Thank you to the Sydney Institute for the invitation to speak this evening.

The Institute has been a feature of public debate in Australia for over three decades. I want to acknowledge Gerard and Anne Henderson for their partnership and everything they have put in to building this forum.

I begin by acknowledging the Gadigal people of the Eora Nation, the traditional custodians of the land on which we are gathered today. I pay my respects to their Elders past and present, and extend that respect to all Aboriginal and Torres Strait Islander peoples here today.

Over the last two years, data centres have become one of the most contested pieces of economic infrastructure around the world. To advocates, they are the factories of a new age — a once-in-a-generation opportunity for Australia to capture a new wave of economic value, as well as securing and shaping its future. To critics, they are giant sheds full of computers, soaking up electricity and water, creating few jobs, and risking some of the same mistakes we made in the resources boom.

Both views contain important truths. And our job — as a country — is not to pick a side. It’s to see clearly enough to hold both ideas in the light, assess them with honesty, and make good decisions in the national interest.

There isn’t a question of whether the boom will happen. It’s happening. It’s accelerating. And it’s likely to be larger and more consequential than anything we’ve lived through in recent decades.

The real question – which I will return to many times - is simpler: Do we shape the boom before it arrives - or do we react after it’s here?

This means ensuring Australians have confidence that the growth of data centre sector will be sustainable when it comes to energy, water and local communities. It also means planning how we as a country will capture the economic upside of this growth, through investment, revenue and the buildout of our innovation ecosystem.

And so tonight, I aim to make a contribution that is neither boosterism nor alarmism. A contribution that is anchored in some uncomfortable lessons Australia learned from decades of experience with the resources sector.

That lesson, in my view, is not that Australia should blindly accept or reject investment — Rather, Australia should actively set the terms on which that investment occurs, consistent with our values and aligned with our long-term interests.

MORE PROFOUND THAN THE RESOURCES BOOM

When people ask, “Why all the fuss about data centres?” — it’s a fair question. Why, suddenly, is the entire world fighting over warehouses full of computers?

The answer is scale. Moody’s expects global data centre investment to reach three trillion dollars over the next five years. The annual capital budgets of individual firms like Amazon and Google now rival the economies of mid-sized nations. Comparable surges, such as railways, electricity and telecommunications, rolled out over generations. This one is similar in size, but it being compressed into a single decade.

The immediate effect of the data centre rush is a shot in the arm for global economic growth.

To give you a sense of how important this is to the global economy right now, hold these two numbers in your head: in the first half of 2025, investment in data centres and the information-processing equipment that fills them made up about four per cent of the United States economy, but accounted for roughly ninety-two per cent of its economic growth. Strip out that single category, and American growth in that period rounds to zero. Not one per cent. Not half. Zero. AI investment is propping up an otherwise sluggish global economy.

In Australia, the numbers are similarly meaningful. Over time, the AI boom may have a more profound impact on the Australian economy than the resources boom.

And this is the deeper reason for the global competition for data centres. Countries are not competing for computer warehouses. They’re competing for what those warehouses produce — intelligence, capability, economic growth and security. Computing power is rapidly becoming a necessary factor of production in its own right. In the nineteenth century, economic power rested on coal. In the twentieth, it rested on oil. In the twenty-first, it will rest on computation; and data centres are where computation is made. AI is the story. Data centres are simply where the AI happens.

Amid this global competition, Australia has quietly become one of the most attractive destinations in the world, ranked second globally in 2024, behind only the United States.

Why is Australia so attractive for these projects? The answer is fairly simple. If you are a global company choosing a location for a data centre, you’ll consider many factors including costs, taxes, planning and regulation. But there is one overriding criteria that sits above all of these. If you are making a ten billion dollar investment that will manage petabytes of sensitive data, your main fear is that your investment could become stranded during its 30 year useful life. To these investors, Australia offers an attractive combination of political stability, Five-Eyes security, and unrivalled potential for renewable power.

For these reasons, just as we were the lucky country during the resources boom, so we are turning out to be the lucky country in the new data centre boom. Investors want to be here. The question for us is: on what terms should we allow them to come?

THE BENEFITS OF DATA CENTRES

The answer to that starts with a clear-eyed assessment of the benefits and costs of data centres.

The first benefit is investment. A single hyperscale facility – the kind of large, compute intensive facility typically built and operated by major tech companies for their own use - can involve tens of billions. For Australia, total investment in different types of data centres could grow into the range of hundreds of billions of dollars. In the most recent National Accounts, investment in equipment and machinery approached record highs. In some states it is already rivalling the peak of the resources boom. However, because much of the heavy machinery and tech hardware comes from overseas, the overall net impact on domestic GDP is lower than the record capital investment figures suggest. In the first quarter of this year, the import of all those server racks caused Australia to record its first international trade deficit in ten years. The impact of this boom is positive, but, in terms of investment alone, not as big as its boosters might suggest.

The second potential benefit is jobs. The construction jobs that data centres generate are real. AirTrunk’s second Melbourne facility is expected to support more than four thousand workers. However once built, these places run with remarkably few people. If the argument for data centres were direct ongoing jobs alone, it would be a weak argument.

