It could be a new world record, although no one involved wants to talk about it. In the south of France, a collaboration among 35 countries has been birthing one of the largest and most ambitious scientific experiments ever conceived: the giant fusion power machine known as the International Thermonuclear Experimental Reactor (ITER). But the only record ITER seems certain to set doesn’t involve “burning” plasma at temperatures 10 times higher than that of the sun’s core, keeping this “artificial star” ablaze and generating net energy for seconds at a time or any of fusion energy’s other spectacular and myriad prerequisites. Instead ITER is on the verge of a record-setting disaster as accumulated schedule slips and budget overruns threaten to make it the most delayed—and most cost-inflated—science project in history.

ITER is supposed to help humanity achieve the dream of a world powered not by fossil fuels but by fusion energy, the same process that makes the stars shine. Conceived in the mid-1980s, the machine, when completed, will essentially be a giant, high-tech, doughnut-shaped vessel—known as a tokamak—that will contain hydrogen raised to such high temperatures that it will become ionized, forming a plasma rather than a gas. Powerful magnetic and electric fields flowing from and through the tokamak will girdle and heat the plasma cloud so that the atoms inside will collide and fuse together, releasing immense amounts of energy. But this feat is easier said than done. Since the 1950s fusion machines have grown bigger and more powerful, but none has ever gotten anywhere near what would be needed to put this panacea energy source on the electric grid. ITER is the biggest, most powerful fusion device ever devised, and its designers have intended it to be the machine that will finally show that fusion power plants can really be built.

The ITER project formally began in 2006, when its international partners agreed to fund an estimated €5 billion (then $6.3 billion), 10-year plan that would have seen ITER come online in 2016. The most recent official cost estimate stands at more than €20 billion ($22 billion), with ITER nominally turning on scarcely two years from now. Documents recently obtained via a lawsuit, however, imply that these figures are woefully outdated: ITER is not just facing several years’ worth of additional delays but also a growing internal recognition that the project’s remaining technical challenges are poised to send budgets spiraling even further out of control and successful operation ever further into the future.

The documents, drafted a year ago for a private meeting of the ITER Council, ITER’s governing body, show that at the time, the project was bracing for a three-year delay—a doubling of internal estimates prepared just six months earlier. And in the year since those documents were written, the already grim news out of ITER has unfortunately only gotten worse. Yet no one within the ITER Organization has been able to provide estimates of the additional delays, much less the extra expenses expected to result from them. Nor has anyone at the U.S. Department of Energy, which is in charge of the nation’s contributions to ITER, been able to do so. When contacted for this story, DOE officials did not respond to any questions by the time of publication.

The problems leading to these latest projected delays were several years in the making. The ITER Organization was extremely slow to let on that anything was wrong, however. As late as early July 2022, ITER’s website announced that the machine was expected to turn on as scheduled in December 2025. Afterward that date bore an asterisk clarifying that it would be revised. Now the date has disappeared from the website altogether. ITER leaders seldom let slip that anything was awry either. In February 2017 ITER’s then director general, the late Bernard Bigot, discussed its progress with DOE representatives. “ITER is really moving forward,” he said. “We are working day and night.... The progress is on schedule.” The timeline he presented implied that everything was on track. Construction of the ITER complex’s foundation, which incorporates an earthquake protection system with hundreds of tremor-dampening rubber- and metal-laminated plates, should have been almost complete. From there, assembly of the reactor itself was planned to begin in 2018. At the time of Bigot’s remarks, two of its major pieces—a massive magnetic coil to wrap around the doughnutlike tokamak and a large section of the vacuum vessel that makes up the tokamak’s walls—were supposed to be ready to ship within the month and by the end of the year, respectively. Instead the coil would take almost three more years to complete, as would the vessel sector. The pieces were completed in January and April 2020, respectively. In fact, a large proportion of the big components of the machine were behind schedule by a year or two years or even more. Soon ITER’s official start of assembly was bumped from 2018 to 2020.

Then, in early 2020, the COVID pandemic struck, slowing manufacturing and shipping of machine components.

In late 2021 the ITER Council quietly asked for a revised schedule and estimate of costs, which was eventually presented at a closed meeting in June 2022—almost precisely a month after Bigot died from an unspecified illness. Some months later, when I asked Laban Coblentz, ITER’s head of communications, what exactly that revised schedule was, like everyone else on the project, he refused to disclose this information—or any other hint of how grave the delays or cost overruns were likely to be. According to Coblentz, Bigot’s death had pushed ITER into a “rather traumatic transition in leadership” that effectively rendered the revised schedule moot. There wasn’t, he said, “any relevance to providing you with an internal document, circulated to the ITER Council in June [2022], which is no longer current or in any sense accurate.”

In response to this stonewalling, earlier this year I initiated a lawsuit under the U.S. Freedom of Information Act seeking to reveal the extent of ITER’s expected schedule and cost troubles. So far, the lawsuit has been partially successful. It has extracted partially redacted documents revealing that in November 2021 ITER’s internal estimates showed the project already facing about 17 months of delays. By the time of the June 2022 ITER Council meeting, the number had doubled to roughly 35 months of delays—enough to easily add billions of dollars to ITER’s already bloated budget. But this timeline didn’t reflect other events bound to introduce even more delays.

