Australia Is Mandating What US Data Centres Only Negotiate

Australia's national cabinet agreed on 26 August to press ahead with a framework that gives large AI data centres a specific legal duty: put at least as much power into the grid as they draw out, with legislation due in early 2027 according to the official readout of the meeting. Every state and territory leader signed off, including Queensland's David Crisafulli, who had been resisting the framework since the Commonwealth first proposed it in July, according to Bloomberg. In the US, the same problem, who builds the power a data centre needs and who absorbs the risk if it is late or short, is being handled deal by deal, as in the 48-megawatt Texas solar contract ENGIE signed with Blackstone's QTS earlier this month.

The scale behind both approaches is comparable. Data centres are the fastest-growing new source of demand on Australia's grid: the market operator AEMO expects their consumption to rise from about 5 terawatt-hours today, 3% of everything drawn from the National Electricity Market, to 34 terawatt-hours by 2035-36, 13% of the total, in its own 2026 Electricity Statement of Opportunities. Both a legal mandate and a negotiated contract are trying to absorb that growth without leaving the grid short.

The mandate versus the deal

Under the framework the Commonwealth first set out in July, large data centres face a legal obligation to underwrite their own new power supply, pay their full share of grid-connection costs, and be able to reduce their draw when the grid needs it. That duty sits on the operator's own balance sheet, fixed in law before a project is allowed to connect.

The US version runs through contract. In the QTS deal, ENGIE sits between the data centre and the actual generation, sourcing power from a third party's solar project to back its supply commitment to QTS rather than QTS contracting for that generation directly.

Routing the PPA through an intermediary spreads that basis risk across more parties, but it also puts the risk on a company, ENGIE, that did not create the demand behind it. Under Australia's model, the obligation stays with the operator whose data centre is drawing the power.

For anyone paying an electricity bill rather than reading a term sheet, that difference is the part that matters. If an operator can size its own power commitments and let the mismatch fall to whichever counterparty is holding the contract when it comes due, the cost of a shortfall tends to land on the shared grid, and on the households connected to it, rather than on the company that caused the demand. A rule that makes the operator answer for its own new load before it is allowed to connect is a bet that keeping the bill with the party that created it is cheaper for everyone else.

What a legal default buys

Australia's bet is that fixing the obligation in law, before construction starts, forecloses the dispute a contract defers until later: whether the counterparty actually delivers what it promised, at the moment demand and supply diverge. AEMO's own forecast is the case for urgency. Consumption nearly tripling within a decade to reach 13% of the entire market's demand is not a load a case-by-case PPA market has had to absorb before, and a legislated default gives every project the same starting obligation rather than whatever terms it can negotiate.

The mechanism the US does not have

The first caveat is that legislating an obligation does not build the power plants or transmission lines needed to meet it. A net-contributor rule only works if operators can actually secure new generation fast enough, and AEMO itself flags how uncertain that pipeline gets once you look past 2030.

The bigger caveat is structural. Australia can do this because national cabinet lets the Commonwealth and every state and territory leader agree a single set of rules and commit to them together. The US has no equivalent body. Power markets there sit with state regulators and regional grid operators such as ERCOT, and federal jurisdiction stops well short of dictating what a data centre must build before it connects. If Australia's model works, it is not obvious any US jurisdiction has the authority to copy it.

The test will not be the law itself, due in early 2027, but what happens before then: whether operators can secure new generation on the timeline AEMO is forecasting, and whether the PPA deals stacking up on US grid nodes eventually push American regulators toward something like the same rule.