Why the Sudden Rush to Build Data Centres?

If it feels as though data centres have suddenly appeared everywhere, you're not imagining it.

Western Sydney. Melbourne. Regional Victoria. Canberra. Brisbane. Perth.

Projects measured not in a few megawatts, but hundreds.

Individual campuses approaching — and in some cases exceeding — the electricity demand of major industrial facilities.

Billions of dollars of investment are being announced. Global technology companies are expanding. Governments are talking about fast-tracking infrastructure and attracting international capital.

But data centres aren't new.

We've been using cloud computing for years.

Governments moved services online years ago. Businesses migrated to the cloud. Australians have been streaming television, storing photographs, banking online and using social media for well over a decade.

So why the sudden rush?

The answer is artificial intelligence.

Something changed in 2022

The emergence of generative AI dramatically accelerated the global race for computing power.

When ChatGPT was released publicly in late 2022, generative AI suddenly became visible to hundreds of millions of ordinary users. At the same time, competition between the world's largest technology companies to develop increasingly capable AI models intensified.

Those models require enormous amounts of computing power to train and operate.

Microsoft, Google, Amazon, Meta and others accelerated investment in AI models, specialised chips and the infrastructure required to support them.

Governments also began treating access to advanced computing capacity as an economic and strategic issue.

Computing power was no longer simply another part of digital infrastructure.

It was becoming an economic and geopolitical asset.

And that requires data centres.

Lots of them.

These aren't yesterday's data centres

This distinction matters.

The public discussion often treats "data centres" as though we're talking about one homogeneous type of infrastructure.

We're not.

Australia unquestionably needs data centres.

Government services need them.

Banks need them.

Hospitals need them.

Defence and national security agencies need them.

Universities and researchers need them.

Australian businesses need cloud storage and computing capacity.

And as Australia develops its own AI capability, our domestic demand for advanced computing infrastructure will increase.

But the scale of the new generation of projects raises a different question.

Are we simply building the infrastructure Australia needs?

Or are we building infrastructure in Australia to service the global AI economy?

Those are two very different propositions.

The global AI race needs somewhere to live

Artificial intelligence may feel intangible.

It isn't.

Every AI query, every model being trained and every cloud service ultimately depends on physical infrastructure.

Buildings.

Servers.

Semiconductors.

Cooling systems.

Electricity.

Water.

Land.

Transmission infrastructure.

Fibre connections.

Backup power.

The digital economy has a very physical footprint.

And as AI models have become larger and more computationally intensive, the infrastructure required to support them has expanded dramatically.

That has triggered a global search for locations capable of hosting the next generation of computing infrastructure.

Australia has some obvious attractions.

We have political stability.

Large amounts of land.

Strong international connectivity.

An established financial and regulatory system.

Potential access to enormous quantities of renewable electricity.

And we sit within the rapidly growing Indo-Pacific region.

Our government has noticed.

Australia doesn't just want to participate. It wants to become a hub.

This is the part of the data centre story Australians need to understand.

The current expansion isn't occurring in a policy vacuum.

The Australian Government's National AI Plan treats advanced data centres as "smart infrastructure" and seeks to attract international investment into Australian AI and computing infrastructure.

Government agreements have gone further, explicitly describing Australia's potential role as a trusted regional hub for AI infrastructure serving Australia and the broader region.

Australia is therefore being positioned not simply to meet Australian computing needs, but as a location from which global technology companies can service a much larger market.

That could represent a significant economic opportunity.

But it fundamentally changes the question Australians should be asking.

If a data centre exists primarily to support Australian government, business, research or national-security needs, there is a clear domestic rationale for committing Australian resources to it.

If a facility is being built largely to service international AI and cloud customers, the proposition is different.

It may still be enormously beneficial to Australia.

But the benefit needs to be demonstrated.

The scale tells us something

The numbers involved can be difficult to comprehend.

In May 2026, CDC Data Centres announced a 555 megawatt contract with a single United States customer.

On its own, 555MW doesn't mean much to most people.

So consider it this way.

According to CDC Data Centres shareholder Infratil, that one contract is equivalent to around 40 per cent of the operating capacity of every data centre in Australia in 2025.

One customer.

One contract.

The equivalent of almost half the capacity of Australia's existing data centre industry.

And the agreement runs for 30 years, with renewal options that could extend it by another 20.

