KUALA LUMPUR, March 13 — Ever wondered where your photos, videos, banking details and personal data go when you upload them online?

Despite confusion of terms like cloud storage, they are actually kept in physical massive data centres either owned or rented by the companies managing your digital footprint.

Malaysia has been home to data centres since the early 2000s, with 25 currently operating in Cyberjaya, Selangor.

But the rise of artificial intelligence (AI) is pushing these facilities to their limits, driving up energy consumption and water usage at an unprecedented rate, which Malaysia needs to look into sooner rather than later even as the government is gaining interest and investments into this field.

What’s inside a data centre?

At its simplest, a data centre is a physical facility that organisations use to house their critical applications and data.

Think of a data centre like a giant, super-powerful computer room that has:

• Servers: The main body that processes and stores data.
• Network equipment: Routers and switches that connect you to the internet.
• Storage: Hard drives and other systems to save information.
• Cooling systems: Cooling is needed to keep the servers running 24/7.
• Backup power: Generators and batteries to keep things working in case of power outages.

A data centre’s design is based on a network of computing and storage resources that enable the delivery of shared applications and data.

How do data centres work?

When you use an app or website, the data centre handles your request to process and/or store the item.

Firewalls and cybersecurity tools keep hackers out. Air conditioning and water cooling prevent servers from overheating, while backup power ensures no disruptions.

If one server is too busy, another steps in to keep things fast.

Types of data centres

1. Hyperscale data centres

There was a time when data centre expansion meant adding more servers to a cabinet, more racks inside a cage or renting an additional office space to meet demands.

In the age of hyperscale computing, the biggest players are adding entire buildings.

For example, Meta owns 24 hyperscale data centre campuses, of which 20 are in the US and covers 48 million square feet with close to US$25 billion (RM111.6 billion) in investments.

Hyperscale data centres are massive facilities housing thousands of servers, designed to serve tens of thousands of customers (think 10,000-40,000) and support large-scale applications like cloud computing and AI.

These centres are the most resource-intensive, consuming 300 to 1,000 megawatts (MW) of power and enough cooling. Other companies that operated hyperscale centres are tech giants like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.

While Malaysia has no hyperscale data centres yet, global players like Google are investing heavily in the country, with a US$2 billion (RM8.2 billion) data centre under construction in Elmina Business Park, Selangor.

Prime Minister Datuk Seri Anwar Ibrahim together with Alphabet and Google President and Chief Investment Officer Ruth Porat attend the groundbreaking ceremony for Google’s first data centre in Malaysia. — SoyaCincau pic

2. Privately owned data centres

Facilities owned and operated by individual organisations, such as banks or corporations, to store and process internal data. These centres are less resource-intensive than hyperscale facilities but still require significant cooling and power.

The benefits of these are organisations have full control over their data and can customise the facility to meet specific needs.

Some companies use a mix of private and public data centres, reserving private facilities for sensitive data and relying on public centres for less critical tasks.

3. Co-location data centres

These are facilities where businesses rent space to house their servers and IT equipment, sharing infrastructure like power, cooling, and security with other tenants.

While less resource intensive than hyperscale centres, co-location facilities still require robust cooling systems and reliable power supplies.

The benefits of having a shared space is its cost effective and scalable, allowing businesses to expand their IT infrastructure without building their own data centres.

Why do data centres need so much energy and cooling?

The short answer is to prevent servers from overheating and crashing.

These servers generate huge amounts of heat as they power everything at high speed, from social media platforms to AI applications like ChatGPT.

Such a workload comes at a cost.

Traditional air conditioning was once sufficient, but modern facilities now depend on water-based cooling methods, which are far more effective at absorbing and dissipating heat, making it the preferred choice for keeping servers at optimal temperatures.

A hyperscale centre would require from 300 to 1,000MW of power, which is like powering a small city. Building the infrastructure to generate that energy would take years and its water usage is staggering.

“A hyperscale data centre with its immense energy usage will require around six to seven swimming pools worth of water daily to cool the systems, daily.

“This problem of overreliance and usage of ChatGPT had at one point caused a severe strain on California’s water supply that the governor at one point asked the public to stop using ChatGPT,” said Charles Santiago, chairman of the National Water Services Commission (SPAN).

While Malaysia does not have any hyperscale data centre, even the smaller ones are cooled using water from the same sources for drinking.

SPAN has been highlighting the urgent need for contingency cooling systems.

