Liquid cooling, renewable power commitments and advisory-first operational intelligence are reshaping the investment case for AI-ready colocation, where resilience, energy cost and asset telemetry increasingly determine long-term value.
CETA System Co., Limited is examining how Malaysia’s expanding data-centre corridor is bringing liquid cooling, renewable power procurement and condition-based asset monitoring into sharper focus for operators, hyperscalers and investors assessing AI infrastructure exposure.
Equinix has allocated $190 million to KL2, its fourth Malaysian data centre, positioned less than one kilometre from its existing Kuala Lumpur facility. At full buildout, KL2 is designed to accommodate more than 2,200 cabinets, with significant capacity dedicated to liquid cooling infrastructure for AI and high-performance computing workloads. The facility is also set to operate on 100% renewable energy from initial commissioning, supporting Equinix’s 2030 target for complete renewable coverage across retail colocation services.
Malaysia’s position as a regional digital infrastructure hub gives the development wider significance. Government programmes, including the National Artificial Intelligence Office, reinforce the country’s role along the Singapore-ASEAN corridor, where data-centre expansion is increasingly judged not only by capacity, but by thermal resilience, energy sourcing and operational discipline.
KL2 occupies a 14,300-square-metre site in Cyberjaya, Malaysia’s designated technology hub within the Multimedia Super Corridor. The site was acquired from Cyberview Sdn Bhd for approximately $4.77 million, and its proximity to Equinix’s existing 900-cabinet, 2,630-square-metre KL1 facility creates a campus configuration that can support network density and shared operational intelligence. Adjacent land acquisition also supports future capacity expansion.
According to Lee Tsz-Hin, who serves as chief executive at CETA System Co., Limited, the rise of AI infrastructure has reshaped how investors should read data-centre capacity. “Cabinet count remains important,” he said, “but thermal stability, asset health and energy control now have a direct bearing on operating risk.”
Liquid cooling is central to that shift. AI application rack densities routinely exceed 100 kW, beyond the practical limits of traditional air-cooling systems. Water’s stronger heat-transfer characteristics allow more efficient thermal absorption than air-based configurations, supporting lower component operating temperatures in high-density environments. Three approaches have gained adoption: rear-door heat exchangers, direct-to-chip systems and immersion cooling, with direct-to-chip implementations described as the dominant deployment model. Per-component costs typically range from about $191 to $400.
The investment case is not limited to technical performance. Liquid-cooling systems can reduce energy consumption by more than 27% relative to air-cooling configurations. A 75% transition from air to liquid cooling can reduce facility power use by 27% and whole-site consumption by 15.5%. Direct-to-chip systems can lower energy costs by 31%, while immersion cooling can reduce them by 37% compared with equivalent air-cooled installations. As CETA System notes, cooling optimisation can account for up to 40% of total data-centre energy consumption, making thermal management a material operating-cost factor rather than a narrow engineering issue.
Integration remains a critical challenge. Data centre infrastructure management systems and building management systems can provide unified visibility across IT and building operations, but legacy compatibility limitations persist. Temperature data from building management systems can support dynamic cooling adjustments when aligned with workload monitoring, while expanded interconnectivity also increases the need for access control and threat protection.
The same considerations apply to Malaysia’s broader investment thesis. The country’s data-centre market reached $3.86 billion in the latest annual tally and is projected to exceed $12.88 billion by 2030. North American hyperscalers committed $22.23 billion across the first ten months of the same reporting window. Johor is expected to account for 60% of Malaysia’s capacity by 2030, supported by proximity to Singapore and electricity tariffs averaging $0.10 per kilowatt-hour. The Johor-Singapore Special Economic Zone further positions the corridor as a scalable hub rather than simply a spillover market.
Operational intelligence is becoming part of that capital-efficiency discussion. AI control systems have demonstrated average energy savings of around 30% across multiple data centres by capturing sensor data every five minutes and predicting how operational adjustments may affect future consumption. In mission-critical environments, however, deployment must remain advisory-first and operator-supervised. Operators retain oversight authority and can disable AI control mode at any time.
“Advisory-first automation is the more credible path for critical infrastructure,” observed Lee. “The value lies in faster visibility and better recommendations, while human approval, auditability and operating constraints remain central to trust.”
Outage economics underline the point. Unplanned downtime costs for AI-ready facilities range from $110,688 to more than $553,441.9 per hour, depending on sector. Large enterprise installations face average hourly costs of $597,717.4. Industry survey data indicates that 70% of data-centre outage incidents cost $110,688 or more, with 25% exceeding $1.1 million. A single 90-minute outage can result in more than $559,649.7 in combined revenue loss and productivity impact, while Fortune Global 500 companies collectively lose an estimated $1.7 trillion annually due to unplanned downtime.
Human error and UPS system failures account for more than 50% of all outages, indicating that many incidents involve preventable asset degradation or operational missteps. That places predictive maintenance, telemetry integration and vendor-agnostic monitoring at the centre of operational resilience. Single-vendor alarm infrastructure can increase long-term costs through lock-in and limited flexibility, while proprietary communication protocols can add complexity and error risk during power-system failures.
Equinix KL2 illustrates the convergence of high-density AI workloads, renewable power commitments and proximity-based network density within Malaysia’s digital infrastructure corridor. For CETA System, the development reinforces a wider market priority: AI-ready facilities will need advisory-first cooling optimisation, condition-based monitoring and vendor-agnostic telemetry if operators are to manage energy cost, outage exposure and compliance pressure while maintaining capital discipline.
About CETA System
CETA System Co., Limited is a Hong Kong-incorporated technology company founded in 2017, delivering artificial-intelligence solutions for data-centre infrastructure. Its vendor-agnostic platform combines HVAC and chiller-plant energy optimisation with predictive maintenance for critical assets, including UPS systems, generators and chillers, and integrates with existing building-management and DCIM environments through an advisory-first deployment model for colocation, enterprise and hyperscale operators across Asia-Pacific and beyond.
- Website: https://cetasystem.com
• Registered business: CETA System Co., Limited (Hong Kong BRN 67731517; CRN 2533166)
