There is a principle that most centre-right economists would regard as elementary: those who create a cost should bear it. Applied to the electricity grid, that principle has been spectacularly violated by the artificial intelligence industry over the past several years. The United States, under President Donald Trump, has now moved to reassert it.
During his State of the Union address on Tuesday, Trump announced that his administration had negotiated a "ratepayer protection pledge" with tech companies, under which companies would have "the obligation to provide for their own power needs" and build on-site power generation for new data centres. The formal signing is scheduled for 4 March at the White House. Companies participating in the event include Amazon, Google, Meta, Microsoft, xAI, Oracle, and OpenAI.
The political logic behind the move is straightforward. In 2025, regulators across the United States approved $11.6 billion in electricity rate increases to accommodate AI's growing power demands, increases that have since gone into effect and generated a backlash from lawmakers and local communities alike. Trump framed it bluntly from the floor of Congress: "I'm telling them, they can build their own plant. They're going to produce their own electricity. It will ensure the company's ability to get electricity, while at the same time, lowering prices of electricity for you."
The scale of the problem is not in dispute. According to estimates from the Lawrence Berkeley National Laboratory, power demand from US data centres doubled between 2018 and 2024 and could triple by 2028. An ageing national grid, built decades ago, was never designed to handle electricity demand rising at this pace. The argument that the cost of upgrading it should fall on ordinary households rather than the corporations generating the demand has always been difficult to defend on any coherent fiscal or ethical basis.
For its part, the technology industry has been quietly moving in this direction already. In January, Microsoft said it would pay utility rates high enough to fully cover its data centre energy costs, while OpenAI committed to "paying our own way on energy" so that its operations do not increase electricity prices. Google says it already covers its own data centre energy costs, and Meta has similarly said it pays the full cost of the electricity used by its data centres. The pledge, in other words, formalises what the industry's largest players were already claiming to do voluntarily, which raises the question of what, precisely, changes on 5 March.
Where the Sceptics Have a Point
Energy experts have offered a more sobering read of the initiative. Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School's Environmental and Energy Law Program, has argued that the White House is "putting this pledge on the wrong entities," because the details of how energy costs are distributed among ratepayers are determined by utilities and state regulators, not by tech giants. That is a structurally important observation. Even if every hyperscaler builds its own power plant tomorrow, the costs of upgrading poles, wires, substations, and transmission infrastructure would remain. As grid expert Brandon Owens put it: "Most of today's cost pressure is coming from transmission, distribution, and system readiness, not energy supply. Those costs remain even if a data centre self-supplies generation."
Trump's ratepayer protection pledge lands in the middle of a growing backlash against data centre construction, with governments at the state and local level introducing measures to offset their costs to communities, and some cities enacting moratoria on new construction. At least six US states have proposed bills that would temporarily ban all data centre construction. That kind of legislative pressure reflects genuine community frustration, and no voluntary pledge architecture resolves it if enforcement mechanisms are absent.
Several Democratic lawmakers quickly said Trump's initiative does not do enough to shield consumers, a position shared by some clean-energy groups. Aliya Haq, president of the nonprofit Clean Economy Project, said pledges are "Band-aids that avoid fixing the real problem" and that serious reforms to electricity markets and grid permitting are overdue. That critique deserves engagement rather than dismissal, because it identifies a genuine gap between the political appeal of a pledge and the regulatory architecture needed to make it stick.
The Australian Dimension
For Australian readers, this is not a distant American policy curiosity. The same structural tension is building here. Australian data centres consumed an estimated 3.9 terawatt-hours of electricity in FY2025, representing approximately 2 per cent of National Electricity Market grid-supplied consumption. That figure is set to grow sharply. The Australian Energy Market Operator's Step Change scenario projects data centre consumption reaching 12 terawatt-hours by 2029, equal to today's entire Tasmanian load.
Data centre investment in Australia could represent between 8 and 11 per cent of the nation's projected electricity consumption by 2035, with up to $135 billion in associated investment. Australia is a strong contender in the Asia-Pacific region, offering significant renewable energy generation potential, land availability, a favourable geopolitical position, and robust data sovereignty frameworks. But those advantages come with a caveat: if the cost of powering that buildout is quietly socialised onto household power bills, the economic case for hosting hyperscale facilities will need serious re-examination by policymakers and regulators alike.
The Australian Energy Market Commission has already begun grappling with this in its draft rules for large energy users, proposing that very large customers meet system-security standards before energisation. The debate in Canberra is behind where Washington now finds itself, but the underlying numbers are pointing in the same direction.
A Pragmatic Assessment
The ratepayer protection pledge reflects a sound intuition: corporations that impose grid costs on communities should internalise those costs. The principle of fiscal responsibility demands nothing less. But a principle without enforcement is a press release. Trump did not name the specific companies that signed the pledge during his address, nor did he spell out how it would be enforced. Those are not minor administrative details; they are the substance of whether the policy functions or merely signals.
The honest centre-ground position is this: the US has identified the right problem and applied political pressure in a direction that is broadly correct, while leaving the hardest regulatory questions unanswered. Tech companies should bear the costs of their own energy footprint, including transmission and grid infrastructure, not merely the supply portion. Voluntary pledges are a starting point, not an end state. What follows the White House signing ceremony will matter far more than the ceremony itself. Australia's energy regulators and policymakers would do well to watch that process closely, and to begin asking the same hard questions of the hyperscalers now arriving on our own grid.