Why Are So Many Data Centres Being Built, and What Will They Be Used For?

TLDR at the bottom of the article. Data centres are being proposed at an extraordinary rate across Australia and around the world.

The immediate explanation is artificial intelligence. Technology companies need enormous amounts of computing power to train AI models and operate services such as Microsoft Copilot, ChatGPT, image generation, automated customer service and AI-assisted search.

However, AI is only part of the story.

Our cars, shops, phones, workplaces and security systems are also collecting more information than ever before. As organisations find new ways to analyse this information, demand is shifting from simply storing data to processing it continuously.

This raises two important questions:

1. Is the current data-centre boom justified by genuine demand?

2. What are the privacy and environmental costs of building this infrastructure?

Existing data centres are not necessarily suitable for AI

It may appear that we already have enough data-centre capacity for websites, cloud storage, Microsoft 365 and other existing services.

The problem is that many older facilities were not designed for modern AI hardware.

AI systems use extremely powerful processors, high-bandwidth memory and specialised networking equipment. These systems consume far more electricity and produce significantly more heat than conventional business servers.

A facility may have available floor space while lacking the power, cooling or network infrastructure required for high-density AI equipment.

This is why many companies are constructing entirely new facilities rather than simply adding more servers to existing buildings.

The real bet is that AI will become part of everything

The technology industry is not only planning for more people to use AI chatbots.

It is betting that AI will become a background component of almost every digital service.

This could include:

• software automatically summarising and processing information

• AI assistants performing administrative tasks

• automated cybersecurity monitoring

• real-time translation and voice processing

• computer vision analysing security footage

• vehicles interpreting road and driver information

• retailers monitoring stock, checkouts and customer activity

• AI-generated images, audio and video

Training an AI model requires considerable computing power, but operating that model for millions of users can require even more over time.

The International Energy Agency reported that data centres consumed approximately 415 terawatt-hours of electricity globally in 2024, around 1.5% of worldwide electricity use. It projects consumption could rise to approximately 1,200 terawatt-hours by 2035.

Surveillance is increasing, but some claims are exaggerated

Retail surveillance has expanded significantly.

Australian supermarkets now use CCTV, self-checkout cameras, product-recognition systems, overhead cameras and other technologies intended to detect theft, improve stock availability and protect staff.

This does not mean every camera is performing facial recognition.

Woolworths states that its general CCTV, staff safety cameras and Scan Assist checkout technology do not use facial recognition. Its checkout systems instead analyse products and movement around the checkout area.

Facial recognition has, however, been trialled by major Australian retailers.

Bunnings used facial recognition in 63 stores across Victoria and New South Wales between 2018 and 2021. Customers’ faces were compared with a database of people considered to present a safety or criminal risk.

In 2026, the Administrative Review Tribunal upheld findings that Bunnings had failed to provide appropriate notice and maintain sufficient privacy governance. It also found that limited use of the technology to address serious retail crime could potentially be lawful under strict conditions.

The important distinction is that a system can monitor behaviour without necessarily knowing someone’s identity.

For example, a camera system may follow an anonymous person through a store, detect an unscanned item and retain a short video of the transaction. That is still surveillance, but it is different from creating a permanent biometric profile linked to a person’s name.

Modern cars collect valuable personal information

Connected vehicles are another genuine privacy concern.

Depending on the manufacturer and services enabled, a vehicle may collect:

• precise location information

• journey history

• speed and acceleration

• hard braking events

• diagnostic information

• battery and charging activity

• voice commands

• infotainment usage

• driver-assistance camera data

Claims that ordinary cars continuously upload terabytes of information are generally misleading.

Modern vehicles may generate enormous amounts of raw sensor data internally, particularly when they have multiple cameras or advanced driver-assistance systems. Most of that information is processed within the vehicle and discarded, overwritten or reduced to smaller events and statistics.

The more realistic concern is that relatively small amounts of data can still be highly revealing.

In the United States, the Federal Trade Commission took action against General Motors and OnStar over allegations that they collected and sold precise location and driving-behaviour information without adequate informed consent. The data reportedly included speeding, hard braking and late-night driving and could be used by consumer-reporting agencies involved in insurance decisions. A final order was approved in January 2026.

A company does not need every second of camera footage to understand someone’s habits. Location points, purchase records, device identifiers and driving events can be enough to create a detailed behavioural profile.

Data centres make large-scale analysis possible

Surveillance footage and connected devices are not solely responsible for the current construction boom.

Simply storing transaction records or driving logs is relatively inexpensive. The more demanding workload is analysing that information at scale.