Based on data from SPAN, Selangor houses 27 data centres, which consume 79 million litres per day (MLD) while Negeri Sembilan has a single data centre that uses four MLD.

Johor alone has 72 data centre applicants who have requested a combined 673 MLD; and so far, 17 centres have been approved, requiring 59.06 MLD of water.

Johor Menteri Besar Datuk Onn Hafiz Ghazi and Investment, Trade and Industry Minister Datuk Seri Tengku Zafrul Abdul Aziz visit the 150MW JH1 Data Centre Campus belonging to Princeton Digital Group at Sedenak Tech Park in Kulai on July 4, 2024. — Picture from Facebook/Onn Hafiz Ghazi

This growing demand for data centres, particularly those supporting AI, is putting immense pressure on Malaysia’s water and energy resources.

SPAN said 81 per cent of Malaysia’s drinking water comes from rivers while 17 to 18 per cent from dams and the remainder from underground sources.

However, pollution and climate change are reducing river yields, exacerbating water scarcity.

To address these challenges, SPAN is pushing for stricter regulations and sustainable practices.

Data centres are now required to have at least two days’ worth of water storage to ensure uninterrupted operations during supply disruptions.

Additionally, operators are encouraged to use alternative water sources, such as reclaimed water, rainwater harvesting, and desalination.

“Data centre operators and contractors must have alternative water sources such as reclaimed water, seawater and so forth. The use of alternative water sources depends on the availability and adequacy of such alternative sources in the proposed developmentareas of data centres.

“Therefore, during the stage of approval for application, water operators must assess the need for data centres to secure water supply from alternative sources other than treated water meant for drinking. Related policies will be proposed to be included in related subsidiary legislation under Act 655 or existing guidelines,” said Charles.

Act 655 refers to the Water Services Industry Act 2006.

Data centres currently use treated water — the same resource essential for drinking and cooking — to cool the intense heat generated by servers and electronic devices. — Picture by Sayuti Zainudin

SPAN proposed several reclaimed water projects and implementation of water supply infrastructure projects to increase reserve margins in the state of Selangor such as the Rasau Phase 1 Water Treatment Plant (WTP) (700 MLD), Rasau Phase 2 WTP (700 MLD), Langat 2 Phase 2 WTP (760 MLD).

Implementation of water supply infrastructure projects to increase reserve margins in the state of Johor such as the Semangar 3 WTP (200 MLD), Layang 2 Phase 2 WTP (160 MLD), Sedili Besar WTP (600 MLD) and Kayu Ara Pasong WTP (90 MLD).

SPAN has suggested several alternative sources, namely reclaimed water from Pelangi Indah Sewage Treatment Plant (STP) for Johor.

SPAN is proposing distributing reclaimed water from Perintis, Puteri Harbour, Medini, Bukit Indah, Serantau Pulai Mutiara and Kangkar Pulai STP in Johor to cool down the other data centres in Cyberjaya and Sungai Besi in Selangor, Nilai Impian in Negeri Sembilan and Kawasan Perindustrian Cemerlang in Johor Baru.

The challenge is that these WTPs are managed by different concessionaires: Indah Water Konsortium Sdn Bhd runs those in Selangor while Johor Special Water is the state’s utility company.

Despite these challenges, Malaysia is emerging as a regional hub for data centres.

Investment, Trade and Industry Minister Datuk Seri Tengku Zafrul Abdul Aziz said that since 2021, Malaysia has approved RM123.5 billion in data centre investments through the Malaysian Investment Development Authority, and the Malaysia Digital Economy Corporation.

The approved investments have positioned the country as a preferred regional data centre hub, and collectively, these investments are expected to create more than 64,000 high-value jobs.

Prime Minister Datuk Seri Anwar has touted the Google investment as proof of growing confidence in his economic policies as the tech giant is expected to create 26,500 jobs and contribute more than US$3 billion (RM13 billion) to Malaysia’s economy by 2030.

As technology advances and more and more ways are found to improve the use of data centres and energy.

China’s HiCloud is using seawater as a natural cooling source for its underwater data centre in Lingshui, Hainan Province, eliminating the need for using land and fresh water sources as well as saving on electricity consumption.

According to reports from China’s state-owned news broadcaster CCTV, the addition boosted the computing power of the submerged cluster to the equivalent of 30,000 high-end gaming computers operating simultaneously and could support 7,000 conversations per second with the Chinese AI chatbot DeepSeek.