Firmus Technologies eyes record ASX float as Australian AI company targets valuation of up to $90 billion

As trillions of dollars pour into artificial intelligence, one of the industry’s biggest problems is no longer finding customers. It’s finding enough electricity, data centres and computing power to serve them.

Australian-founded business Firmus Technologies is developing data centres packed with specialised computers that can run AI models, and plans to sell access to that computing capacity to companies like OpenAI that need it to serve users.

The business plan is relatively simple, what’s more complex is is putting a value on a seven-year-old company whose biggest projects aren’t even built.

Firmus wants to list on the ASX later this year at a valuation estimated between $US30 billion to $US60 billion ($45 billion to $90 billion), according to news reports.

This would make it the biggest sharemarket listing in Australian history, with an eye-popping valuation that reflects bets AI will prove a revolutionary technology for users and wealth creator for large stakeholders.

The company’s ambitions extend across Australia and Asia, with proposed AI factories in South Australia, Tasmania, Malaysia, Singapore and Indonesia.

Divides opinion

This week, its management team has been meeting Asian investors as it seeks the cash required to build physical data centres also known as AI factories.

But the mooted valuation up to $US60 billion is dividing Australian investment circles, almost as sharply as the AI boom itself.

“You have two groups of fund managers those that want to assassinate it (Firmus) because they’re not there (invited to buy shares), and those that want to cheer it because they are,” said Gary Rollo a Portfolio Manager at Montgomery Investment Management.

Investment bankers and Firmus’ management have reportedly told potential investors that the company is targeting $US5 billion of earnings before interest and tax (EBIT) in the 2028 financial year.

On that forecast, they argue it can be fairly valued at $US60 billion on 12 times forecast EBIT.

That calculation and whether the forecasts are realistic is at the heart of the debate.

Much of Firmus’ construction plans and financial forecasts only exist on drawing boards and spreadsheets.

The forecasts for the company’s future profits therefore depend heavily on projects that are yet to be completed or operated at scale.

The main evidence supporting the growth projections is the portfolio of contracts Firmus has signed with customers including large tech companies including Instagram-parent Meta and ChatGPT-owner OpenAI.

Those agreements commit customers to paying for data-centre capacity on the assumption that Firmus can deliver the projects on time, or at all.

Firmus also needs to raise huge amounts of cash by borrowing money, or issuing shares, to fund its giant construction projects.

That leaves investors making a judgement not only about the future demand for AI computing, but about Firmus’ ability to turn contracts and development plans into profitable operating assets.

Comparison to CoreWeave

The comparison reportedly being put to investors is CoreWeave, the US-listed AI infrastructure company whose shares trade on the Nasdaq exchange in the US.

Firmus’ investment banker advisers have reportedly argued that a multiple of 12 times forecast EBIT would be reasonable because it represents a discounted valuation to CoreWeave, which has briefly traded at 17 times forecast EBIT in 2026.

The comparison is straightforward. CoreWeave also builds and leases data centres containing the high-performance computing equipment required for AI and investors often use listed peers to value companies coming to market for comparable valuations.

This is much like a house’s value can be compared to a similar house sold down the road on the same street.

But Rollo says the comparison with CoreWeave glosses over an important difference, because the US group has already demonstrated that it can build and operate multiple facilities and deliver contracts for demanding customers.

“We haven’t seen that yet with Firmus, all it’s done has sign contracts,” he says. “Firmus are also probably arguing these contract counterparties are the smartest guys in the room, so why would they sign a contract with us if they didn’t think we could do it? And there’s some validity to that.

“But I don’t think many Australian fund managers are persuaded by the argument the smartest blokes in the room think we can do it, so don’t worry about it, and just trust us on these big numbers we’re putting out.”

The distinction between signing contracts to rent out data centres and actually building them is important because the biggest risk for Firmus investors is not a lack of demand from its customers.

It’s whether the company can build enough data-centre capacity, at a low enough cost, to generate attractive returns on the capital being invested.

The economics could also change in that today demand for AI computing is so strong that customers are willing to pay huge sums in contracts for access to scarce capacity. If supply catches up with demand, those contracted prices could sink.

Firmus itself declined to comment on its contracts or what valuation it expects to achieve at IPO.

Debt risks

Elsewhere for the company, the vast amount of borrowed money required to build AI factories creates other risks.

