Where Value Accrues in Australia’s AI Infrastructure Build

Firmus Technologies withdrew its $7.2 billion ASX float on 9 October 2026, after seeking a valuation of about $44 billion. The listing would have been Australia’s second-largest IPO, behind Telstra in 1997. The asking price was roughly three times the valuation set in a private funding round two months earlier.

Datt Capital examined the float in September, before the bookbuild opened. The withdrawal now tests that analysis, and it raises a more durable question for investors: where value accrues across the AI infrastructure build, and which layers hold the pricing power to justify a premium.

Public investors were asked to pay three times the private-round price

Datt Capital received the Firmus term sheet. The gap between the private and public valuations was the central issue.

An IPO price has to leave room for new investors to earn a return. A threefold step-up in two months, with little change in the company’s operating assets, leaves new shareholders relying almost entirely on future execution for that return. Even at half the proposed valuation, the uplift on the private round would have been close to twofold.

“It was obviously being pitched at a valuation that we thought was incredible rather than credible, especially given the recent private financing round that was conducted at a valuation one third of what the company was seeking from public market investors.”

Firmus had two operational data centres, in Melbourne and Singapore. Most of the valuation rested on facilities still to be built.

The middle of the AI value chain is the most commoditised layer

The AI value chain has three broad layers. At the top sit chip designers and model developers, where intellectual property creates pricing power. At the end sit the businesses that apply AI to their own operations. In between sit the data centre developers and operators that sell computing capacity.

That middle layer competes largely on cost and availability. Capacity from one operator is broadly interchangeable with capacity from another, so pricing power is limited. The business is also capital-intensive, and much of its value depends on cash flows several years away. That makes valuations sensitive to interest rates. With the US 10-year Treasury yield at 5.23 per cent, its highest level in more than two decades, distant cash flows are worth less in today’s terms.

Emanuel Datt sees investor expectations adjusting to that reality.

“Infinite growth is no longer being expected by investors in the space, which does affect valuations.”

The same logic applies across commodity markets. As set out in [Where Value and Profits Accrue Across Today’s Energy Markets]([URL if a web version exists]), value concentrates where supply is constrained and pricing power is strongest, and it rarely spreads evenly across every participant in a value chain.

Service providers capture the build-out only when their clients are funded

Datt Capital’s September analysis of the Firmus float argued that value in the AI build-out sat with the service providers building it, and that execution risk sat with the platform. The first part of that view holds. The second needs refining.

Australia has no structural advantage in frontier AI model development. It does need local AI facilities, and someone has to build, power and equip them.

“Companies that are providing services such as in the construction or provisioning of these centres are probably the most sustainable or most obvious candidates who’ll benefit from the roll out in these AI facilities locally,” Emanuel said.

Service providers earn revenue during construction, years before an operator’s facility produces its own cash flow. That timing is a real advantage. The failed float showed its limit. As doubts over the listing grew, a listed contractor holding more than $1.2 billion in Firmus work fell 21 per cent in a single session. Execution risk sits with the platform, while funding risk travels down the supply chain to every contractor whose order book depends on that platform raising capital.

Investors assessing an AI infrastructure supplier can apply three checks:

1. Client funding certainty. Is the client’s build-out funded with committed capital, or does it depend on a future raise?
2. Customer concentration. What share of work on hand comes from a single developer?
3. Balance sheet exposure. Does the supplier also hold equity in its client, doubling its exposure to one outcome?

These checks separate suppliers with diversified, well-funded clients from those concentrated on a single developer that is still raising capital.

A withdrawn float shows public market pricing discipline at work

The withdrawal has been described as a setback for the ASX, which has had few large technology listings in recent years. Emanuel takes a different view.

“I actually don’t think it’s a loss for the ASX. I think that ultimately it’s a business model that is still unproven.”

Index inclusion adds weight to that view. If Firmus had joined major share indices, index funds would have been required to buy it, giving superannuation members exposure without an active decision. Institutional investors declined the price before that point, which is the role public market pricing is meant to play.

Portfolio relevance

Since September, large caps have driven the Australian market’s gains, while small caps have been sold as investors reduced risk. Many of the local businesses positioned to service the AI build-out sit in the small cap segment, and relative value against global peers is drawing takeover interest.

“It seemed that global investors are recognising the opportunity, and it’s probably just a matter of time until our local investors do as well.”

For investors, the Firmus withdrawal is a prompt to separate the AI theme from the price paid for it. Exposure to the build-out is better assessed supplier by supplier, on the funding strength of each client as well as the size of the contract book.

Conclusion

Firmus sought a public valuation three times its last private round, for a business in the most commoditised layer of the AI value chain. Service providers remain the more durable way to gain exposure to the local build-out, provided their clients are funded and their order books are diversified. The market’s refusal to pay a speculative premium reflects pricing discipline at work.

To learn more about how we assess opportunities in Australian small companies, visit our Small Companies Fund page or contact our Distribution Manager, Daniel Liptak, at 0419 004 524 or by email at daniel@datt.com.au.