Powered by
MOMENTUM
MEDIA
Australia’s data-centre boom requires long-term commitments in a rapidly changing market. Understanding the economics behind those commitments is critical to assessing value and allocating risk, write Christine Polek and Theresa Bischof.
Not long ago, all the computing power in the office of a small economic consulting firm could be kept in a storage closet. Every so often, someone would pop their head in to check the temperature and reposition the fans. The hum and whir were the background noise as we ran regressions, reviewed case materials and streamed music from the internet.
Today, that storage closet has become something closer to a small city.
Whether you deploy a swarm of AI agents at work or ask a chatbot to plan a family holiday, these interactions depend on infrastructure that is already enormous, with data-centre spending forecast to reach US$1.1 trillion a year by the end of the decade.
To build a data centre, a developer must coordinate land, planning approvals, grid connections, long-lead electrical equipment, construction, financing and customer commitments. Each input operates on its own clock, under its own regulatory, commercial and technological constraints.
The infrastructure also has a geography. Similar-looking data centres can face radically different economics and timelines depending on local grid design, planning regimes, supply chains and trade policy. Moreover, AI capabilities, customer requirements and the economics of computing can shift before a single planning process is complete.
The result is a basic economic tension: capital must be committed years in advance to serve a market whose technology and demand can change in months.
Australia is a case in point. Still a small market by US or Chinese standards, it is attracting extraordinary interest. As of June 2026, prospective projects had generated 66.9 gigawatts of grid-connection requests across the National Electricity Market. That compares with about 2.2 gigawatts of existing rated grid-connection capacity.
Australia appeals as a regional digital hub, with subsea cables to Asia, strong renewable energy potential, and large sites. But attracting projects is easier than getting them up and running. AEMO is targeting a roughly two-year process for large data centres to move from connection application to energisation. Oxford Economics, in modelling for AEMO, found that more than one-third of connection requests in AEMO’s 2025 database were withdrawn over the preceding 12 months.
The challenge is also financial: securing capital today depends on convincing investors of what future infrastructure can earn. In early October, Firmus, which develops data centres and provides computing capacity for AI workloads, withdrew its planned ASX listing. It had sought to raise $7.1 billion at an implied valuation of around $44 billion, which would have been Australia’s largest IPO in nearly three decades. The company cited market volatility. Some investors reportedly wanted more detail in the prospectus; others doubted that AI infrastructure spending would pay off soon.
The withdrawal doesn’t mean demand for AI computing is disappearing. But the debate surrounding its valuation highlights the difficulty of pricing capacity and earnings that may take years, and substantial additional capital, to deliver.
If funding falls through, power arrives late, customers move elsewhere, or new technology changes the economics of computing, a project’s value may change. Assessing any resulting loss requires more than adding up foregone revenues.
It requires assessing what would likely have happened but-for the disruption, taking into account changes in technology, prices and market demand. Even without a delay, a facility might have faced lower prices, different customer requirements or higher costs. Would the customer have stayed? Was demand lost or merely deferred? Could the capacity or equipment have been redeployed, and what costs were avoided?
These economic questions matter long before a dispute arises. Long-term customer commitments can help secure financing, but reduce flexibility. Rights to exit or change a project can preserve flexibility, but leave another party bearing costs it cannot recover.
The agreements underpinning a project may also allocate risk differently. A construction contract might allow extra time for a late grid connection, while a customer agreement still requires service to begin on the original date. The resulting exposure depends on both the contractual terms and the commercial consequences: whether customers can wait, capacity can be sourced elsewhere, or equipment can be redeployed.
The storage-closet economy is gone. What has replaced it demands vastly more capital, depends on many more moving parts and carries far greater consequences when expectations and reality diverge. For lawyers advising on these commitments, understanding how value changes when assumptions fail is central to understanding the risks their clients are taking.
Christine Polek is a senior principal and economist, and Thresa Bischof is a consultant and engineer, at Keystone Strategy.
