Published: 6 November 2026

Google will invest $15bn in Malaysia across 2027 and 2028, building new data centre campuses as the country tightens the rules on who gets power and water to run them. The package, confirmed in Kuala Lumpur on 4 November with prime minister Anwar Ibrahim present, covers data centre campuses, more than 1GW of new clean energy capacity for the national grid, water security projects and training for 800,000 Malaysians by 2031, per Google .

Key points

• Google noted construction would add an average of $1.1bn a year to Malaysia’s GDP and support more than 37,000 jobs a year, with the campuses running at a power usage effectiveness of 1.12 to 1.14.

• - On 28 October, deputy prime minister Fadillah Yusof, who is also minister of energy transition and water transformation, noted projects in areas with low water reserve margins would not be considered, and that data centres must hold two days of water reserves and source at least 30% of their energy from renewables.

• - Johor’s state coordination committee reportedly rejected around 30% of data centre applications in 2024 for failing sustainability requirements.

• - Johor has the largest incoming data centre pipeline of the Asia-Pacific markets in Knight Frank’s Data Centre Atlas 2026, at 8,542MW, but a colocation vacancy rate of just 0.7%.

• - Singapore’s rate is 4.9%, Jakarta’s 20.5% and Bangkok’s 23.3%.

• Background and analysis. Southeast Asia has emerged as one of the most contested arenas for AI infrastructure. Singapore in particular has become the region's de facto hub for cloud regions, chip design houses and data centre investment, while newer markets are jockeying to host the next wave of capacity. Developments in the region are increasingly watched globally as a bellwether for how AI buildout proceeds in emerging economies.

• Background and analysis. On the facilities side, physical constraints are biting. Power equipment lead times, grid interconnection queues and skilled labour shortages now shape project timelines as much as capital does. Operators that lock in equipment and sites earliest are the ones most likely to deliver on schedule.

• Our take. For anyone building or buying AI infrastructure, the practical lesson is to plan further ahead than instinct suggests. Whether the bottleneck is silicon, transformers or optical links, the winners are securing their constraints years before they need them.

• Source: Capacity, "Malaysia’s data centre ceiling: can it absorb Google’s $15bn?", 5 November 2026. https://capacityglobal.com/news/malaysias-data-centre-ceiling/