Previously, security footage might only have been reviewed after an incident. Modern computer-vision systems can potentially examine footage continuously for:

• missed checkout scans

• suspicious behaviour

• aggressive incidents

• unauthorised entry

• vehicle number plates

• stock shortages

• workplace safety risks

Similarly, vehicle information can be used to improve driver-assistance software, predict mechanical failures, develop subscription services or assess insurance risk.

The transition is therefore not merely from less data to more data.

It is from passive records to automated decision-making.

Are we moving towards social credit scoring?

China’s social credit system is frequently used as a comparison.

Contrary to some popular descriptions, China does not operate one universal numerical score covering every citizen. Its system is a collection of government records, regulatory ratings, financial information, court-enforcement blacklists and local programs.

The concern in countries such as Australia is more likely to involve a decentralised version.

Separate organisations may independently assess people through:

• credit scores

• insurance risk profiles

• fraud-detection systems

• retail watchlists

• identity-verification services

• employment screening

• online platform reputation systems

• government compliance databases

None of these systems alone amounts to a nationwide social credit score.

However, problems can emerge when information is inaccurate, shared between organisations or used to make decisions without transparency or a meaningful appeal process.

The greatest risk occurs when data becomes:

• personally identifiable

• interconnected across services

• retained indefinitely

• used to make automated decisions

The issue is not the existence of data centres themselves. It is how the information within them is collected, combined and acted upon.

The environmental cost is substantial

Modern data centres require enormous amounts of electricity.

The International Energy Agency expects renewable energy to supply a large share of future demand, but also projects increased natural-gas generation and pressure on already constrained electricity grids. It estimates that approximately 20% of planned data-centre projects could face delays if grid and infrastructure constraints are not addressed.

Environmental impacts can include:

• additional gas or coal generation where renewable supply is insufficient

• diesel-generator emissions during testing and outages

• water consumption for cooling

• increased electricity use from waterless cooling systems

• emissions from manufacturing servers and semiconductors

• mining and processing of copper and specialised metals

• electronic waste from rapid hardware replacement

• new substations and transmission infrastructure

The effect depends heavily on where and how a facility is built.

A data centre operating alongside abundant renewable energy will have a different impact from one added to a congested electricity network that still depends heavily on fossil fuels.

Local communities also need confidence that the cost of grid upgrades, water use and supporting infrastructure will not be unfairly transferred to households and other businesses.

Is the data-centre boom a bubble?

There is undoubtedly real demand.

Cloud services continue to expand, AI workloads are extremely resource intensive, and the amount of information generated by connected devices is increasing.

However, that does not guarantee that every proposed data centre will be needed or financially successful.

Large technology companies are building capacity before demand is fully proven because data centres, electrical connections and supporting infrastructure can take years to develop. Being caught without enough capacity could leave a company unable to compete.

This creates an arms race in which overbuilding may appear safer than underbuilding.

The result could resemble the telecommunications boom of the late 1990s. The internet ultimately justified enormous infrastructure investment, but many individual companies and projects still failed because too much capacity was built too quickly.

Both conclusions may therefore prove correct:

• AI and cloud demand may grow enormously.

• Some data centres may still be unnecessary, poorly located or commercially unsuccessful.

What should businesses take from this?

For most small businesses, the practical lesson is not to abandon cloud computing.

Cloud services provide genuine advantages, including remote access, collaboration, disaster recovery, scalability and enterprise-grade security capabilities that would otherwise be difficult for smaller organisations to maintain.

The better response is to understand what information the business collects and where it goes.

Businesses should regularly review:

• what customer and employee information they retain

• whether all collected information is genuinely required

• which cloud providers and third parties can access it

• how long information is retained

• whether AI tools are receiving confidential material

• whether camera systems use facial or biometric recognition

• whether automated decisions can be reviewed by a person

• how customers and employees are informed about monitoring

• whether outdated information is securely deleted

The future of computing will involve more data centres, more connected devices and more automated analysis.

The challenge is ensuring that increased capability does not come at the cost of privacy, accountability and responsible energy use.

Technology should serve people and businesses, not quietly remove their ability to understand or challenge how decisions are being made about them. TL;DR

The rapid growth of data centres brings genuine benefits, including better cloud services, improved cybersecurity, more capable AI tools and new ways for businesses to work efficiently.

However, it also creates legitimate concerns around electricity use, pollution, water consumption, surveillance and the increasing use of personal information to make automated decisions.

The issue is not simply whether data centres or AI are good or bad. It is whether companies and governments are transparent about what information is collected, how it is used and what environmental costs are being created.

Consumers are not powerless. The services we choose, the permissions we grant and the privacy standards we demand can influence how these technologies develop. Meaningful change is most likely when people remain informed, support businesses with responsible practices and push for laws that protect privacy, transparency and the right to challenge automated decisions.