Chart shows International Thermonuclear Experimental Reactor project’s shifting completion dates and increasing budget estimates for each year from 2006 to 2023.
Credit: Amanda Montañez; Source: Charles Seife

In addition to some of ITER’s components arriving far behind schedule, some of that machinery also turned out to be defective. Several thermal shields meant to keep ITER’s liquid helium refrigerant cold and protect the walls of the machine corroded and cracked because of the way the welds interacted with an acid used to wash the metal. This needs to be repaired. “So all in all, it’s removing about 20 kilometers of very thin piping, replacing that—in most cases, repairing the thermal shields, in some cases, making new ones,” Coblentz says. “That is not a high-cost component in ITER terms.” In addition, some of the puzzle-piece-like parts of the vacuum vessel—intended to fit together with submillimeter precision—proved not to be manufactured as precisely as needed. “You can call that a manufacturing flaw, legitimately,” Coblentz adds. In November 2022 the ITER Organization decided not only to halt assembly of the vacuum vessel but also to remove the already installed segment for repairs. Even so, Kathryn McCarthy , director of the U.S. ITER Project at Oak Ridge National Laboratory, testified to Congress just this week that ITER’s “continued project progress shows us that it is possible to achieve engineering precision, at the millimeter-scale, on ship-sized fusion components.”

On top of that difficulty, in January 2022 the French Nuclear Safety Authority (ASN) put a stop to ITER assembly entirely. ASN is unconvinced that, among other issues, the planned amount of radiation shielding around the machine will be adequate, and the authority won’t let the assembly go forward until ITER can prove that it can keep personnel safe. But adding much more shielding might pile on more weight than the rubber-and-metal earthquake-resistant foundation can bear. “ASN will reconsider lifting the tokamak assembly hold point on the basis of a self-supported dossier [ASN] requested from the ITER organization,” wrote ASN spokesperson Evangelia Petit in an e-mail to me. This dossier must address, among other things, biological protection against radiation hazards. Coblentz, however, says the impasse has been caused by “excessive conservatism” and suggests that the situation might be resolved by allowing ITER to run at low power so that the radiation hazard can be mapped and understood more fully before switching over to high-power operations.

In late 2022 Bigot’s replacement, Pietro Barabaschi, admitted that the problem with the vacuum vessel and thermal blankets would wreak havoc on the timing of ITER’s much vaunted initial run, its so-called first plasma date. “We are of course very much aware of the consequences as far as schedule and cost are concerned—and they will not be insignificant,” he said in a November 2022 ITER press release. The length and cost of those delays are still unclear, however, and Barabaschi’s statement didn’t address the supply chain issues—or the regulatory ones, which have not improved. In March 2023 ASN found that the qualifications of certain welders—who have to make nuclear-plant-grade welds between metal parts—had been falsified. ITER officials subsequently banned the vendor that supplied the welding services from any activity on the worksite, but ASN required ITER to go through all the relevant contractor contributions and prepare an impact statement about the falsifications. Coblentz believes that an approximately three-year delay caused by late components and supply chain issues discussed in 2022 and a roughly two-year delay caused by the defective vacuum vessel segments and thermal blankets won’t be additive—that many of the problems can be worked out in parallel. In fact, he says, the ITER organization will possibly start installing equipment not needed until after the first plasma date—a date, he suggests, that might not even be a relevant goalpost anymore.

Whether the wait slips four or five or even more years, ITER is far from the only big scientific project to face enormous delays, cost growth and moving goalposts. Such obstacles, its advocates say, are unavoidable when attempting ambitious tasks that require large amounts of technological development. Proponents of megaproject largesse may cite the James Webb Space Telescope (JWST) as an apt example: intended to be completed in a decade at a cost of a bit more than $1 billion, it took 20 years and more than $10 billion to get the telescope off the ground. Those overruns were especially painful for astronomers but in hindsight seem justified, given that they ensured JWST’s successful launch, deployment and ongoing revolutionary observations in deep space.

But ITER and JWST are not remotely the same. ITER’s gestation has been even longer—stretching back to a handshake agreement between Ronald Reagan and Mikhail Gorbachev in the mid-1980s—and its cost is higher than any scientific endeavor in history. Adjusted for inflation, its price is about the same as that of the Manhattan Project, which made the first atomic bombs—and is almost certain to get larger. As early as 2018 the DOE’s undersecretary for science told Congress that the machine was going to cost much more than the then official price tag of $22 billion. ITER officials vigorously disputed this claim, but the as-yet-undisclosed effects of the project’s latest setbacks makes it clear, at least, that the final bill will be billions more still.

And unlike JWST, which began full operation mere months after launch, ITER won’t be fit for purpose for years after its construction ends. The real purpose of ITER—to run high-power fusion experiments using a mixture of the heavy hydrogen isotopes deuterium and tritium—won’t happen until more than a decade after the machine hits its first plasma milestone. (Originally those experiments were supposed to take place just five years or so after ITER’s debut. Over time, that turned into 10 years: the scheduled 2025 turn-on date would have meant a 2035 start to deuterium-tritium operations.) A further slip to ITER’s start date is likely to cause a corresponding delay in the deuterium-tritium experiments.

When assailed by costly, acrimony-inducing delays, the architects of ITER, JWST and other scientific megaprojects typically respond by reminding the public and policymakers that great monuments take time to build. The plans for Notre Dame and other Gothic cathedrals, for example, were of such grand scale and intricacy that, from their outset, everyone knew their creation would span generations; no one present at Notre Dame’s beginnings assumed they’d live to see it finished. ITER’s designers, however, did not initially hold such lofty expectations for the project. Instead they fully believed they’d see it completed within a couple of decades. Yet the project is now entering its third generation of planning and construction, and its important experiments are at least another generation away. ITER has become the Gothic cathedral of our time: a beautiful but immensely complex structure that we pray will help us find salvation from our energy and climate woes.

Then again, perhaps a cathedral is the wrong metaphor: while Notre Dame took a century to complete, it became an active structure much more quickly, one that was used for its intended purpose less than a generation after construction began. Nobody can say when that will be true for ITER. With each passing decade, this record-breaking monument to big international science looks less and less like a cathedral—and more like a mausoleum.