The market will continue to finance artificial intelligence, but it no longer seems willing to accept any price. The proof comes from Australia, where Firmus Technologies withdrew its listing application after unsuccessfully trying to save one of the largest IPOs in the country’s history.
The data center operator, backed by names such as Nvidia, Blackstone and Coatue, was aiming for a valuation close to A$44 billion and raises of around A$7 billion. In a few days the price offered to investors fell from 11 to 8.25 Australian dollars per share. The consultants would later test the market even at $5.50, half the initial value.
It wasn’t enough. Firmus backed out of its debut on the Australian Stock Exchange, attributing the decision to volatility and market conditions. It will now seek money through a new private placement.
The listing fell through, but its financial significance goes beyond a single company: the market has begun to ask how much of the value of AI comes from already operational infrastructures and how much from plants, revenues and profits still to be made.
The maxi IPO of artificial intelligence stops
Firmus builds and manages infrastructures for the processing of artificial intelligence models. The project was of interest to investors because data centers represent one of the main bottlenecks of the technological revolution: they require processors, electricity, cooling systems, land and network connections.
This need is real. It is more difficult to establish how much a company is worth today that promises to satisfy you tomorrow.
According to data circulating during the placement, Firmus had about 42 megawatts of built capacity, compared to a pipeline of close to a gigawatt. Much of the evaluation therefore depended on structures still to be completed, future contracts and particularly ambitious economic objectives.
| What attracted investors | Which increased the risk |
| Growing demand for computing power | Much of the capacity yet to be built |
| Backed by Nvidia and big funds | Rating increased rapidly |
| Agreements with important technological groups | Very high initial investments |
| Expansion into Australia and Southeast Asia | Uncertain times, energy and authorizations |
| Profitable potential billionaires | Business still making a loss |
The problem was not necessarily the industrial project. It was the distance between the project and the price required to participate in it.
Because Nvidia is not a price guarantee
Nvidia’s presence made Firmus instantly recognizable. The chipmaker can directly benefit from the construction of data centers, because many of the new infrastructures use its processors.
Being supported by a world leader, however, does not automatically make any assessment correct. An investor must distinguish between the quality of the partners, the solidity of the project and the price paid for the shares.
The point is particularly important after Nvidia’s record revenues showed how much data center demand weighs on the group’s growth.
The chipmaker cashes in when it sells the hardware; the data center operator must, instead, build the facility, find energy, finance it, and keep it occupied long enough to recoup the investment.
They are two different positions along the same value chain. The success of the supplier does not guarantee that of each of its customers or subsidiaries.
The transition from private markets to the stock exchange
As long as a company remains in the hands of funds and professional investors, its valuation comes from relatively narrow negotiations. An IPO, or the offer with which a company opens its capital to the market, introduces, however, more severe control: new investors must accept that price using real money.
Firmus had rapidly increased its valuation in previous private rounds. The placement was supposed to transform those estimates into a capitalization recognized by the stock exchange. The insufficient demand showed that the two values did not coincide.
The return to private financing, the territory of private equity, does not eliminate capital needs. It simply shifts it to investors willing to bear greater risk, often in exchange for more favorable conditions.
The signal has already reached the shares
The rejection was not confined to Firmus. Maas Group, a listed company that owns around 3.2% of the operator, lost 30% during the session and closed with a drop of more than 22%. In one day, approximately 517 million Australian dollars disappeared from its capitalization.
It demonstrates how a private valuation can quickly transfer to public markets when a listed company owns a stake in the affected company.
Firmus is not present in the ETFs because it never appeared on the stock exchange. His case, however, may change the multiples recognized for the entire supply chain: data centers, energy infrastructures, semiconductors, cloud and companies committed to financing expansion.
What those investing in AI need to check
The IPO withdrawal does not prove that AI is a bubble and does not negate the growth of the sector. It demonstrates that technological innovation and financial convenience are not synonymous.
The principle also applies to those who own ETFs exposed to artificial intelligence and semiconductors. The risk comes not just from the presence of fragile companies, but from the concentration on the big stocks themselves and the expectations embedded in their prices.
As the funding raised by Big Tech to support the race for AI also shows, the digital revolution requires enormous capital. The decisive question is not whether AI will grow, but who will be able to turn those investments into cash flows.
Firmus does not mark the end of the race. It perhaps marks the end of its most lenient phase: the Nvidia logo is no substitute for a balance sheet, a pipeline is not the same as a plant, and a forecast is not yet a profit.
