SoftBank, Grab and PETROS Commit $1.85B to Malaysia AI Build
3 min readSoftBank Group, Grab and Sarawak’s state energy company PETROS have signed joint development agreements to put US$1.85 billion of seed capital into Malaysia AI infrastructure, with a long-term target of five gigawatts or more of compute capacity. The deal, announced Monday from Tokyo and Singapore, turns SoftBank’s October pledge for Malaysia into funded projects with named partners.
Background
Malaysia has become one of Southeast Asia’s busiest data center markets, led by the Johor corridor across the strait from Singapore. In October, Masayoshi Son said SoftBank intended to invest US$50 billion in the country’s AI and data center sectors, and SoftBank signed a framework agreement with Sarawak in Kuching on October 8. Grab, the Singapore-based ride-hailing and payments group, was founded in Malaysia and has been looking for growth beyond its core apps. Both projects are tied to Malaysia’s AI Nation 2030 national action plan.
What Happened
According to SoftBank’s announcement, the money splits two ways. A US$1 billion Sarawak project takes US$800 million from SoftBank, US$100 million from Grab and US$100 million from PETROS, with SoftBank as majority shareholder and Grab holding an option to raise its stake. A US$850 million project in Peninsular Malaysia takes US$650 million from Grab and US$200 million from SoftBank, with Grab as founding shareholder and lead developer. In total SoftBank commits US$1 billion, Grab US$750 million and PETROS US$100 million, or about RM7.5 billion, as Malay Mail reports.
Grab chief executive Anthony Tan will oversee both projects and appoint dedicated management teams, and SoftBank has agreed in principle to a five-year lock-up on the Grab shares it holds directly. Funding is staged against development and customer milestones, and the partners say they will seek additional investors and hyperscaler customers. The facilities are to follow Malaysia’s renewable energy guidelines and use designs that cut water use and reliance on the local grid, TechNode notes.
Why It Matters
Five gigawatts would rank among the largest AI infrastructure programs in Asia. For comparison, Credit Agricole has estimated that all of Malaysia will reach roughly 4 to 6 gigawatts of data center capacity by 2030. The deal also marks an unusual pivot for Grab, which is putting three quarters of a billion dollars into power and compute rather than apps, with its founder personally running the build. For SoftBank, it adds a Southeast Asian anchor to a portfolio of sovereign-scale AI projects. The open questions are who the anchor tenants will be, since no hyperscaler is named yet, and how quickly Sarawak’s hydropower-heavy grid can absorb the load.
Definitive agreements and the first site announcements come next. The joint ventures still have to be formed, so watch for the first customer contracts.