Housing’s new rival: Why AI capital could change Australia’s property economy

Australia’s housing argument is still being fought as if the old economy will last forever: more people, more workers, more mortgages, more roads, more suburbs and then, somehow, enough homes.

But a different argument is beginning to form beneath it. What if the next economy does not need labour in the same way? What if capital no longer sees housing as the safest place to compound? What if the next great claim on money, power and energy is not land, but compute?

That is the question raised by veteran macro investor Jordi Visser, whose recent conversation with Anthony Pompliano captured a growing split in global markets. Visser’s point was simple enough: investors are still reading interest rates through the memory of an economy driven by housing, cars and debt, while the marginal growth engine is shifting towards artificial intelligence.

“Part of this fear over rates is the history of an economy that was driven by housing and was driven by autos,” Visser said. In his view, the AI buildout has far less sensitivity to rates than the old economy because the leading models and infrastructure companies have high margins, urgent demand and customers still racing to adopt. “This is all about AI agents. This is all about consumer agents. This is all about enterprise adoption,” he said. “Rates are just not going to be as important as they’ve been in the past.”

For Australia, that lands awkwardly. No developed country has made housing more central to its household balance sheet, banking system and political economy. The Australian Bureau of Statistics put the value of residential dwellings at about $12.8 trillion in the March quarter of 2026, an asset class larger than the sharemarket, the super system and household deposits combined ABS. Housing credit now accounts for about 62 per cent of total private credit, making property not just a social issue, but the core of bank balance sheets.

That is the old model. Population growth feeds housing demand. Housing demand feeds bank lending. Bank lending supports land prices. Rising land prices support household wealth. Household wealth supports consumption. Governments then try to manage the pain with supply targets, grants, guarantees and planning reform.

AI optimism ranges from the cautious to the extreme. Goldman Sachs has argued generative AI could lift global GDP by 7 per cent and raise productivity growth by 1.5 percentage points over a 10-year period Goldman Sachs

The political class sees the problem as supply because, in the present tense, that is where the pain is. Rents are high, vacancy rates are thin, and the National Housing Accord target of 1.2 million homes remains hard to meet. The National Housing Supply and Affordability Council forecast about 980,000 homes during the Accord period, with the target pushed beyond the original timetable NHSAC. The Productivity Commission has found Australia now completes roughly half as many homes per hour worked as it did in 1995 Productivity Commission.

So the supply argument is not wrong. It is incomplete.

The deeper question is whether Australia is still allocating capital around an economic model that may be fading. Data centres are no longer a niche infrastructure class. AEMO-linked forecasts prepared by Oxford Economics Australia estimate Australian data centres consumed 3.9 TWh of electricity in FY25 and could reach 12.0 TWh by FY30 under the Step Change scenario, about 6 per cent of NEM grid-supplied electricity AEMO/Oxford Economics Australia. The Clean Energy Finance Corporation says data centres could account for up to 11 per cent of national electricity consumption by 2035, up from about 1 per cent in 2025 CEFC.

That is capital with a different logic. It wants power, cooling, fibre, chips, secure sites, sovereign data rules and customers who need computation. It does not need a kitchen renovation, a mortgage broker and a negative-gearing schedule. It is still physical infrastructure, but its yield is tied to information demand, not population growth alone.

Visser’s more radical claim is that AI agents change the unit of economic demand. Businesses may not need to hire in the same numbers if a small group of people can manage agents that work continuously. “They never sleep and they’re always available,” he said. He described holding a staff meeting with his AI agents because that was how he would manage humans. The point was not science fiction. It was that labour can be reorganised before it is formally replaced.

“The deflationary pressures from AI are far more extreme than people realise,” Visser said, adding that people may not feel the full effect until humanoid robots become real

This matters for immigration. Australia’s recent growth model has depended heavily on population inflow. ABS figures show net overseas migration of 306,000 in 2024-25, down from the post-pandemic peak but still the main driver of population growth ABS. If the economy continues to need labour at scale, migration remains part of the answer. If AI and robotics reduce the need for marginal workers, the case changes.

