At a glance
CDC Data Centres chief strategy officer Dr Jack Dan told the CEDA State of the Nation conference in Canberra on 25 June 2026 that the boom is attracting “fly-by-night, cowboy-type” speculators, and called for stringent upfront rules to “separate what is real investment” from quick-money plays.
The congestion he described is real and measured. Network providers passed roughly 44GW of data centre connection requests to AEMO, but Oxford Economics, commissioned by AEMO, found only about 6GW is needed under the Step Change scenario, leaving roughly six of every seven megawatts as “phantom demand.”
The wider claim is looser. Dan named no operator and gave no example of one behaving as a “cowboy,” and land speculation is not the same activity as building data centres quickly.
The warning lands while the Australian Government’s data centre inquiry is open to submissions and the AEMC is finalising new connection rules, the moment when the entry bar for everyone in the queue is being set.
Tighter vetting is sound policy for a clogged queue, and it also favours operators with capital, track record and certification, which is why a warning from an incumbent and a fix that raises the bar point the same way.
A warning aimed at the rule-writers
Australia’s data centre pipeline is large and, in its committed core, credible. Westpac estimates the investment pipeline will exceed A$155 billion, a scale economists have begun comparing to the mining investment boom, and data centre spending has already lifted business investment to its fastest quarterly growth in a decade. The financed, contracted projects behind those numbers are real.
What CDC put on the table at CEDA was the rest of the queue, where lodged applications outnumber the projects that will ever break ground. Speaking at the State of the Nation conference in Canberra on 25 June, CDC Data Centres chief strategy officer Dr Jack Dan said the speed of demand has pulled in applicants with no capital, no underwriting and no customers, who can still lodge connection and planning requests that public agencies are obliged to process.
“There is this narrative that data centres will grow at an almost exponential rate and, with any industry where you have significant growth, that industry also attracts a significant amount of speculators, a significant amount of fly-by-night, cowboy-type behaviour,” Dan said. “One of the things we have to ensure is that those actors don’t ruin it for everyone.”
State of the Nation is a policymaker forum, and the warning arrived while three rule-writing processes are live: the Australian Government’s data centre inquiry, open to submissions until September; the AEMC’s new connection standards for large loads, due to be finalised in the second half of 2026; and the New South Wales consultation, expected to report over the same period. A call for “stringent and clear requirements upfront,” delivered to the people drafting those requirements, is a submission made out loud. It bundles two claims that are worth pulling apart, because they do not carry the same weight.
Phantom demand is real
The first claim is the strong one. Network service providers passed roughly 44GW of data centre connection requests to AEMO for its 2025 planning inputs. Independent forecasting by Oxford Economics, commissioned by AEMO, screened those requests and found only about 6GW is required to meet demand under the central Step Change scenario. On that screening, roughly six of every seven megawatts requested is “phantom demand” that is not expected to reach the grid. Separately, around 5.4GW now sits formally in the transmission connection queue across eleven projects, split close to 60 per cent in New South Wales and 40 per cent in Victoria, as we set out in our analysis of AEMO’s 5.4GW connection-queue disclosure.
Oxford Economics found that removing requests unlikely to proceed explains almost 90 per cent of the gap between total connection requests and the load expected at maturity. That is the quantitative version of the congestion Dan described: the bulk of what is lodged reflects firms competing to capture future demand rather than a reliable guide to what will be built.
Australia’s data centre demand | Capacity | What it represents |
Connection requests to AEMO (2025 inputs) | ~44GW | All data centre interest passed up by network providers, unfiltered |
Required under Step Change (Oxford Economics screening) | ~6GW | Capacity the central scenario expects to be built |
Formally in the transmission queue (Q1 2026) | 5.4GW | Eleven projects progressing through connection, ~60% NSW / 40% VIC |
Source:Oxford Economics phantom-demand screening commissioned by AEMO (2025 IASR inputs); AEMO Quarterly Energy Dynamics Q1 2026.
The bottleneck this creates sits upstream of any contest between operators. Phantom applications consume the assessment capacity of network planners, water utilities and councils, each of which must treat a request as genuine until it proves otherwise. On that much, the case is sound, and it is a problem the renewable-generation queue has carried for years.
Cowboy operators
The second claim outruns the evidence. Dan's own example is land speculation, and that much is well documented: a landowner sees a power line at the back of a property, lodges a data centre plan and a connection request, then waits to on-sell the site at a higher price. Extending that to 'fly-by-night, cowboy-type' operators who could 'ruin it for everyone' is a far broader charge, and Dan gave no named example to support it.
The two groups are not the same. A speculator lodging a phantom request to flip land is not a builder. A fast-moving, lightly capitalised operator that connects quickly is not a speculator. That distinction carries real weight here, because the cure on the table, tighter vetting at the point of application, is aimed at the queue but lands on everyone applying to it. A filter built to catch land-bankers can also raise the bar on legitimate new entrants whose edge is speed rather than balance-sheet depth. Speed and low capital intensity are not the same thing as bad faith, and a rule that treats them as proxies for it would screen out builders the market needs.
Who a vetting gate favours
The consequential effect of the “cowboy” framing is not on speculators, who fail at the first gate regardless. It is on where the entry bar sits for everyone already inside the queue. CDC is the largest operator making the case: an independent valuation put the company at A$15 billion in March, and a 555MW contract signed in May, the largest data centre deal in Australian history, took its contracted capacity past 1GW. Higher, clearer upfront requirements are good policy for a congested queue, and they also favour operators with capital, track record and certification over those without. Both are true at once, which is the honest way to read an incumbent calling for tougher rules: the public interest in clearing the queue and the commercial interest in raising the drawbridge point in the same direction.
Australia has the beginnings of that filter already. The Australian Government published five national expectations for data centre developers in March 2026, covering energy, water, security, community and economic contribution, and is taking submissions to its data centre inquiry until September. We examined whether those expectations would attract investment or deter it when they landed. At state level, New South Wales has tried to convert intent into throughput, moving from A$136 billion in proposals to twelve-month approvals through its Investment Delivery Authority, a process whose value depends on whether the proposals entering it are genuine.
The energy counterweight
The CEDA panel was not uniformly cautionary. AGL chief executive Damien Nicks told the forum that data centres, run on the right terms, are an asset to the energy system rather than a drain on it. “Data centres are energy intensive but managed in the right way, with the right settings, they can help underwrite and support new renewable investment, improve the system utilisation and help lower costs over time,” he said, adding that the sites of retiring power stations could host them and keep regional jobs.
A data centre is net-positive for the grid when it brings its own firmed supply, sits where the network can carry it and can flex when the system is tight. The same load is a problem when it queues speculatively, assumes supply it has not secured and offers no flexibility. The filter on speculative requests and the upside Nicks describes are the same mechanism from two ends: keep the unbacked load out, and genuine load can be matched to new generation rather than competing with households for existing supply. A Climate Council report counts more than 160 data centres already built in Australia and around 90 more in planning. The opposition that helped produce these inquiries centres on power, water and bills, which we have tracked across the country in our coverage of data centres and social licence, and that is a separate question from who in the queue is genuine. Both are in play at the same forum.
What to watch
The Australian Government’s data centre inquiry closes to submissions in September 2026, and how it treats applicant vetting and upfront conditions will show whether the reform clears the queue or narrows the field. The Australian Energy Market Commission has separately proposed new connection standards for large data centre loads, including ride-through obligations, with a final rule expected in the second half of 2026. Together with AEMO’s next connection-queue update, those are the instruments that will decide where the line between real and speculative is drawn, who clears it, and how quickly the phantom megawatts make way for the financed pipeline.