That tells us something important about what is driving the current expansion.

This isn't simply Australians storing more photographs in the cloud or businesses moving their accounting systems online.

We are talking about a global market for computing capacity operating at an entirely different scale.

And CDC Data Centres isn't an isolated example.

Near Morwell in Victoria, Singapore-headquartered Keppel has secured a 123-hectare site for a potential 720MW AI data centre campus. Keppel says it is already in discussions with hyperscalers and "neocloud" providers about using that future capacity.

In Western Sydney, the proposed Mamre Road Data Centre Campus is larger again.

The NSW Planning Portal describes a campus with 1GW of power capacity, six four-storey data centre buildings, 936 cooling units, 852 diesel backup generators and 14.43 million litres of diesel storage.

One gigawatt.

For one data centre campus.

These aren't server rooms.

They are major pieces of industrial infrastructure.

And their scale tells us why the distinction between Australian need and international demand has suddenly become so important.

At these scales, major projects can represent decades-long commitments of electricity, land, water and supporting infrastructure — although headline megawatt figures should be treated carefully, because contracted capacity, maximum demand and actual operating consumption are not necessarily the same thing.

That doesn't make these bad projects.

But before Australia commits those resources, we should know what the infrastructure is for, who will use it and what Australia receives in return.

But who's adding it all up?

This is where the scale of the current expansion raises another problem.

A 555MW contract can be assessed as one commercial development.

A 720MW campus can be assessed as another.

A 1GW proposal can move through its own planning and approval process.

But Australia's electricity system, water resources, transmission networks and communities don't experience these projects one at a time.

They experience the cumulative demand.

And that cumulative picture is becoming enormous.

In the March quarter of 2026, 11 large data centre projects representing around 5.4GW of maximum demand were progressing through the transmission connection process in the National Electricity Market.

That does not mean 5.4GW will all be built or consumed. Many projects remain at early stages, and data centres typically ramp their electricity use over time.

But that uncertainty is itself part of the problem.

Australia needs to know which projects are credible, which are speculative, how much capacity is likely to become operational, and what the cumulative resource requirement will be if several large projects proceed simultaneously.

Governments are already actively encouraging data centre investment and positioning Australia as a regional AI and cloud hub. Yet we are still trying to establish something much more fundamental: how much data centre capacity is actually being proposed across Australia, how much electricity and water it will collectively require, and how much of that capacity is intended to meet Australian needs rather than international demand.

That sequencing matters, and carts before horses rarely end well.

Because Australia risks discovering the aggregate scale of the commitment after the strategy to attract the industry is already well underway.

The question shouldn't simply be whether each individual project can secure enough electricity, water and land to proceed.

It should also be:

What happens when we add them all together?

And what if the technology changes?

There is another question that seems almost absent from the rush to build.

What if we don't need these enormous data centres forever?

Look at the history of computing.

The first commercial computers occupied entire rooms. Computing power that once required enormous machines is now carried around in our pockets.

Technology gets smaller.

Chips get faster.

Cooling improves.

Software becomes more efficient.

More computing power is squeezed from less physical infrastructure.

There is no reason to assume artificial intelligence has somehow reached the end of that technological progression.

In fact, I think it would be extraordinary if the enormous AI data centres being proposed today represented the permanent physical form of computing for the next half-century.

Perhaps demand will grow so quickly that efficiency gains are swallowed entirely by increased use. That's certainly possible, and current forecasts expect data centre demand to keep growing strongly in the immediate future.

But that's a forecast.

Technology has a habit of making long-term forecasts look rather silly.

And Australia is making decisions now that could last for decades.

The 555MW agreement announced by CDC Data Centres with one US customer runs for 30 years, with options that could extend the relationship for another 20.

Fifty years.

Think about computing technology fifty years ago.

Then try to imagine confidently predicting what computing infrastructure will look like fifty years from now.

Yet Australia is being encouraged to commit electricity generation, transmission infrastructure, water, land and other resources around today's expectations of what the AI industry will require.

I think that's an extraordinary gamble.

Perhaps these enormous campuses will still be essential in 2050.

Perhaps they won't.

Perhaps future AI chips will use a fraction of today's electricity.

Perhaps computing will become far more distributed.

Perhaps technologies we haven't invented yet will make today's hyperscale facilities look like the room-sized computers of the past.

We simply don't know.

And that uncertainty should make governments more cautious, not less.

