SoftBank, Grab and Petros put $1.85 billion into Malaysian AI infrastructure, aiming for 5 gigawatts of data centres in Sarawak and the peninsula

How much are SoftBank, Grab and Petros putting into Malaysia's AI build out? 1.85 billion US dollars to start. SoftBank will provide 1 billion dollars of seed capital, Grab 750 million dollars and Petroleum Sarawak, known as Petros, 100 million dollars for two AI infrastructure projects, with a long term goal of at least 5 gigawatts of data centre capacity across Sarawak and Peninsular Malaysia, the companies said in a joint statement reported by VnExpress.

The money turns an October framework agreement, signed in Kuching to explore an AI and digital infrastructure platform in Sarawak, into committed capital. It also gives Malaysia a high profile anchor for its ambition to become Southeast Asia's main AI hub, at a time when both SoftBank and Grab have been knocked back elsewhere in Asia.

What the three partners bring

Each partner fills a gap. SoftBank brings capital and a long history of betting on AI and the hardware behind it. Grab, the Singapore based superapp that started in Malaysia in 2012, brings a regional customer base across eight countries and a strong interest in building AI tools in local languages and for local conditions. Petros, Sarawak's state owned energy company, brings what data centres need most: power, land and a supportive state government.

"AI infrastructure will power the next wave of growth, and Malaysia has the talent and ambition to help lead it," Masayoshi Son said. "Together with Grab, we will build the capacity needed to support Malaysia's digital ambitions and accelerate AI innovation across Southeast Asia."

Grab's chief executive Anthony Tan framed it as a homecoming. "With SoftBank Group, we plan to build critical AI infrastructure in Malaysia, where Grab began, for all of Southeast Asia," he said. "This infrastructure will allow us to accelerate the development of localized AI solutions for our communities."

The split of the money is telling. SoftBank's 1 billion dollars makes it the lead investor, while Grab's 750 million dollars is a large commitment for the ride hailing and payments group, signalling that it sees computing capacity as core to its future rather than a side bet. Petros contributes the smallest cash share, but its real contribution is access to energy and state support, which in a data centre project can be worth more than the capital itself. The joint statement did not say how the two projects divide between Sarawak and the peninsula, or when construction will begin.

Why Sarawak matters

Sarawak is betting that cheap, plentiful energy can make it a destination for power hungry AI campuses. The state has large hydropower resources and gas, and its government has pushed a digital economy agenda, including a sovereign data platform. A partner such as Petros can help secure electricity on the scale that a gigawatt class data centre needs, which is the single biggest constraint on AI build outs worldwide.

The peninsula offers the other half of the equation: proximity to Singapore, fibre links and an existing cluster of operators. Johor in particular has become one of the fastest growing data centre markets in the world since Singapore restricted new data centres, and we have reported on the strain this places on the grid in our analysis of Johor data centres and power demand.

Malaysia's data centre capacity is projected to rise to 6.8 gigawatts by 2030 from 1.3 gigawatts as of June, according to Cushman and Wakefield. A target of 5 gigawatts for this one partnership would therefore be very large relative to the national market, which is why the companies describe it as a long term goal rather than a near term plan.

Gigawatts, not dollars, are the real currency of the AI race. Sarawak's pitch is simple: it has the power, and now it has partners with customers.

The bigger SoftBank picture

The Malaysia deal sits alongside a much larger figure. In October, Prime Minister Anwar Ibrahim said Son planned a 50 billion dollar AI and data centre investment in Malaysia, centred on Kuching and the peninsula, a figure SoftBank has not publicly confirmed. We examined that claim in our report on Anwar's 50 billion dollar mystery investor. The 1.85 billion dollar seed commitment is the first concrete money attached to that vision.

It comes in a difficult period for SoftBank. The group reported a quarterly net loss of about 7 billion dollars this week, citing interest costs, administrative expenses and a 3.7 billion dollar charge in the first half for revising procurement plans for power equipment at US AI data centre projects, according to Free Malaysia Today. Last week China blocked its 5.4 billion dollar acquisition of ABB's robotics business.

Grab, too, has faced resistance. In October, Taiwan rejected its proposed 600 million dollar takeover of Foodpanda's Taiwan business. For both companies, Malaysia offers a friendly government, a ready market and a chance to show investors that their AI strategies are producing physical assets.

What it means for Malaysia and the region

For Malaysia, the deal strengthens its position as Southeast Asia's data centre capital. Several factors make it attractive to AI investors:

• available land and power in Sarawak and Johor
• government support at federal and state level
• proximity to Singapore's finance and connectivity
• a growing pool of engineering talent
• competitive costs compared with Singapore and Japan

There are risks. Data centres draw heavily on water and electricity, and public pushback has grown in parts of the region, from Johor to Indonesia. US scrutiny of AI chip flows through Southeast Asia is also rising, as shown by the investigation into Singapore's Megaspeed in our report on the Jensen Huang affair. Projects that rely on advanced chips will need clean compliance records.

For Grab's users, the benefit may be less visible but real. Locally hosted computing could cut costs and latency for AI features in ride hailing, food delivery, payments and lending, and could support models trained on Malay, Bahasa Indonesia, Thai and Vietnamese data. That is the strategic logic behind Tan's talk of localised AI.

SoftBank, Grab and Petros have turned an exploratory pact into 1.85 billion dollars of seed money, and Malaysia has another flagship for its AI ambitions. The 5 gigawatt target will take years and far more capital to reach, but the deal confirms where global AI money is heading in Southeast Asia: to places that can offer power, land and political goodwill at the same time.