But direct jobs alone were never the real argument and if we stop here, we misunderstand the opportunity. The real argument is the one history makes about every great piece of economic infrastructure. Nobody remembers the railway boom for how many people it employed. We remember the railway boom because it was the infrastructure that made the industrial revolution possible. Data centres are the same. They are the engine room beneath cloud services, software firms, medical research, financial systems and the next generation of Australian start-ups. Their worth is not just what happens inside the shed. It is what they make possible everywhere else.

The third benefit is sovereignty. As a nation heading into a digital economy, we have to ask ourselves: Where are we comfortable having our citizens’ data stored? And increasingly, we need to be concerned with the deeper, related question: Where is our nation’s computing intelligence being produced? The nations that hold large-scale computing infrastructure will have far greater capacity to develop, deploy and shape advanced AI, and to do it under their own laws, for their own purposes. When Australian government activities, Australian research and Australian AI models run on Australian soil, under our Security of Critical Infrastructure regime, we are not a tenant in someone else’s digital economy. In a contested region, that is not a luxury. It is national resilience.

AI infrastructure can scaffold an innovation ecosystem to attract scientists and engineers with world-class capability, and retain Australia’s best science and technology talent before it goes offshore. Building a deep Australian talent pipeline and innovation capability is arguably the most important long-term driver to our economic success. The average age of top ten companies in the Australian ASX is over 120 years old. In contrast the average age of the S&P and NASDAQ 10 is over 40 years old. We need Australian-made AI to secure our economic future by enabling new businesses, new research and new industries we haven’t imagined yet. In a digital world, that matters. A country that depends entirely on critical infrastructure located elsewhere is, in a sense, renting its future. A country that helps build it — and shape it — has more control of its future.

Finally, if they are built within the right framework, data centres can accelerate our energy transition rather than threaten it. Globally, technology companies signed around forty per cent of all corporate renewable power-purchase agreements in 2025. A large, steady, predictable customer is exactly the kind of anchor that gets a new wind or solar farm financed and built, adding power to our grid that might not otherwise exist.

Designed as a flexible load, AI infrastructure can even dial their demand up and down to help steady the grid – with the potential to reduce retail cost pressures for consumers at peak times. It can be the reason a renewable project gets built, not the reason your power bill goes up.

The Government’s Data Centre Expectations released in late March this year outline three specific goals:

• data centres bring new and renewable energy supply to offset their demand
• data centres pay their full share of network infrastructure costs to ensure those are not passed onto consumers or businesses; and
• data centres provide demand flexibility and cooperate with market operators to strengthen the energy grid.

Commonwealth State and Territory Energy Ministers made positive progress on all four fronts in May.

Work has been tasked to expert market bodies, and we expect further progress when Ministers meet again in July to consider that advice.

These three elements – bring your own supply, cover your grid connection costs, and be demand flexible will be Australia’s triple lock.

THE COSTS OF DATA CENTRES

So that is the promise of data centres. If that were the whole story, the decisions we need to make would be easy. But that isn’t the whole story. And there are a range of legitimate concerns that needs to be discussed.

The biggest pressure is energy. Projects are lining up that would require power on the scale of major power stations, with limited plans for how that power will be supplied. If that demand arrives faster than supply, the consequences are obvious: Prices rise. Pressure builds. Households feel it. We’ve seen how this plays out elsewhere.

Today, data centres use around two per cent of the electricity in our National Electricity Market or roughly the consumption of seven hundred thousand homes. That sounds modest, but the trajectory is steep: demand nearly doubled in Victoria and rose eighteen per cent in New South Wales in a single year.

The New South Wales pipeline alone holds forty-four projects seeking eleven gigawatts of capacity. That is close to four times the output of Eraring, our largest coal-fired power station. Not all of it will be built. But large new loads arriving faster than generation and transmission can expand could push up prices for everyone if not carefully managed.

In the United States, there is already evidence that data centre demand is being fed directly into household power bills in the Mid-Atlantic. In the PJM grid (the largest in the country) the independent market monitor attributed sixty-three per cent of one year’s surge in capacity prices to data centres: more than nine billion US dollars, billed straight back to households, in a single year.

There are also legitimate community concerns about water. In 2024–25, Australian data centres consumed an estimated seven gigalitres of water for cooling. This is around 0.04 per cent of Australia’s total industrial water consumption. That’s around one per cent of the water consumed by mining, and about one and a half per cent of manufacturing, which is relatively small. But we know from overseas experience it can be locally acute. For example, in one Georgia county, residents reported taps running dry during construction and water bills rising by a third in two years. Market rules and new technology mitigate water impacts by requiring “closed loop” designs are here – which use up to ten times less water than their open loop predecessors, and are between 5 to 15 per cent more energy efficient too.

In addition to water, communities have expressed concern about accumulated noise of data centre clusters, visual impact, and the conflicts where industrial sites are proposed near homes.  People have real concerns that deserve real answers. We are seeing it right here in Sydney: a single proposed data centre at Lane Cove West drew a record 374 objections (and just nine submissions in support) from residents worried about noise, diesel backup generators, and a facility rising barely 160 metres from the local primary school.

