Asian AI Data Center IPO Wave Hits Headwinds as Multiple Companies Face Public Market Valuation Tests
nashnova research
Several Asian AI data-center companies have seen IPOs blocked or scaled back while listed peers suffer sharp sell-offs — public markets are losing patience with sky-high valuations, construction delays, and fragile contracts.
Whose IPOs got blocked?
Australian developer Firmus pulled its IPO — planned to raise up to A$7.9 billion (~US$5.5 billion) — after failing to attract enough investor demand. The company, which started in Bitcoin mining, sought a ~US$30 billion valuation despite much of its capacity remaining unbuilt. It is now turning to private funding.
SoftBank-backed DayOne, which operates data centers in Southeast Asia and Japan, is considering halving its planned US IPO raise from up to US$5 billion or delaying the listing entirely. The company had been targeting a ~US$20 billion valuation.
Another Australian operator, AirTrunk, has pushed its planned ~US$1.5 billion Singapore REIT — a real-estate investment trust that packages data-center properties into tradable fund units — to next year.
SoftBank's US arm SB Energy has also delayed its Nasdaq IPO over valuation disputes; reports suggest the listing may slip to December or later.
Are listed players faring any better?
Australian data-center developer NextDC plunged nearly 42% in a single day, wiping out ~A$3 billion in market value. This means → the public market delivered the bluntest possible message: plans alone are not enough — delivery is what counts.
The trigger was specific: a power-expansion project at a large Melbourne data center was blocked by local residents, deferring ~A$140 million in contracted revenue. The company also froze new developments and land acquisitions.
South Korean AI-infrastructure provider Elice dropped over 30% on its Kosdaq debut, then bounced nearly 30% the next day — the wild swings themselves show the market has no consensus on how to price these companies.
Where does the money go, and where does it get stuck?
Data centers are a classic "spend big first, collect slowly later" business: massive upfront capital, heavy power infrastructure, and community approval — with revenue often years away.
In plain terms = you spend billions building server halls and power grids, negotiate with neighbors, and wait for clients to start paying only after construction is done. Any single link in that chain breaking cuts off cash flow.
Doyle Securities CEO John Athanasiou put it plainly: "The market is going through a reality check, and investors are realizing the risks that come with building data centers."
What does the "house of cards" risk really mean?
Tamim Asset Management's Ron Shamgar flagged the highest risk at "neocloud" operators — companies that lease AI infrastructure and rely on debt financing. If construction delays prevent on-time delivery, client contracts can be terminated.
His exact words: "It's a bit like a house of cards that, as delays happen, could start to tumble." This means → the business model stacks like playing cards: borrow to build → lock clients with contracts → collect rent after delivery. Pull one card out and the rest collapse.
Yet Shamgar also sees long-term demand for advanced chips and compute power holding firm. He compared this pullback to the dot-com era: "These large drawdowns are a natural part of this mega-theme that will play out over about a decade."
What is the public market actually testing?
This reflects a core shift: public-market pricing for AI data centers is decoupling from private-market valuation logic. Private investors can bet on a long-term story; public markets want to see what you can deliver today.
In plain terms = the private market acts like venture capital — it prices vision. The public market acts like a bank — it prices cash flow. The same company can get valuations an order of magnitude apart.
Whether construction delays and community resistance can be effectively managed will determine if this sector regains market confidence — this is not a demand problem but a delivery problem.
市场有风险，内容仅供研究参考，不构成投资建议。