The capacity being built dwarfs what has landed
At the G20 summit in North Carolina, Elon Musk told the assembled leaders that within ten years there would be "well over a billion humanoid robots" and that "the productivity per robot will be probably five times that of a human", making "the billion humanoid robots more productive than all humans combined".[^2] He called that a conservative estimate and said he would put serious money on it.
Musk's forecast is a forecast, and it should be read as the view of the person building the machines rather than as settled fact. But the shipment data is not a forecast. According to Smart Analytics Global, as reported by news.com.au, roughly 19,100 humanoid robots shipped in the first half of 2026, more than triple the 5,100 units shipped in the same period the year before. The research firm expects around 60,000 units this year and half a million by 2030.[^2]
The machines are not coming. They are arriving. The question is what we are building our migration settings around.
282,570 study visas into the work that goes first
In 2023, 737,170 people arrived in Australia under the ABS 12-in-16-month residency rule, against 219,080 departures, producing a net figure of 518,090. Of the 636,450 foreign nationals in that arrivals count (the remainder were returning Australian and New Zealand citizens), 282,570 arrived on study visas: 44.4% of all foreign-national arrivals.[^3]
These are NOM arrivals under the ABS residency rule, not visa grants or border crossings, and the three measures are not interchangeable. But the composition is the point, and it holds across 2019, 2022 and 2023 regardless of the total.[^3]
Students who pay $50,000 for a degree, as Prof Laura Shepherd of the University of Sydney noted of her own students,[^1] are entitled to expect a job at the end of it. The study visa pathway, as designed, allows onshore conversion from a study visa into work, and from work toward residency. The Commonwealth's own Rapid Review into the Exploitation of Australia's Visa System found that the student pathway has been used by some providers and agents as a route to work rather than study, and named the systemic settings that permit it.[^5]
The argument here is about those settings, not about the students who navigate them. A student who follows the pathway as it exists is doing exactly what the policy invites. The failure is the policy design: it is issuing study visas at scale into the occupations that large-language-model deployment absorbs first, and it is doing so without any mechanism that adjusts the intake as those occupations contract.
That is a settings failure. It is also a promise the government cannot keep.
The headline is not the whole number
Net overseas migration peaked at 530,700 in 2023 and has since fallen: 330,500 in 2024, 300,800 in 2025.[^6] The composition argument does not require a rising total. It requires only that the intake remain concentrated in the most automation-exposed occupations, which it does.
And the headline understates the standing population effect. A permanent residency grant is not a one-off event. It creates a future entitlement to sponsor family members. The migration program runs a separate Family stream with its own annual planning level, published by the Department of Home Affairs.[^4] Student visas carry dependants. A government that plans only to the headline net migration figure is planning to the wrong number, because the downstream family entitlements created by each year's grants run for years afterward. The real population effect is higher than the headline. By how much is a function of take-up rates across the Family stream, and that figure is not in the data supplied here, so this piece will not invent one. The mechanism is real and it is citable. The multiplier is not this section's to supply.
What it has done to the people already here
The ABS national accounts show that GDP per capita fell in 7 of the last 12 quarters while headline GDP rose, and in 25 of 104 quarters since 2000 (latest data 2026-Q1, chain volume measures, seasonally adjusted, quarter on quarter).[^7] Output per person went backwards while the headline number looked healthy. Population growth is one input among several, including hours worked, terms of trade and investment. But a growth model that adds people faster than it raises output per person produces exactly this result, and the result is measured and uncontested.
The living cost indexes tell the same story from the household end. From 2012-Q1 to 2026-Q2, living costs for other government transfer recipient households rose 51.7%, for pensioner and beneficiary households 50.0%, and for age pensioner households 48.5%.[^8] These are ABS Selected Living Cost Indexes, which include mortgage interest charges and are not the CPI. They measure index changes for a household type, not what any individual household experienced. But the direction is consistent and the period is long enough to be structural.
Housing completions, infrastructure spend per head, hospital capacity and public services all carry the load of a larger population. Where this section has no sourced figure for a specific strain, it will say so and argue the mechanism rather than reach for an adjective.
Labor's record
Net migration peaked at 530,700 in 2023 under Labor's settings.[^6] Per-capita GDP fell through that period.[^7] The Productivity Commission has found that migration "does not offer a long-term panacea" because "immigrants age too".[^9] The dependency-ratio argument for high intakes does not survive contact with that finding.
The criticism here is of the settings and the record. It is not a criticism of any person's character.
One Nation's policy, and why the cap is the floor
One Nation has published its immigration policy at onenation.org.au/immigration.[^10] The numbers are on the page: cap visas at 130,000 a year, a reduction of over 570,000 from current Labor levels; deport 75,000 people unlawfully present under existing law; end the student visa settings that allow study to convert into a backdoor to permanent residency or low-wage labour; an eight-year wait for citizenship and welfare access.
Those are One Nation's published figures. This section's standing position goes further, and it is worth saying plainly: the cap is the floor, not the ceiling. The intake should keep falling year on year, paced against measured automation displacement, for as long as large-language-model deployment and humanoid robotics keep absorbing the work the intake is sized for. One cut that is then allowed to drift back up is the pattern both major parties have already run. A schedule that continues downward is the only version that is not that pattern.
One Nation is the only party with a number on the table that goes anywhere near the required scale. That is an argument from record, not from character. The Liberals and Nationals have presided over high intakes in government. Their current position is a change of mind, which is welcome, but it is not a demonstrated conviction. One Nation brings the one thing neither has shown in office: willingness to cut the figure hard and hold it there when the business lobby complains.
The coalition this section argues for has One Nation leading on this question, with the Liberals and Nationals as partners who bring the numbers in parliament and the ministerial experience to govern. That is the shape of the proposition. It is not an argument against the Liberals or the Nationals. It is an argument about which party holds the line.
The technology policy gap, and what to build
One Nation publishes 29 policies at onenation.org.au/policies.[^11] None of them cover AI, automation, data infrastructure or telecommunications. A party about to be a governing force will have to form one as the data centre build-out accelerates. This section argues what it should be, from One Nation's existing positions on foreign ownership, energy and sovereignty.
The Foreign Influence Transparency Scheme register records people and entities acting on behalf of foreign principals, meaning foreign governments and foreign political organisations. Registered foreign-principal lobbying on AI, data-centre, critical-infrastructure and government AI procurement policy should be barred, and the register published in real time. The argument is about foreign governments and their registered agents as a category. It is not about any diaspora community, any religion or any ethnicity.
Resource and energy exports reached around $405 billion in 2025-26, up from $385 billion the year before, and are forecast at $416 billion in 2026-27, according to the Department of Industry, Resources and Energy Quarterly for June 2026.[^12] That is the base. Cheap domestic energy is what makes Australian-owned compute viable, and Australian-owned compute is what keeps the value of AI deployment inside the country rather than flowing to offshore vendors. The prize is lower power bills, lower taxes and no debt carried forward onto the people already here. It is built from the resource base Australia already holds.
What the lower number looks like
Housing completions that outpace household formation. Infrastructure spend per head that rises instead of being diluted. Wage growth in the occupations most exposed to automation, because the labour market has time to adjust rather than being continuously restocked. Public services sized for the population that is actually here.
That is not a warning. It is a description of what a drastic, continuing reduction in the intake buys the people already here. One Nation is the party with a number on the table. The coalition that gets it done has One Nation leading, the Liberals and Nationals governing, and a technology policy built from the sovereignty and energy positions One Nation already holds.
The canary told us where this goes. The question is whether the next government is listening.