To The Moon

The postponement of a major data centre listing by Blackstone Group LP has capped a week in which three Asia-Pacific data centre float plans faltered, exposing a deeper structural contradiction: the AI infrastructure expansion boom was built on 2024's low-rate assumptions, while current financing costs have returned to 2007 levels.

According to media reports, citing "higher interest rates and market volatility that dampened investor demand," Blackstone Group LP's data centre operator AirTrunk has decided to postpone its Singapore real estate investment trust (REIT) listing to the first or second quarter of 2027, depending on market conditions. The deal was originally sized at approximately US$1.5 billion (about S$2 billion).

The impact of this delay extends beyond a single issuer. Last week, Nvidia-backed Australian data centre company Firmus withdrew its ASX listing plan, while Vitro, the data centre unit of Philippine telecom operator PLDT, also pushed back its Philippine REIT IPO to 2027.

Analysts at Goldman Sachs forecast in a recent report a A$50 billion to A$65 billion capital expenditure supercycle for digital infrastructure in Australia and New Zealand, but explicitly noted that financing is the real bottleneck constraining the industry's development.

Three Setbacks in a Single Week

AirTrunk is the third data centre listing in the Asia-Pacific region to be halted or delayed in roughly one week. Firmus withdrew its ASX listing application on October 8 after being forced to cut its A$11 offer price by as much as 25% during bookbuilding. Vitro announced the postponement of its Philippine REIT IPO, citing "current market conditions and rising interest rates." Earlier, SoftBank-backed SB Energy suspended its listing roadshow on September 22. The three cases are tightly clustered in time, reflecting a synchronised cooling of investor sentiment across the region rather than isolated problems with individual projects.

Firmus Was a Valuation Problem; AirTrunk Stumbled on Rates

Although the outcomes were similar, the failure logic of the two deals was entirely different, and distinguishing between them helps to accurately understand the real crux of the current market. Firmus's problem was essentially a valuation issue. The company asked public market investors to subscribe at a valuation of approximately A$43.7 billion (about US$30.4 billion), nearly three times its August private funding valuation of US$10.5 billion, and cited unconventional metrics such as "EV+1/EBIT+2" to justify its pricing. Australian local fund managers were broadly sceptical, with Jun Bei Liu of Ten Cap publicly stating, "My view is that [the IPO valuation] should be below A$10 billion."

AirTrunk's situation is entirely different. It is a hyperscale operating platform spanning Australia, Japan, Singapore, Malaysia, Hong Kong and India, acquired in 2024 by Blackstone Group LP and CPP Investments for A$24 billion, the largest data centre transaction on record globally at the time. Its REIT was positioned as a "steady path" to invest in AI infrastructure — assets already in stable operation, tenants being large hyperscale cloud computing companies, and generous dividend returns.

Yet AirTrunk stumbled on rates. According to Bloomberg, before the offering AirTrunk had considered raising the yield from an original 5.5% to about 6% to attract investors, but still failed to complete the deal. On Tuesday, the US 10-year Treasury yield closed at 5.22%. At a "optimised" 6% yield, the AirTrunk REIT would offer only about 78 basis points of excess return — compensation investors deemed far from sufficient for a single-sector illiquid instrument carrying currency risk, tenant concentration risk and construction risk.

AirTrunk had also planned to arrange approximately S$2 billion in loans for asset acquisitions and debt refinancing, which Bloomberg described as "a key test for Asian data centre financing, as local banks have approached their sector exposure limits." That test also ended in failure.

The Market Demands Higher Risk Compensation

Although the two deals operated through different mechanisms, they sent the same signal to the market: investors are demanding more attractive pricing to compensate for the risks they bear. For Firmus, that means a lower valuation; for AirTrunk, that means a higher yield — and mathematically, a higher yield is also equivalent to a downward shift in the valuation of the assets held by Blackstone Group LP.

The bond market is sending the same warning. Credit spreads on data centre debt have continued to widen, with Coreweave's unsecured bond spread briefly exceeding 1,000 basis points for the first time, and Oracle's unsecured bonds trading at deeply junk-level prices. Broadcom CEO Hock Tan warned on October 13 at a conference hosted by DBS Bank in Singapore that financing could replace component shortages as the industry's main bottleneck. JPMorgan technology analyst Joshua Meyers characterised the remarks as the "most direct cause" of the tech sector's decline that day.

Demand Exists; Financing Is the Bottleneck

Market setbacks do not mean underlying demand has disappeared. According to Goldman Sachs' sales team's summary of its latest research report on Asia-Australia digital infrastructure, as AI adoption accelerates, Australia and New Zealand will face a 0.7 gigawatt supply gap by 2030, triggering a A$50 billion to A$65 billion capital expenditure supercycle. However, in maintaining a neutral rating on listed company NEXTDC (NXT AU), Goldman Sachs offered a telling rationale — financing itself is the constraint. Goldman Sachs analysts warned that NEXTDC faces near-term margin dilution pressure during its transition to hyperscale cloud customers, and has a funding gap of more than A$1 billion.

The contradiction points to the core: tens of billions of Australian dollars in capital expenditure require funding sources, and the two most obvious public market financing channels — ASX growth IPOs and SGX income REITs — clogged up in the same week. The alternative path leads to private markets, including private credit, insurance capital and special purpose vehicles, at the cost of higher financing costs. Goldman Sachs economists stated bluntly in this week's "Global Economic Analysis" report: "Global interest rates have risen to their highest level since 2007." Income products priced under 2024 rate assumptions will always face an inherent disadvantage in the 2026 rate environment.

Structural Mismatch in the Cost of Capital for AI Infrastructure

AirTrunk's postponement is more cautionary than Firmus's. Firmus can be attributed to a single issuer's self-inflated valuation; AirTrunk represents the most credible version of this trade logic — real assets, real tenants, Blackstone backing, and a proactive yield increase. If such a project still cannot complete its offering at 6% with the 10-year Treasury yield at 5.2%, the problem is no longer an isolated misstep by an aggressive issuer, but a structural mismatch in the cost of capital for the entire AI construction wave. This expansion boom started with 2024 rate assumptions as its premise and is now seeking refinancing at 2007-level costs. The difference in between will ultimately be borne by one party. For now, the public market's attitude is clear — that party will not be them.

Internal discussions at AirTrunk regarding listing size and timing details are still ongoing. The next notable milestone is Firmus's partner in Batam, Indonesia, DayOne, which is reportedly planning to seek up to US$5 billion in financing in the US market as early as November.