The metric that will decide the next collider is a ratio, not a total

Fifty-four per cent of CERN's 2024 grid draw went into a single machine ^[1]. Inside the technical estate that takes about 95% of the lab's power ^[1], the LHC is 57% on its own ^[6]. Smooth the whole load across the 8760 hours in a year and it annualises to roughly 147 MW ^[3]; everything outside the accelerators, detectors and computing, offices included, is about 65 GWh ^[2]. That remainder is the only part a director can economise on without touching physics, and it is a twentieth of the bill.

So the number that gets quoted is a ratio. A doubling of data per joule is arithmetically the same statement as a halving of joules per unit of data ^[5], which is a real efficiency gain and a genuinely useful thing to have measured. It is also a quotient, and a quotient rises whenever the numerator grows faster than the denominator, including when the denominator is growing too ^[7]. Nothing in a data-per-joule figure caps the megawatt-hours.

The June guidelines are more candid than the headline metric. Their 110 recommendations reach into tunnelling, shielding, component design, procurement and waste management ^[11], and they cover decommissioning alongside planning, construction and operation ^[10]. None of that appears in joules per unit of data. A 91 km ring would be about 3.4 times the LHC's circumference ^[4], and the excavation and concrete happen once, before there is any data to divide by. The comparison the piece uses for annual consumption, a mid-sized European city ^[6], is the one a grid operator would recognise; the ratio is the one a physicist would.

Hannah Wakeling of the John Adams Institute at Oxford, who led the report, says colleagues across the community are "eager to incorporate environmental sustainability into their work and to share lessons learned" ^[13]. Her own work covers carbon footprint, raw materials and biodiversity ^[12]. The document is equally plain about why little of this happens: labs lack funding, staff and time, generic sustainability advice does not transfer to accelerator technology, and there is no settled method for folding it into daily operations ^[15]. It is a living document offering advice, which means uptake is each facility's choice.

Physics World's account also ties the urgency to energy prices, citing oil and gas costs rising after what it describes as the effective closure of the Strait of Hormuz ^[16]. A buyer with that annual average load is a price taker, and the electricity line is the one that moves with events no lab controls.

The High Luminosity LHC is due online by 2030, the 91 km Future Circular Collider is under consideration, and China is looking at 100 km ^[8]. When those proposals reach funding agencies, data per joule will be in them, because it is the metric that has moved since 2020 ^[5]. What arrives on a host country's grid is the total.