SoftBank, Grab and Petros Commit $1.85 Billion to Malaysian AI Data Centres
The partners aim for 5 gigawatts or more of AI computing across Sarawak and Peninsular Malaysia, with SoftBank leading one project and Grab the other.
By Nakoda News Desk
·2 min read
Prefer Nakoda News on Google

SoftBank Group, Grab Holdings and Petroleum Sarawak Bhd (Petros) have signed joint development agreements to build artificial intelligence data centres in Malaysia. They are putting up $1.85 billion in initial seed capital for a multi-year plan that they say could eventually reach 5 gigawatts or more of computing capacity.
Grab announced the agreements on Monday. SoftBank is contributing $1 billion, Grab $750 million and Petros, the Sarawak state energy company, $100 million. The companies said the money will be released in stages as the projects reach development and customer contract milestones, and that it is subject to standard closing conditions.
The seed money is small next to the long-term target. Most of the cost of reaching several gigawatts would have to come from elsewhere, and the partners said they plan to bring in more investors and hyperscale cloud customers as the projects grow. They said that once the facilities are fully running they would be among the largest AI infrastructure of their kind in Asia, giving governments, multinational companies and small businesses access to advanced computing.
The plan has two parts with different leaders. The Sarawak project starts with $1 billion: $800 million from SoftBank and $100 million each from Grab and Petros. SoftBank will lead it and hold a majority stake. Grab has an option to raise its holding, which would reduce SoftBank's share, though SoftBank would remain the majority owner. The companies did not give ownership percentages or the terms of the option.
The Peninsular Malaysia project starts with $850 million: $650 million from Grab and $200 million from SoftBank. Grab will lead its development and intends to bring in long-term strategic investors as it expands. Petros is not listed as an investor in this part.
The deal moves Grab, the Singapore-listed ride-hailing and delivery company that started in Malaysia, into owning AI infrastructure. Anthony Tan, Grab's co-founder, chairman and chief executive, said the company would build the facilities in Malaysia to serve all of Southeast Asia, and that the work would speed up locally built AI services and create skilled jobs and AI training.
Masayoshi Son, SoftBank's chairman and chief executive, said in the statement: "AI infrastructure will power the next wave of growth and Malaysia has the talent and ambition to lead it."
The companies said the plan was developed with Malaysia's federal government and the Sarawak state government. It is aligned with the National AI Action Plan 2026-2030, known as AI Nation 2030, and with Sarawak's Post Covid-19 Development Strategy 2030. They said the sites will follow Malaysia's guidelines on renewable energy and will be designed to use less water and depend less on the local power grid. The partners also said they will look at other areas of common interest, including advanced manufacturing in Malaysia.
The announcements do not say where in Sarawak or on the peninsula the data centres will be built, how much capacity the first phases will have, or when they are due to start operating. Because each tranche of funding depends on signing customers, the timing of the first contracts will show how quickly the plan moves beyond its initial $1.85 billion.
Sources
- DatacenterDynamics · SoftBank, Petros, and Grab plan Malaysian data center buildout with "long-term target" of 5GW
- TechNode Global · SoftBank, Grab commit $1.85B to Malaysian AI infrastructure with PETROS
- Malay Mail · SoftBank, Grab and Petros to build 'among Asia's largest' AI facilities in Malaysia with US$1.85b seed capital
Produced with AI assistance from the sources above. How we work
By
Nakoda News Desk
Nakoda News covers AI and business. Every story links the sources it draws on; how we work is set out in our editorial standards.


