2026-07-14 · 8 min read · Raw Drafts / Artificial Intelligence / Infrastructure
The AI Future Is Becoming a Land-Use Problem
The AI race is moving from models and chips into planning approvals, electricity grids, water systems, and the local politics of where intelligence is allowed to live.
For years, the artificial intelligence race has been described as a competition between models. Which company has the smartest system? Which country has the most advanced chips? Who has the best researchers, the largest training clusters, and the deepest reserves of capital?
Australia's emerging national AI framework suggests that the next phase of the race will look considerably less glamorous. It will take place through planning approvals, electricity grids, water allocations, construction projects, and negotiations between different levels of government.
The AI future is becoming a land-use problem.
The Australian government is establishing an Office of AI within the Department of the Prime Minister and Cabinet. Its purpose is to bring together work that has previously been handled issue by issue across copyright, employment, education, defence, national security, and infrastructure. One of the immediate problems it is expected to address is the fragmented way Australian states approve data centres.
This might initially sound like an administrative detail. It is actually a glimpse of what AI becomes when it leaves the demonstration screen and enters the physical world.
A chatbot appears weightless. You type into a box and receive a response from somewhere called the cloud. The interface removes almost every sign of the industrial system behind it. There are no smokestacks, turbines, substations, or cooling towers on the screen. The result arrives with the apparent effortlessness of thought.
Behind that moment is a rapidly expanding network of data centres, power-generation facilities, transmission lines, cooling equipment, fibre connections, and semiconductor supply chains. The International Energy Agency puts the relationship plainly: there is no AI without energy, specifically the electricity consumed by data centres.
The cloud has always occupied the ground. AI is making its physical footprint much harder to ignore.
Australia has already started defining what it expects from companies building this infrastructure. Data-centre developers are being asked to support the energy transition, cover their share of transmission and distribution costs, use water responsibly, create Australian jobs, strengthen local research capability, and demonstrate that their investments serve the national interest. The government describes these expectations as the foundation of the industry's social licence to operate.
The phrase social licence reveals how much the conversation has changed. AI companies once presented themselves as software businesses offering digital services. They increasingly resemble mining, energy, and property-development companies. Their growth depends on access to finite resources and public infrastructure. Their projects affect surrounding communities. They require political approval, long-term construction, and negotiations over who pays for the systems they need.
The cloud is becoming a local resident.
Once this happens, the politics of AI begins to resemble the politics of every other large physical development. A data centre may promise investment, tax revenue, jobs, research capacity, and strategic importance. It may also consume electricity that must be generated somewhere, use water that already has other claimants, occupy valuable land, and require costly upgrades to the grid. The benefits may be distributed nationally or flow primarily to a foreign technology company. The pressure on infrastructure is experienced in a particular town, district, or state.
This produces a fundamental mismatch of scale. AI is sold as a national opportunity and a global necessity, while many of its consequences are negotiated locally.
A prime minister can announce that a country intends to become an AI power. A local planning authority still has to decide whether a particular building can be constructed on a particular piece of land. An energy provider must determine whether the grid can support it. A water utility must assess the local supply. A community has to decide whether the promised economic benefits justify the demands placed on its surroundings.
National ambition eventually arrives as an application form.
This is where the AI race may be won or lost. The leading countries will need more than talented researchers and access to chips. They will need electrical capacity, functioning planning institutions, public trust, and places willing to host the infrastructure. The speed of development will depend partly on model architecture and partly on whether a transmission line can be approved and built.
That does not make the technological race less important. It means the race has expanded. Intelligence has become dependent on urban planning, environmental policy, energy markets, and regional development.
The familiar map of AI power centres around Silicon Valley, Shenzhen, London, and major research laboratories. A more revealing map might show substations, available grid connections, water stress, industrial land, renewable-energy projects, and the jurisdictions capable of approving construction without creating a political backlash.
The geography of intelligence may ultimately follow the geography of electricity.
This also changes the relationship between governments and technology companies. For much of the internet era, states often appeared to be responding to systems that had already become global. Social networks, search engines, and digital advertising platforms grew faster than the institutions attempting to oversee them. Their infrastructure was physical, but the public experienced them primarily as services delivered through screens.
AI infrastructure gives governments a different kind of leverage. A company may operate globally, but its data centres must exist somewhere. They require connections to national energy systems, access to land, water, and cooperation from public institutions. Governments can attach conditions to those needs.
The Australian framework points toward this emerging bargain. Technology companies receive clearer and potentially faster approval processes. In return, governments expect them to contribute to energy supply, manage water responsibly, create local capabilities, and demonstrate benefits beyond the boundaries of the facility.
The question is whether those conditions will have real force.
A framework can coordinate government departments while still leaving the largest decisions to the companies making the investments. Faster approval can become the main objective, with sustainability and public benefit treated as supporting language. The promise of joining the AI economy may encourage governments to compete against one another by offering cheaper power, easier land access, or weaker requirements.
Australia's patchwork of state approaches already reflects this tension. Some states created special pathways to attract data-centre investment. As public concern grew, the conversation shifted from attracting any project toward attracting the right projects, those that produce wider economic benefits while managing pressure on energy, water, and nearby communities.
The same conflict will appear elsewhere. Every government wants the jobs, investment, and strategic capacity associated with AI. Few governments want rising electricity prices, stressed grids, environmental pressure, or large industrial projects that generate limited local employment after construction ends.
AI policy can therefore no longer remain confined to model safety, copyright, and automation. It must address who supplies the physical resources, who pays for new infrastructure, how the costs are distributed, and who receives the economic value created by it.
These questions are political because land and energy cannot be copied infinitely. A software product can theoretically serve another million users with little visible change to the interface. The infrastructure behind those interactions still has limits. At some point, more intelligence requires another building, another grid connection, another cooling system, and another agreement with the place that must host them.
The most significant AI disputes may therefore emerge far away from the laboratories producing the models. They may appear in municipal meetings, environmental assessments, utility negotiations, and public arguments over power lines. They may involve farmers, construction unions, property developers, water authorities, and communities that have never thought of themselves as participants in the technology industry.
This is how technological futures usually become real. The dramatic invention receives the attention. The slower process of fitting it into the physical world determines what it eventually becomes.
Railways had to cross privately owned land. Cars required roads, parking, fuel systems, and the reconstruction of cities. Aviation required airports, air-traffic systems, and agreements over noise. The internet required cables, spectrum, and data centres. AI is now entering the same stage of development, when the technology stops appearing as a discrete invention and starts reorganising the systems around it.
Australia's national AI framework is therefore interesting for reasons larger than its specific policies. It acknowledges that artificial intelligence has become an infrastructure question.
The next generation of AI will still be shaped by breakthroughs in algorithms and computing. It will also be shaped by zoning rules, energy markets, water policy, grid capacity, and the ability of governments to negotiate with companies whose resource demands are growing alongside their ambitions.
The artificial intelligence industry is becoming physical enough to argue about.
The future may still arrive through a chat window. Somewhere behind it, a planning department will have approved the building.
Credit
This essay was prompted by Clare Armstrong's ABC News report, *Anthony Albanese maps out AI future with new national framework*, published on July 14, 2026. The report identified the fragmented approval of data centres across Australian states as a central reason for the government's new national framework. The land-use framing and broader interpretation developed here are my own.
AI authorship note
This essay grew from Kris Haamer's questions about Australia's new national AI framework and his observation that the strongest Raw Drafts angle was the transformation of AI into a land-use problem. AI supported source research, drafting, editing, and implementation. Kris selected the central thesis, rejected an earlier generic governance angle, and made the decision to publish the article with explicit credit to the original ABC News reporting.