders' Economic Summit in Gyeongju, South Korea. Prime Minister Anwar Ibrahim, Nvidia founder and chief executive Jensen Huang and Investment, Trade and Industry Minister Tengku Zafrul Abdul Aziz attended the announcement.
The campus occupies part of the 600MW YTL Green Data Centre Park in Kulai. Its compute is built on Nvidia's liquid-cooled NVL72 Grace Blackwell, or GB 200, GPUs, paired with ILMU, the large language model that YTL has positioned as Malaysia's own. A GB 200 NVL72 system links 72 Blackwell GPUs through one high-bandwidth fabric and is cooled by liquid rather than air, a design that fits far more compute into the same floor space than earlier air-cooled generations.
The launch came 22 months after Anwar and Huang first met in Kuala Lumpur in December 2023. In the same period the government established a National AI Office under the Digital Ministry, launched an AI Cloud initiative and rolled out AI skills training and adoption programmes in both sectors. ILMU was released in August at the ASEAN AI Summit and is described as Malaysia's first homegrown large language model.
Compute, model and policy in one package
Sovereign AI means a country owns the compute, data and models behind its own public services instead of renting it from foreign providers. Pairing domestic compute with a domestic model is what gives the Kulai campus that character. YTL Power managing director Datuk Seri Yeoh Seok Hong said it has become increasingly critical for nations to build their own clouds to host government and citizen-facing AI services.
- 600MW YTL Green Data Centre Park in Kulai
- Nvidia liquid-cooled NVL72 Grace Blackwell (GB 200) GPUs
- ILMU, Malaysia's first homegrown large language model
- Announced at the 2025 APEC summit in Gyeongju
The National AI Office under the Digital Ministry is the body meant to coordinate that work, while the AI Cloud initiative gives public agencies a route to compute and the training programmes aim to widen the pool of people able to use it. Malaysia has set itself the goal of becoming a leading AI nation by 2030, and the money committed in the 2026 budget is attached to that ambition.
From a first meeting to a live campus
A gap of 22 months between a first meeting and a live AI campus is short by the standards of digital infrastructure, where land, power agreements and construction usually stretch the timeline further. The Kulai site already houses YTL's green data centre park, giving the project an existing land and power position. The presence of the prime minister and the Nvidia chief executive at the announcement signalled the political weight attached to it.
- Anwar and Huang first met in Kuala Lumpur in December 2023
- 22 months between that meeting and the launch
- ILMU released in August at the ASEAN AI Summit
- National AI Office set up under the Digital Ministry
The Kulai facility at the 600MW YTL Green Data Centre Park is equipped with Nvidia's liquid-cooled NVL72 Grace Blackwell (GB 200) GPUs, together with ILMU, which represent a major step forward in building Malaysia's own AI.
ILMU gives the campus a domestic model to run, and the National AI Office gives the government a single point of coordination for AI policy. The AI Cloud initiative and the training programmes are the other two pieces: one supplies access to compute, the other supplies people able to use it. Read together, they describe an attempt to build a full stack rather than to buy capacity case by case.
What the launch means for the region
For Malaysia the significance is less about one building than about the combination it demonstrates: Nvidia hardware, a domestic model and government policy arriving at the same time. Johor has drawn the bulk of the country's data centre investment, and an AI-focused campus there strengthens the case that the state can host advanced workloads, not only standard colocation. It also gives the government a reference site for services it wants to run on its own cloud.
Tengku Zafrul said the Government has earmarked RM5.9bil in its 2026 budget to bolster the nation's artificial intelligence (AI) sector.
What to watch next
The test for Kulai is how quickly the capacity is taken up and whether the domestic model keeps pace with the hardware underneath it. Compute of this class loses value over time and needs a steady flow of workloads, whether from government services, local enterprises or research. The goal of becoming a leading AI nation by 2030 depends on the demand side developing as quickly as the supply side has.
For cloud and data centre operators in Malaysia, the project sets a marker for what an AI-focused campus looks like: liquid cooling, rack-scale GPU systems and a domestic model layer rather than general-purpose colocation. Operators weighing similar investments in Johor will read the Kulai launch as evidence that the policy, power and hardware pieces can be assembled in the state within a single cycle.
- How quickly the GB 200 capacity is taken up
- Whether ILMU keeps pace with the hardware
- How the National AI Office coordinates policy
- How the 2026 budget allocation is spent
The broader signal for Malaysia's cloud and data centre market is that AI infrastructure is now discussed in the same breath as national capability. That framing tends to bring policy support, skills spending and long-term demand from the public sector, which makes the local market more attractive to operators. It also raises expectations, because a campus built for sovereign AI has to show that domestic workloads actually arrive.
Source: The Star