Firmus will need substantial amounts of debt to finance construction and the cost of borrowing has been soaring with global interest rates.

On Tuesday, the US central bank, the Federal Reserve, lifted benchmark cash borrowing rates 25 basis points to 3.75 to 4 per cent.

Today, an investor can buy a “risk free” US or Australian Government 10-year bond and earn a steady, or linear 5 per cent a year return.

Firmus therefore needs to generate returns comfortably above 5 per cent to compensate investors for the risks involved in construction, financing, tech, regulation, competition, and customer demand.

“[Firmus] is obviously an area of the market unlikely to be linear,” says Rollo.

“You’re not going to put money into it and sleep nicely for the next three years. It will be volatile. And there’s also the monster operational execution for management to deliver on.

“But if it all goes well, could you make a good solid return? Of course, and that’s the dangle the bankers are putting in front of the investors.

“Those investment bankers do a great job [telling the media and investors to value it like CoreWeave] because operationally Firmus has achieved very little, but commercially quite a bit. And they’re saying we want you to value it at a modest discount to CoreWeave, but I think for people to pay that level will be generous.”

Firmus’ secret weapon

Australia’s biggest AI star also argues that one of its tech advantages is its use of liquid cooling (rather than conventional air cooling) in its data centres across a system it calls ‘Hypercube’.

According to the company, its AI servers are cooled with a specialised fluid that pulls heat from high-density computing equipment far more efficiently than conventional air-conditioning.

Firmus boasts to investors the ‘Hypercube’ tech can reduce the water and power required to operate its data centres, which lowers the cost of running them versus peers.

But the claims of low costs and increased efficiencies in sales pitches to investors are yet to be tested in the real world.

“Firmus say we’re building 30-year assets for fractions of the unit costs of capital than other [peers] and those things will last that duration period, but that’s not proven,” Rollo says.

“The assets are likely to be different to those the big boys like CDC [Data Centres] and NextDC are building for a much higher unit capital cost, but we won’t know any payback until we get through the lifecycle of these assets and understand their true economics and how much capital they require.

“So, maybe it’s a good trade, but maybe it isn’t?”

Prison barber to AI billionaire

Another question hanging over the proposed sharemarket float is why Firmus has chosen Australia at all.

It was co-founded in the country by Sydneysider Oliver Curtis and University of Melbourne graduate Tim Rosenfield, but most of its operations are in Asia, where Curtis is a resident of capital gains tax haven Singapore.

All its AI factory peers in CoreWeave, SharonAI, Nebius and IREN trade on the Nasdaq, which offers the deepest pool of tech investors and vast networks of AI expertise.

This week it has been courting overseas asset managers in Asia and its cornerstone debt and equity investor is Wall Street royalty in Blackstone.

So why the ASX? Firmus declined to comment.

The decision has prompted speculation that Curtis’s 2016 conviction and imprisonment for insider trading made a US listing more difficult. His story from prison barber to AI billionaire in 10 years is one of the harder ones to sell to Wall Street.

Rollo declined to comment on the issue.

Instead, he argues that a major tech listing would be a huge win for the Australian market after a period in which rising bond yields and higher interest rates have made IPOs more difficult.

This is because large investors are less willing to risk capital losses in exchange for uncertain future capital growth, if they can earn returns in excess of 5 per cent over 10 years for zero risk on bonds.

“We in Australia should welcome businesses like this and it’s right there’s a healthy argument about how much investors should pay,” says Rollo. “Firmus are talking about investing a significant amount of capital into Australia and deserve a run at meeting the big investors.”

There may also be a more technical benefit to an ASX listing.

A company like Firmus worth $50 billion dollars would instantly be large enough to attract significant S&P/ASX 200 index-fund ownership, forcing passive asset managers that track the market to buy the stock after it hits the boards.

“That’s also an opportunity for bankers to exploit here,” says Rollo. “If you list something for a large value in a small market like Australia it will get significant index fund buying. And the bankers can go to [the large fund managers] and represent this [Firmus] is going in the index that will give it a boost, and it is too big for you to ignore.”

Over the weeks ahead, Australia’s biggest AI star - on paper - is looking for professional investors who believe the story that its future profits will justify the sky-high valuations being angled for.

Once that step is complete, the next test for Curtis and Rosenfield will be to deliver on its construction ambitions.