Charlton's local AI push collides with Chalmers' startup tax fight

Andrew Charlton wants Australia to buy local AI. In a speech this month, the assistant minister for science, technology and the digital economy argued the country is in a “critical window” that will decide whether it builds a domestic AI industry or becomes “a permanent renter of intelligence from abroad”. The line lands. It names a real strategic anxiety: that local companies end up paying overseas model vendors for the most important software layer of the decade while the jobs, margins and intellectual property pile up elsewhere. But Charlton’s message is landing at the same time as Jim Chalmers’ 2026-27 tax reform package changes the after-tax maths on startup exits. For founders, that does not look like a joined-up industry strategy. It looks like two parts of government arguing with each other.

For an early-stage software company, especially one trying to build around AI, patriotic procurement and capital gains tax are not separate debates. One tells customers and agencies what to buy. The other shapes whether founders, staff and early investors decide the risk is worth taking in Australia rather than in a deeper, faster-moving market. The budget says the 50 per cent capital gains tax discount will be replaced by inflation indexation and a 30 per cent minimum tax on capital gains from 1 July 2027. Treasury can make a clean-policy case for taxing real gains rather than windfalls flattered by inflation. But in a sector that runs on equity promises long before profits arrive, the timing could hardly be worse. The change lands just as ministers are telling the market it needs more local AI companies, not fewer.

Charlton is not wrong about the market problem. If Australian businesses, universities and government agencies default to US platforms for every model, copilot and workflow tool, local founders will struggle to win reference customers at home — the kind that help with recruiting, fundraising and selling abroad. His call for buyers to “lean into buying Australian AI technology and services” is procurement advocacy wrapped in the language of industry policy. It assumes local vendors can win enough contracts to survive the long stretch between first product and meaningful scale. This is where the tax fight turns into something bigger than a Treasury detail. AI startups burn cash on engineers, compute and sales before they generate steady revenue. They recruit senior staff with equity, not just salary, and rely on early backers staying patient for years. That funding loop is why AFR reporting on the startup carve-out fight has landed so hard in the sector. According to the Financial Review, investors including Seek co-founder Paul Bassat have been pressing Treasury to treat startups differently rather than force them into a regime designed for more conventional capital gains. The argument is not simply that founders want a lower tax bill. It is that startup returns are unusually skewed: most bets fail, a few winners carry the fund, and employee equity only matters if the eventual gain still looks meaningful after years of risk and dilution. In AI, where Australia is already competing with US pay packets, offshore venture pools and faster procurement cycles, founders say the country can ill afford to make local equity feel less valuable at the exact moment ministers are urging customers to back domestic builders.

The government’s defence is that the budget is not just taking with one hand. Alongside the CGT overhaul, the 2026-27 budget productivity measures include refundable startup losses for companies with turnover up to $10 million, a move the government says will help about 25,000 young businesses each year. That is a material concession for companies still in the expensive early stage, when payroll, cloud bills and compliance costs can easily outrun revenue. Treasury can fairly argue the package still counts as pro-innovation. Cash relief earlier in a company’s life can be more useful than a distant tax concession if a founder is trying to keep engineers on staff or stretch a runway. The problem is that the levers work at different points in the company life cycle. Loss relief helps on the way in. Exit taxation shapes behaviour on the way out.

Canberra is subsidising startup formation while making the eventual payoff look less compelling.

There is also a practical policy problem hiding beneath the rhetoric. Once Chalmers signalled he wanted to get the CGT change “right” for startups, the debate shifted from principle to line-drawing. John Storey, chief executive of Slyp, told ABC News there “shouldn’t be carve outs” because exemptions would make an already complex tax system more complex. Storey’s point is a serious one. Any carve-out has to answer awkward questions about sector definitions, holding periods, turnover thresholds and who counts as an active founder rather than a passive investor. Yet refusing to draw a line is a choice as well. It means treating a speculative investment in an early-stage AI company much like any other capital gain, even though policymakers spend the rest of the year saying they want more local risk-taking, more commercialisation and more Australian-owned technology champions.

Charlton’s speech, seen through this lens, is not a complete policy. It is one part of a larger, still unresolved bargain. Telling Australian organisations to buy local AI may help create demand, and it may become more important if geopolitical tensions or model concentration make reliance on offshore suppliers look riskier. But domestic demand alone will not build durable champions if the surrounding capital settings convince founders and staff that the upside is better captured elsewhere. The question for Australia is not whether it should want its own AI companies — of course it should. The harder question is whether the tax, procurement and startup policy settings are pointing in the same direction. Right now the answer is mixed. That is why the fight over capital gains has become a tech policy story, not just a Treasury one.