AirTrunk just put its US$1.5 billion listing on hold. In a volatile data centre market, qualification is no longer optional.

AirTrunk has reportedly delayed its US$1.5 billion Singapore REIT listing to 2027, weeks after Firmus scrapped its IPO and Goodman withdrew its A$1.2 billion Project Mars.
Thailand paused 166 data centre projects in a single decision on 4 September, while Australia plans to legislate mandatory national data centre standards by the end of 2026.
Around a third of new data centre projects proposed in the NEM were cancelled in the past year, according to AEMO.

An eight-question test any organisation should apply before committing people, capital or capacity to a development-stage data centre project:

Is the power secured?
Is the land controlled and the community on side?
Will the design survive the rules still being written?
Is there a tenant?
Is the next phase funded?
Is the client’s strategy stable?
Do the terms pay for early-stage risk?
Can your firm absorb it stopping?

AirTrunk is delaying its planned Singapore REIT listing until next year, according to people familiar with the matter quoted by Bloomberg, Reuters and The Wall Street Journal on Tuesday. The Blackstone-backed operator had intended to start formally testing investor interest in the US$1.5 billion offering at the start of October. Higher interest rates and market volatility kept investors cautious.

It is the latest in a run of signals that the data centre market has become volatile. In the past month, Goodman withdrew Project Mars after conceding it hadn’t won a social licence. Firmus lost its partnership with CDC and scrapped its IPO. Thailand froze 166 projects while it writes new rules, and Indonesia halted a 640 MW campus weeks after ground was broken.

None of this means the boom is over. It means the pipeline is far less reliable than its headline numbers suggest. For any organisation that wants a share of the work while projects are still in development, qualifying projects properly is no longer good practice. It’s a necessary capability.

The investment trend is still real

The long-run numbers remain enormous. PwC puts global data centre spending at US$31.6 trillion through 2050. Asia Pacific may need to mobilise around US$772 billion over the next five years for data centres, computing and networks. Westpac values Australia’s pipeline at about A$175 billion, with roughly A$110 billion likely to be invested over the next three years.

The demand story is intact. What has changed is how reliably any single project turns into work. Four pressures are driving that volatility, and all of them are building at once.

Investor sentiment is turning

AirTrunk is not a company in trouble. Blackstone and CPPIB bought it for a record A$24 billion in 2024, and it remains the region’s largest data centre operator. It had even considered lifting the yield on offer from 5.5% to around 6% to support demand. If the best-capitalised developer in Asia Pacific is waiting for better conditions, every developer’s funding assumptions deserve a harder look.

It isn’t alone. PLDT’s data centre arm deferred its own REIT listing to 2027 last week, citing rising interest rates. Singapore’s main REIT index has fallen 12% since late July. And as I’ve written previously, Firmus scrapped its planned IPO shortly after CDC Data Centres dropped it from their joint Australian AI infrastructure project, reportedly over Firmus’s South East Asian expansion.

REIT listings, IPOs and asset sales are how developers recycle capital into the next build. When that window narrows, the pressure doesn’t land on operating facilities. It lands on projects that haven’t reached financial close, which is exactly where consultants, advisers, early contractors and suppliers carry most of their exposure. Firmus also shows that a client’s strategy and partnerships can change mid-project, and the scope of everyone working on it changes with them.

Regulators are rewriting the rules mid-pipeline

In Australia, the Senate opened an inquiry into AI data centres in May. Labor plans to legislate national data centre standards by the end of 2026, including a requirement for new facilities to fully offset their electricity use with additional renewable generation, despite opposition from Queensland and the Northern Territory. The AEMC has proposed new technical rules for large data centre loads, and Transgrid is steering developers towards regional NSW.

South East Asia, often seen as the growth market for firms watching Australian projects stall, is moving faster. On 4 September, Thailand’s Data Centre Business Policy Committee paused 166 projects pending new national siting and environmental standards. Forty-nine were already under construction. Johor asked operators to suspend expansion of water-cooling systems for 18 months, and Malaysia’s Prime Minister told parliament in February that applications for new non-AI data centres would stop. In Indonesia, BDx broke ground on a 640 MW AI campus in West Java on 22 September, and was ordered to halt within weeks over missing approvals.

Every one of these changes can turn a design in progress into a redesign. Some stop the project altogether.

Social licence has become a project-ending risk

Goodman’s Project Mars would have sat around 20 metres from homes in Lane Cove. Goodman conceded at the Senate inquiry that it had not won a social licence, and withdrew the application on 28 September. In Perth, GreenSquare withdrew applications at two sites after community and council backlash. In Tasmania, Firmus’s Bell Bay project has faced opposition over when its diesel generators can run.

