Quoting a price for a major new scientific instrument is notoriously tricky. Researchers have to estimate costs for equipment that has never been built, forecast expenditures years in advance, allow for unknown contingencies, and win approval from sceptical politicians who always want the project to cost less.

So it is not a complete surprise that a recently finished design review of ITER, a major fusion experiment to be built in Cadarache, France, is forecasting a delay of 1–3 years in its completion date and a roughly 25–30% increase in its €5-billion (US$7.8-billion) construction cost (see page 829).

The seven international partners in ITER (the United States, the European Union, Russia, China, Japan, India and South Korea) will no doubt be displeased by the news. They reached a final agreement to go ahead with ITER in 2006 based on a partially incomplete 2001 design, and may well suspect that the scientists were deliberately quoting an over-optimistic price in order to sell the project.

Whatever truth there might be in that allegation, the fusion community was making its estimate under less than ideal circumstances. ITER had been something of a political football since 1985, when it began life as part of the cold war détente. The collapse of the Soviet Union began a decade of political limbo for the project. Scientists had to radically downsize it at the end of the 1990s to appease the budget concerns of skittish member states.

As international partners came and went (and, in the case of the United States, came again), ITER subsisted on a shoestring. Meanwhile, politicians fought over the project's location. Until that debate was settled in mid-2005, only limited revisions to the design could be done. The redesign has been a top priority for the new ITER team ever since, and the group should be commended for coming forward with a higher estimate of costs after the full review.

ITER may yet follow the path of other projects whose costs spiralled out of control

What is worrying is that even this new price tag might not reflect the true cost of the machine. Crucially, it does not include the soaring price of commodities such as steel and copper, which are used in large quantities in the giant reactor. The ITER team claims that these costs can be excluded because individual member states will contribute finished components rather than raw materials, but this seems disingenuous. Already, the US government has doubled its estimated maximum contribution to the project, and other countries will probably have to follow suit.

This suggests that ITER may yet follow the path of other projects whose costs spiralled out of control once they were given a political imprimatur. The danger to the project itself may seem to be limited because of its international nature, but strictly speaking there is nothing to prevent a cancellation of the sort that ended the US Superconducting Supercollider. Congress halted that experiment 15 years ago, even as the tunnels were being dug in Waxahachie, Texas.

The more likely outcome is that overruns will further undermine the credibility of science at a time when it is increasingly dependent on multinational collaborations to build instruments and data networks. Future projects such as the International Linear Collider, a next-generation particle accelerator for high-energy physics, may well face more sceptical funders if ITER's costs aren't contained.

The independent scientific and management advisory committees overseeing ITER should take a hard look at whether the latest estimates are truly realistic. If they are not, then the committees should demand that the budget include adequate contingencies for factors such as increased energy and commodity costs, as well as scenarios for construction with less than full funding. Even if it means more pain in the short run, this kind of discipline will ultimately lead to a better machine and a better future for all international collaborations.