The final category of costs are the opportunity costs. These are the costs an economist is obliged to account for. This massive investment boom is diverting other parts of our economy of investment, power, land and human resources. Australia needs to build more homes, needs more skilled trades people and is in the midst of a challenging energy transition. It is fair to ask whether that same electricity, land, capital and talent might generate more value elsewhere.

LESSON FROM THE RESOURCES BOOM

Some of Australia’s response to the data centre boom is shaped, rightly or wrongly, by our experience of the resources boom. And no part of that experience is rawer, or more topical, than gas.

We had abundance — but we didn’t lock in our advantage early enough. We built for export before securing supply at home. We became one of the largest gas exporters on earth and then watched households and factories pay more for gas dug up beneath their feet. Successive governments allowed global markets to effectively set the domestic price. We let the boom set the terms, instead of setting the terms of the boom. Australians remember that. And they are right to ask whether data centres will now do to electricity what exports did to gas: a giant new buyer, bidding up the price of something we all rely on.

The Albanese Government has moved to repair the gas market through a domestic reservation policy, setting aside Australian gas for Australians. That is the right fix. But it is a fix arriving a decade after the damage was done. That is the real lesson from gas for data centres: it’s far better to get the rules right at the outset than to try to put them right a decade later.

LESSONS FROM OTHER COUNTRIES

If that is the lesson we learned at home on gas, it is worth asking what lessons other countries have learned from the data centre boom abroad, because almost every other country that has faced this boom acted only after the strain appeared.

The United States largely let the market lead. The result was extraordinary scale and extraordinary backlash. The rapid growth of data centres has lost social licence in many parts of the country. States are now implementing improvised, nationally inconsistent, after-the-fact fixes, with utilities only now introducing rules that make data centres pay for the capacity they reserve.

Ireland is another cautionary tale. It drew the global technology giants with tax, talent and access to Europe, and it worked at first. Then data centres passed ten per cent of national electricity demand. The grid around Dublin effectively seized up, and new connections were frozen. Only then did regulators require new centres to source the bulk of their power from new domestic renewable generation. The right policy was arrived at too late.

Singapore moved faster and harder. Facing acute limits on land and power, it simply halted new approvals in 2019. When it reopened the door, it did so through competition: you no longer secure capacity by turning up, but by being the most efficient and the greenest, against strict energy-performance targets. Quality-constrained growth, by design.

The Nordic countries took a different road again, turning abundant clean power and cool climates into a pitch as the world’s most sustainable home for digital infrastructure.

And China has treated computing capacity as strategic national infrastructure, steering investment into designated hubs as a matter of industrial policy.

The pattern is unmistakable, and it confirms the lesson of gas. The countries that managed the costs best were the ones that leaned in and set the terms.

AUSTRALIA’S APPROACH

Australia’s approach has been to move quickly to get safeguards into the system while we can shape it. We want to learn the lessons of the resources boom and not repeat the mistakes of other countries.

Under our National AI Plan, in March this year the Government released its Expectations of data centres and AI infrastructure developers. These are not red tape. They are the foundation of the sector’s social licence to operate here: a plain statement of what we expect in return for access to our grid, our land and our market.

We expect data centres to underwrite new renewable power supply. That means they must bring their own generation, not draw down everyone else’s. We expect them to pay their full share of grid connection, so those costs are never passed to households and businesses. We expect them to support the system through demand flexibility, to be a grid asset rather than a grid burden.

Bring new supply, cover your network costs, and be demand flexible –Australia’s triple lock.

On top of this we also expect the global operators to make their computing power available to Australian start-ups and to partner with Australian researchers so that the benefits of this build-out flow to Australian innovation, not merely Australian real estate. That last expectation matters most, because it goes to the question underneath this entire debate.

The AI economy is coming whether we are ready or not. That much is settled. What is not settled is our place in it. We can be a country that merely consumes artificial intelligence built somewhere else, renting our intelligence by the month, on terms set in another hemisphere. Or we can be a country that helps build it: that owns the infrastructure, trains the people, grows the companies, and shapes the technology to our own values and our own interests.

That is the real choice. Not whether the boom arrives, but whether Australians end up tenants in someone else’s digital future, or owners of our own.

We have the leverage and natural advantages now, at the start of the data centre boom, to shape how this unfolds.

We can learn the lessons of our past booms, like LNG, and lessons from countries overseas. If we set the conditions early this boom could leave us with more capacity than we started with, not less.

When we secure investment, we don’t just secure construction jobs, or renewable energy financing. We win the right to host the services, shape safety, influence and train the models, and run the platforms of the digital economy. This foundation also positions Australia to capture the revenue and economic growth that will follow.

But that only happens if we act early. Not after the system is under pressure. Not after people see their communities impacted. Not decades after all the investments are sunk.

That is why the Albanese government is acting and will continue to set the conditions for data centre investment. We will continue to work closely with the states and territories and keep all options on the table when it comes to asserting our national interests.

For decades Australia got the sequence wrong with gas. We do not have to get it wrong again.

That is why getting data centres right matters. And that is what we intend to do.

Thank you