The pattern isn’t limited to Australia. Bangkok authorities paused approvals on three more projects after complaints about a data centre being built next to a hospital. Data Center Watch counted at least 75 projects worth roughly US$130 billion blocked or delayed by local opposition in the first three months of this year.

The combined effect shows up in the pipeline. AEMO estimates around a third of new data centre projects proposed in the NEM were cancelled over the past year. In NSW, Transgrid is tracking more than 20 GW of enquiries against around 1.5 GW of signed connection agreements. That’s not a rounding error.

Why qualification is now a necessary capability

Investors, regulators and communities are each now capable of stopping a project, and none of them moves on the client’s timetable. The risk sits earliest in the project life, where the organisations supporting development do most of their work and carry most of their unrecovered cost.

Chasing the data centre trend isn’t a strategy on its own. The firms that do well from it will be the ones that can tell, early and with evidence, which projects are likely to be built.

Who carries development-stage risk

This isn’t only a problem for engineers. Any organisation that spends money, people or capacity before a project reaches financial close carries the same exposure. That includes:

Engineering and design consultants across civil, structural, electrical, mechanical, HV and grid connection.
Planning and environmental consultants, including the acoustic, traffic, air quality, water, heritage and ecology specialists who prepare DA studies.
Community and stakeholder engagement advisers building the social licence case.
Legal, property and financial advisers working on land, approvals, connection agreements, leases and funding.
Cost consultants, quantity surveyors and project managers engaged before the build is committed.
Contractors in early contractor involvement or paid pre-construction services.
Equipment suppliers and manufacturers taking early orders or reservations for long-lead items such as transformers, switchgear, generators and cooling plant.
Labour hire and specialist recruitment firms building teams ahead of demand.

Each of them is paid, if at all, from a project that may never be built. Each has a reason to qualify that project before committing.

Eight questions to answer before you commit

Each question below maps to one of those pressures and asks for evidence, not assurance. The test is what the client can show you, not what they expect to happen.

#	Question	Green: evidence in hand	Red flag
1	Is the power secured?	Signed connection agreement with capacity and date	An enquiry or application only, in a constrained part of the network
2	Is the land controlled and the community on side?	Land owned or under option, early engagement done, buffer to homes, schools and hospitals	Option pending, DA is the community’s first notice, site near sensitive receivers
3	Will the design survive the rules still being written?	Design already tested against draft national standards, AEMC load rules, local water and siting policy	No view on mandatory renewables, ride-through or water limits, or a jurisdiction with an active pause
4	Is there a tenant?	Signed hyperscaler or anchor lease	Speculative capacity, or tenant talks at heads of agreement
5	Is the next phase funded?	Equity committed or debt arranged for construction	Funding depends on a listing, sale or raise that hasn’t happened
6	Is the client’s strategy stable?	Established operator, unchanged partners, consistent plan	Recent partner exit, pulled IPO, fast expansion into new markets
7	Do the terms pay for early-stage risk?	Paid gates, suspension and termination payments, change-in-law relief	Discounted or at-risk early work, broad indemnities, uncapped liability
8	Can your firm absorb it stopping?	Project within concentration limits, people and capacity redeployable to other sectors such as transmission and renewables	One client or segment dominates revenue, specialists hired for this job alone

Questions 1 to 4 test whether the project can be built. Questions 5 and 6 test whether the client can build it. Questions 7 and 8 test whether your firm survives if neither holds.

Turning the answers into a decision

Rate each question green (evidence in hand), amber (evidence pending, with a date) or red (no evidence). Then let the score set the terms of engagement, not just the yes or no.

Mostly green, no reds. Pursue on normal terms, within your concentration limits.
Reds on 1 to 4 only. The project may not be buildable yet. Offer paid, limited-scope early work such as a feasibility or go/no-go assessment, rather than full scope. Don’t discount early work to buy position.
Reds on 5 or 6. The project may be buildable, but the client may not get there. Engage only with paid gates, suspension and termination payments, and payment security such as a parent guarantee.
Reds on 7 or 8. That’s your firm’s risk, not the project’s. Fix the terms or the resourcing plan before committing, or decline.
Three or more reds anywhere. Decline, or limit involvement to paid advisory work.

Revisit the scorecard at every gate. A project can move from green to red in a fortnight, as funding markets have just shown.

Implication for Board and senior management decisions

The global data centre investment trend is real, and it will create a great deal of work for the organisations that support it. The volatility is real too. Nobody has a reliable view of which of today’s projects will be built, and a checklist won’t provide one. What it does is make the firm’s exposure visible before people are hired and hours are spent.

So the question worth asking isn’t whether your firm should chase data centre work. Most firms with the right capability clearly should. It’s this: how many of the projects in your current data centre pipeline would pass these eight questions today, and are your contracts and resourcing priced for the ones that wouldn’t?

#DataCentres #ProjectDevelopment #RiskManagement #Infrastructure #SoutheastAsia