Because the technology companies aren't offering Australia a guarantee that today's infrastructure model will remain economically valuable for the next fifty years.

Australians, however, may be left with the consequences of the decisions being made around it.

Transmission lines.

Generation infrastructure.

Water infrastructure.

Industrial land committed for decades.

Environmental impacts.

Public expenditure.

And potentially enormous facilities whose original technological purpose has changed or disappeared.

That doesn't mean we shouldn't build data centres.

It means we should stop behaving as though today's AI boom has already revealed the permanent shape of tomorrow's economy.

We are making infrastructure decisions measured in decades for a technology industry that changes in years.

That alone should demand far more caution than we're seeing.

Who are we building them for?

This is one of the central questions raised in SUNA's Data Centres briefing paper.

Australia's domestic need for secure digital infrastructure and the economic case for becoming an international AI infrastructure hub are often spoken about as though they are the same thing.

They aren't.

Both may serve Australia's national interest.

But they should not be assessed using the same evidence.

Domestic infrastructure should be judged partly by whether it gives Australians the secure computing capacity required for government, business, defence, research and economic participation.

Export-oriented infrastructure needs an additional test.

If Australian electricity, water, land and infrastructure are being committed to providing computing services to international customers, what does Australia receive in return?

Jobs?

Tax revenue?

Australian ownership?

Australian businesses in the supply chain?

Research capability?

Skills?

Intellectual property?

Access to advanced computing?

Regional development?

Or simply construction activity and foreign investment?

These are not arguments against data centres.

They are the questions that should come before approval, not after construction.

Demand does not establish national interest

Perhaps the most important distinction in this entire debate is between demand and need.

There is clearly enormous global demand for AI computing capacity.

Technology companies want it.

Investors want to finance it.

Developers want to build it.

Governments around the world want to attract it.

But the existence of international demand does not, by itself, establish that every proposed Australian project is in Australia's national interest.

Australia has finite resources.

Electricity committed to one industry cannot simultaneously be used by another without additional generation.

Water allocated to one facility cannot be used somewhere else.

Industrial land has competing uses.

Transmission infrastructure costs money and takes years to build.

And communities live alongside these facilities.

The bigger the AI race becomes, the more important these trade-offs become.

Australia still has time to ask the question

The United States is already showing what can happen when data centre development races ahead of community confidence and infrastructure planning.

In June 2026, Australia's own Assistant Minister for Science, Technology and the Digital Economy acknowledged that legislation to pause or ban new data centre construction had been introduced in at least 11 US states, while community opposition had blocked or delayed billions of dollars of projects.

Australia is earlier in the build-out.

That should give us an opportunity to avoid making the same mistakes.

But there is an uncomfortable contradiction in the Government's current approach.

It acknowledges the risks. Its own expectations for data centre developers address electricity generation, transmission costs, water use, community impacts and pressure on the electricity system.

At the same time, it is actively pursuing the investment, promoting Australia as a regional AI and data centre hub and encouraging an industry whose potential cumulative resource demand we are still trying to establish.

That is backwards.

Before deciding that Australia should become a regional hub for one of the most resource-intensive industries of the AI era, Australians should have been given some basic answers.

How much data centre infrastructure does Australia actually need?

How much additional capacity are we seeking to build for international customers?

How much electricity, water, land and transmission infrastructure will the strategy require if it succeeds?

What are the cumulative impacts when all of the projects are considered together?

And what enduring benefit will Australians receive for committing those resources?

These aren't obscure technical details to work out after the investment has arrived.

They are the questions that should inform the strategy in the first place.

We don't need to choose between embracing every project placed before us and rejecting data centres altogether.

There is a much more sensible position.

Build the infrastructure Australia needs.

Develop Australian AI capability.

Attract international investment where it produces substantial national benefit.

But before committing increasing quantities of Australian electricity, water, land and infrastructure to the global AI race, establish what we're building, who we're building it for and what Australians receive in return.

Because every major data centre represents a long-term commitment of Australian resources.

The question is not whether Australia should build data centres.

The question is whether the benefits to Australia justify the resources being committed.

This is a SUNA opinion and analysis piece. It draws on research undertaken for our Data Centres briefing paper, together with publicly available government, industry and economic data. It reflects our analysis of Australia's emerging data centre strategy and the questions we believe Australians should be asking about its long-term national benefit.