Grab and SoftBank commit $1.85bn to Malaysia AI infrastructure
Grab, SoftBank and Petros, the energy company owned by the Malaysian state of Sarawak, have announced a joint $1.85 billion investment in AI computing infrastructure in Malaysia. The partnership targets at least five gigawatts of data-centre capacity over time, making it one of the largest AI infrastructure commitments yet announced in Southeast Asia.
The involvement of Petros is notable. Data centres at scale require guaranteed, dispatchable power, and energy companies are increasingly taking equity positions in computing projects rather than acting purely as suppliers. The structure mirrors deals seen in the Middle East, where state-linked utilities have co-invested alongside hyperscalers to secure both power contracts and a share of the upside.
Southeast Asia's AI headroom
The investment rationale rests on a significant consumption gap. Per-capita AI usage across Southeast Asia is estimated at roughly one-tenth of US levels, yet some analysts project that regional demand could grow as much as 60-fold by 2030. If that trajectory materialises, the region will require substantially more GPU compute, networking and cooling capacity than currently exists. Grab, as the region's dominant super-app operator with hundreds of millions of consumer touchpoints, has both the data assets and the commercial incentive to accelerate that growth.
SoftBank's participation extends a strategy the Japanese conglomerate has pursued through its Vision Fund vehicles: anchoring large-scale AI infrastructure deals in markets where land, power and policy conditions are more favourable than in the US or Europe. Malaysia has positioned itself aggressively for this capital, offering competitive land costs, improving grid connectivity and a government willing to provide regulatory clarity for hyperscale operators.
Competitive and regulatory context
Southeast Asia's data-centre market is contested. Microsoft, Google and AWS have each announced or expanded local cloud regions in the past 18 months, and regional operators such as ST Telemedia and Princeton Digital Group have been building capacity ahead of demand. A five-gigawatt target puts the Grab-SoftBank-Petros consortium at a scale that would rival national cloud programmes in several neighbouring countries.
On the regulatory side, Malaysia's Personal Data Protection Act is under revision, and the government has signalled ambitions to attract sovereign AI workloads. Investors will watch whether the consortium secures anchor tenants from the public sector or from regional financial-services firms, which face strict data-residency requirements under their own national frameworks.
The Asia Tech Podcast discussion from which these details emerged also touched on Singapore's decision to pause certain autonomous-vehicle junction tests following two near-miss incidents involving Grab and ComfortDelGro vehicles. That development adds a note of regulatory caution to Grab's broader technology ambitions in the region, even as its infrastructure investment signals long-term confidence in Southeast Asian AI demand.
No timeline for the first operational capacity was disclosed, and the parties have not named a construction partner or confirmed which grid connection agreements are in place.