NOT in my backyard (NIMBY) or in my retirement fund. Australia is giving us a chilling preview of how fast the trillion-dollar artificial intelligence (AI) data centre buildout can unwind.
Public announcements of new projects have gone into overdrive this year. Builders are planning to add around 23 gigawatts of compute capacity, up from just six gigawatts in February, according to the Commonwealth Bank of Australia (CBA).
The country is seen as a prime destination: It’s priced competitively – rental rates in Sydney are half of those in Singapore – and the biggest cities have more space, unlike crowded urban centres like Hong Kong and Tokyo.
But odds are that most of these won’t materialise.
Local communities are increasingly vocal, and successful pushback by one emboldens campaigning by others.
Projects have been halted at an alarming speed. In late September, Goodman Group, the country’s biggest property developer, scrapped a plan to build a 90-megawatt facility in a Sydney suburb.
In early October, Firmus Grid Ltd’s 288-megawatt Bell Bay site in Tasmania was hit with appeals just weeks after receiving local council approval.
Then the most damaging news came last week.
Data centre manager Nextdc Ltd slashed its earnings forecast and placed development plans for one of its major facilities in Melbourne on hold.
Electricity provider Jemena had suspended work to deliver more power to the M3 facility, which already has 60 megawatts in operation, because of local opposition to the installation of 22-meter-high power poles.
Now the company’s best hope is that Jemena will recommence work in February.
Nextdc slashed its earnings forecast over 22-m high power poles.
Only about one-third of planned developments in the pipeline can be delivered by 2030, according to Goldman Sachs Group Inc, which tracks 7.5 gigawatts of projects that are mostly located in Sydney and Melbourne.
Nextdc’s M4 facility, also in Melbourne, for instance, has only a 13% chance of success, according to Goldman.
The 150-megawatt campus, approved by the state government at the start of the year, is facing fierce backlash from residents.
Project withdrawals have occurred at all stages, including more advanced ones when data centre connection requests have been completed.
The median probability of a development progressing from its current status to completion is only 28%, according to the CBA, using data from the Australian Energy Market Operator.
In some ways, these builders resemble biotech firms, risking vast sums of money to develop one product, only to have nothing to show for it in the end.
Only about one-third of planned capacity will be delivered by 2030.
From data centre hotspot to unexpected failures, a drastic shift in Australia’s business conditions perhaps explains why capital markets have reacted so violently.
Firmus had to pull a much-anticipated initial public offering (IPO) in October despite being a venture capital’s darling just two months earlier.
Similarly, Nextdc’s A$1.1bil (US$773mil) convertible bond, issued in September at a meager 1.75% coupon, is now yielding a whopping 12%.
Nextdc’s shares have tumbled more than 40% since the hybrid’s issue.
UniSuper, one of Australia’s top pension funds, decided not to participate in Firmus’ IPO and is steering clear of direct investments in data centres. In hindsight, it’s a good call.
As we’ve seen with Nextdc, using valuation metrics more suited to fast-growing firms, such as comparing a company’s enterprise value to its revenue, or earnings before interest, taxes, depreciation and amortisation, exposes investors to sudden event-driven shocks.
Firmus asked for US$30bil valuation before shelving the IPO.
Rather, it’s clear by now that having permits alone doesn’t necessarily give developers the social licence to operate.
Going forward, we need to look at each project separately – including its location, timing, financing and local political dynamics – before saying with confidence that it will be a cash-generating asset. Are investors prepared to get their hands dirty?
Commenting on Firmus’ IPO, UniSuper’s chief investment officer John Pearce noted that bankers and vendors all got a bit greedy and that the share price would have been “probably closer to US$1 than US$11.”
Indeed, it’s easy for these unicorns to lose their shirt over opposition from a tightly knit community.
Their business model is not only capital intensive, but precarious. — Bloomberg
Shuli Ren is a Bloomberg Opinion columnist covering Asian markets. The views expressed here are the writer’s own.
