How AI is transforming Southeast Asia data centres

How AI is transforming Southeast Asia data centres 

Artificial intelligence (AI) is rewriting the rules of how we build data centres in Southeast Asia. As server racks consume more power than ever before, success now depends not just on speed to market, but also on early-stage decisions regarding power, cooling and procurement. 

Market forecasts predict data centre investment in Southeast Asia could reach $30bn by 2030, but the boom faces a key major constraint: delivering power-dense infrastructure quickly. Rack densities have jumped from a standard 8–12 kW to 25–40 kW and beyond, changing what defines success. 

While available land and construction considerations remain important for developers or owners, utilities, thermal strategy and manufacturing capacity now shape schedules, cost and feasibility long before construction begins.  

A new phase for Southeast Asia’s data centres 

The data centre challenge is no longer only about expansion, but about escalation. The question isn’t ‘where do we build?’ but ‘how do we deliver the power, cooling and resilience quickly that AI workloads demand?’.  

Across Singapore, Indonesia, Malaysia and Thailand, developers are planning for higher rack densities, advanced cooling and longer procurement cycles, however each market absorbs these pressures differently.

Singapore is densifying within tight constraints, while Indonesia and Malaysia are scaling into power-rich locations with longer infrastructure timelines. Meanwhile Thailand is emerging with its own supply chain limitations. 

The real opportunity lies in understanding how these locations play different roles within a wider regional delivery strategy rather than choosing a single market. 

Credibility sets the pace 

Southeast Asia added over 1GW of data centre capacity last year, but across all major markets the shift to liquid cooling and rear-door heat exchangers has had the greatest impact.  

This density shift concentrates risk in three specific areas: 

1. Power readiness  

Utility coordination, substation capacity and grid approvals are critical. In Jakarta for example, high voltage work can take 18-24 months. Approvals in Malaysia often take 10-18 months. Singapore, meanwhile, relies on upgrading existing brownfield sites. 

2. Cooling choices 

Once you define your power, cooling becomes the next constraint. Above 25 kW per rack, liquid or hybrid cooling becomes essential.

These systems account for 15-20% of capital expenditure and require six to seven months lead time. This forces early decisions on water strategy, electrical design and suppliers. 

3. Equipment lead times  

Feasibility relies on manufacturing capacity. For example, generator set up and delivery has a lead time of approximately 40-45 weeks, switchgear takes 25-40 weeks, transformers need 20-32 weeks and liquid cooling units require 28-32 weeks. 

These factory slots must be secured during site acquisition and early design, not after contractor appointment. 

These risks are linked. Power, cooling and long-lead equipment decisions should be made months before construction, and delays in any one area can spread across an entire programme. 

Why traditional delivery models are under stress 

The traditional ‘design-bid-build’ model worked when civil works dominated costs and the lead times were short. In today's market, it simply isn’t fit for purpose. 

Contractor capacity is tight.

Singapore has three or four tier-one contractors, Malaysia has four to five and Indonesia has five to six. With 15-20 major projects running concurrently, relying on sole-source models pile risk onto firms that are already stretched. 

Meanwhile, procurement timing has also flipped. Equipment orders now lag six to eight months behind optimal, exposing schedules to volatile lead times. Design, too, must evolve during delivery. 

AI-driven density, cooling choices and utility interfaces demand flexibility, but traditional models penalise change. This pressure is structural and not just a passing phase. 

What shapes delivery in different markets?  

Understanding the unique challenges and opportunities in each market is key to successful data centre operations in southeast Asia. We’ve provided a breakdown of market-specific delivery considerations for key locations in the region. 

Singapore

Indonesia 

Malaysia 

Thailand

The new cost reality: 2025 benchmarks 

Our Data centre construction cost index 2025 shows cost drivers shifting from civil works and labour towards power infrastructure, cooling systems and imported long-lead equipment.  

Current benchmarks for air-cooled designs are: 

• Singapore: US$14.53/watt  
• Jakarta: US$11.21/watt  
• Kuala Lumpur: US$11.37/watt  
• Thailand: US$10.80-11.50/watt 

Liquid cooling adds between US$1.80-2.40 per watt depending on architecture (rear-door, in-row or direct-to-chip). High-density specifications for AI (40+ kW per rack) typically increase costs by 18-22% over legacy baselines. 

Cost certainty now depends on procurement certainty. Original equipment manufacturing (OEM) slot reservations, freight planning and multi-scenario design reviews must happen before general contractor bidding, not after.  

Escalation risk has shifted from labour and materials to equipment lead-time delays and late-stage design changes. 

From speed to certainty 

With risk shifting upstream and complexity growing, building faster is no longer a competitive advantage. The focus must be on the ability to manage risk earlier in the lifecycle, long before construction begins.  

In this environment, your delivery model is a strategic decision, not a contractual formality. How you structure procurement, allocate risk and control critical technical interfaces increasingly determine whether your projects reaches operation on time, at cost and at density. 

The growing interest in multi-prime contracting reflects this shift. It signals a mature market where early commitment and integrated execution are non-negotiable for delivering high-density capacity with confidence.