But this is where the theory needs discipline. AI optimism ranges from the cautious to the extreme. Goldman Sachs has argued generative AI could lift global GDP by 7 per cent and raise productivity growth by 1.5 percentage points over a 10-year period Goldman Sachs. Daron Acemoglu of MIT is far more restrained, estimating AI may lift total factor productivity by only about 0.71 per cent over 10 years, or around 0.07 per cent a year, with GDP effects of roughly 0.9 to 1.1 per cent if investment effects are modest MIT.

Both cannot be right in the same way. The difference turns on how much work AI can actually do, how quickly firms adopt it, and whether it attacks hard physical tasks or mostly improves paperwork.

Australia’s housing market shows the trap. Robots may make construction less labour-intensive. Prefab, 3D printing, automated approvals and AI-assisted design can help. But housing is not priced only by labour. In Sydney, RBA research found zoning restrictions added $489,000 to the price of an average detached house in 2016, raising prices 73 per cent above the cost of supply RBA. In Digital Media & Co.’s Property Economy paper, the arithmetic was blunt: if robotics halved construction costs, the Sydney house-and-land package might fall only about 17 per cent, because land and permission still do most of the work.

That is why “we can manage with what we have” is only partly true. Australia may not need the same labour intake in an AI-heavy economy. But it still has household formation, ageing, divorce, students, renters, internal migration and location demand. It also has a housing stock badly distributed between where people want to live and what they can afford. Lower labour demand would reduce pressure at the margin. It would not magically unlock well-located homes.

The more likely future is uneven.

Over the next three to five years, housing remains politically dominant because rents and mortgage payments are felt every week. Governments will keep arguing over supply, migration caps, investor taxes and first-home-buyer schemes. Prices may soften if rates stay high and investors retreat, but a deep national collapse still needs a bigger trigger: unemployment, forced selling, credit stress or a genuine loss of faith in land as a store of wealth.

Over five to 10 years, capital allocation becomes the larger story. If AI infrastructure delivers better growth than property, the prestige of housing as the default Australian investment weakens. Super funds, private credit, banks and high-net-worth investors will put more money into digital infrastructure, energy storage, grid connections, chips, automation and businesses that use agents to cut costs. Property will not disappear. It may simply stop being the only game in town.

Over 10 to 20 years, the immigration question could change. A country that can grow output with fewer workers does not need to run the same population machine. But that would require governments to replace population-led GDP growth with productivity-led GDP per capita growth. That is a harder political bargain. It means accepting slower headline growth, rethinking university funding, rebuilding tax bases, and admitting that “more people” has often been used as a substitute for productivity reform.

Visser’s point is useful because it breaks the spell of the old rate-sensitive economy. If the next economy is driven by AI infrastructure and agents, then the old transmission channels weaken. Higher rates hurt households and small businesses, but may not stop the AI firms, chip companies and data-centre developers. “The deflationary pressures from AI are far more extreme than people realise,” Visser said, adding that people may not feel the full effect until humanoid robots become real.

That is the moment when housing’s value proposition faces its hardest test. A house is valuable because land is scarce, planning is restrictive, credit is available and workers need to live near jobs. If work becomes less tied to human labour, and if more economic value is created inside data centres than office towers, the location premium changes. Office property has already learnt this lesson. Residential will learn it more slowly because families still need shelter.

The political danger is that Australia fights the last war. It may build an economy around ever-rising population while capital is already moving towards machines that reduce the need for labour. It may subsidise housing demand while the productive frontier asks for energy, transmission, compute and advanced manufacturing. It may protect land values while younger workers discover that their wages are competing not only with migrants, but with software agents and robots.

The forecast, then, is not that Australian real estate becomes worthless. It is that its era of automatic outperformance is under challenge. Scarce land in good locations will hold value. Poorly located, debt-heavy, low-yield property will be exposed. Construction labour may be disrupted, but planning will still ration supply. Migration may slow, but household demand will not vanish. Data centres will attract more capital, but they will also collide with power prices, grid politics and community resistance.

Australia’s old bargain was that population growth would keep the property machine turning. The new bargain, if it comes, will be harsher and more interesting: productivity instead of people, compute instead of credit, energy instead of land, and capital chasing assets that can grow without adding another lane to the freeway.

That is why the housing debate feels slightly out of date. Supply still matters. Immigration still matters. But the bigger question is whether Australia is prepared for a world in which labour is no longer the binding constraint, and housing is no longer capital’s safest story.