A compute-for-equity deal between Innovation Victoria and Heidi Health, reported by Capital Brief, is being framed as smart economic policy. No it isn’t.
Heidi Health is on the brink of signing what is believed to be the first compute-for-equity deal of its kind in Australia, according to Capital Brief.
The structure, revealed at a Joint Select Committee on Artificial Intelligence hearing in Sydney this week, involves Innovation Victoria and AI infrastructure builder ResetData providing Heidi with access to GPU compute in exchange for an equity stake in the company.
A heads-of-agreement has been signed. The parties expect to finalise the deal within days, though the value has not been disclosed.
ResetData co-CEO Marcel Zalloua described it to Capital Brief as part of the company’s broader Australian AI Growth Program, a structure designed to funnel subsidised compute to local startups, with government and institutional partners underwriting the cost.
Heidi, which secured a $1.26 billion valuation only last month off the back of a USD100 million raise led by Blackbird with a growth investment from General Catalyst, is the foundational partner.
“Being able to partner with the AI darling of Australia in Heidi Health is a great validation,” Me Zalloua said.
Innovation Victoria CEO Rod Bristow told Capital Brief the agency is “continually exploring opportunities to strengthen Victoria’s innovation ecosystem and support the growth of high-potential companies”, but declined to comment on specifics.
Heidi’s spokesperson said innovative funding structures that help companies access compute “can remove a significant barrier to growth and strengthen Victoria’s AI industry”, while cautioning that no arrangement has been finalised.
The program was developed between Innovation Victoria, ResetData and Nvidia, after Innovation Victoria announced plans to collaborate with Nvidia through its local cloud partners in September.
A strategy with precedent and risk
The compute-for-equity model was not invented here. OpenAI popularised it in May 2026, when Sam Altman offered Y Combinator founders up to USD$2 million in AI tokens in exchange for equity.
Nasdaq-listed neocloud CoreWeave launched a venture arm offering the same structure in September 2025.
The logic is that for AI-native companies burning through GPU cycles at scale, it can be more valuable than cash.
Heidi II – the agentic platform Heidi launched last month – is a materially more compute-intensive product than its predecessor, running persistent agents across a clinical day rather than processing a single consult.
Containing that cost through a deal rather than paying hyperscaler spot rates directly improves Heidi’s unit economics.
None of that makes this a risk-free investment for Victoria. It makes it a reasonable investment structure for Heidi.
The day Firmus chose to fall apart
Industry observers will have noted that the same day the Heidi deal surfaces, Firmus – Australia’s most-hyped AI infrastructure company – is in the process of watching its $44 billion ASX float implode.
Fund managers have noted that 97% of Firmus’ promised data centre capacity has not been built, that its valuation jumped from $6.9 billion to $15.5 billion in six months, and that the company is projecting US$30 billion in debt once construction is complete.
A high-profile CDC partnership announced last year has already been dissolved. One co-founder was jailed in 2016 for insider trading.
Heidi isn’t Firmus of course. It has a real product, users and revenue.
But it has a risk profile that would make anyone question why Victoria, essentially a broke state, would make the investment now.
Related
Who is qualified to make this call?
The AI investment cycle is running very hot, valuations are moving faster than fundamentals in some cases, and government agencies are being asked to make calls in a sector that is evolving faster than most investment committees can track.
Who is really qualified in the Victorian government to make such a risky investment call? Even if things work out for Heidi, the investment is in a global play and Victoria is Victoria … a state in a faraway land, with money problems.
Innovation Victoria has previously backed healthcare AI companies including Diag-Nose.io, Deep Derm and AlleSense.
Backing diagnostics startups at early stage is a different exercise from taking an equity position in a company valued at $1.26 billion, in a sector – clinical AI – that that is simultaneously dealing with unresolved questions about regulatory liability, data sovereignty, practice management software terms of service, and hospital procurement cycles.
Heidi is not operating in a protected market. Microsoft, Google, Epic and a growing list of purpose-built clinical AI companies are all moving into the same space.
Heidi’s current advantage is clinician adoption and product velocity. Whether that is durable enough to justify a billion-dollar valuation – let alone deliver returns to a government equity investor – is a difficult question for the best of the world’s investment firms but we know that in Heidi’s last round no US based firm would touch it as a new investor.
That is a very bad sign in AI today.
There is also the geographic mismatch to consider. Innovation Victoria is spending Victorian taxpayers’ money on compute infrastructure that will support a company whose customers are clinicians across every state and territory in Australia but mostly internationally.
The compute may run on Victorian infrastructure. The value creation, if it happens, flows to Heidi’s cap table – dominated by Blackbird, General Catalyst, and the company’s founders.
Victoria’s equity stake will be a small slice of that. The jobs, the tax base, the long-term economic dividend that would justify the risk – it is not obvious how much of that actually lands in Victoria.
Mr Zalloua’s argument to the investment committee – that without this kind of support, AI startups will move to the US to access the capital they need to scale – is misleading.
Heidi couldn’t get any new investors from the US in its latest round.