As helium-3 runs scarce, researchers seek new ways to chill quantum computers

Helsinki—One morning in November 2025 here on the outskirts of Finland’s capital, David Gunnarsson slips off his winter coat and steps into a factory where the coldest temperatures in the universe are forged. Gunnarsson is the principal scientist for Bluefors, the world’s leading manufacturer of dilution refrigerators—researchers’ portal into the quantum world.

All around him, dozens of machines, draped with pipes and wires like golden chandeliers, shuttle between workstations on an overhead conveyer belt. Inside each device, liquid helium chills stacks of copper plates to ever-colder temperatures. At the bottom, a gold-plated finger will freeze a tiny electronic chip to just a hair above absolute zero (0 K, or –273.15°C). “The magic happens there,” Gunnarsson says.

Quantum technologies—from detectors that map the faint afterglow of the Big Bang to microscopes that image atoms—flourish at those temperatures, where thermal motion nearly vanishes. But the biggest booming application is quantum computers—the futuristic devices that may be able to solve problems traditional computers can’t. Tech giants such as Google, Microsoft, and IBM are pouring billions of dollars into quantum computers—and each of them demands fridges like the ones built at Bluefors.

As quantum computers get bigger, so, too, must the fridges. In a room at the factory, Gunnarsson shows off Bluefors’s next-generation fridge, a towering device the size of an elevator car. Called Kide, after the Finnish word for snowflake, it’s designed for today’s largest quantum computers, which can contain more than 1000 qubits—the quantum equivalent of transistors.

Around the corner, Gunnarsson opens a closet hiding a stack of scuba tanks filled with the fridges’ precious fuel: helium-3, a rare isotope of the gas that fills birthday balloons. This exotic ingredient is both the technology’s secret sauce and its Achilles’ heel. Helium-3 is primarily sourced from the decaying components of nuclear weapons, making it one of the most expensive substances on the planet.

Its scarcity could prove a serious bottleneck for emerging quantum technologies. Some researchers are keen to head off a supply crunch by developing new helium-3 sources—perhaps even mining it on the Moon. But others are trying to sidestep the problem entirely by reinventing how ultracold temperatures are reached. In Germany, one company is resurrecting a century-old magnetic cooling technique. And at Aalto University, a 30-minute train ride away from Bluefors, two groups are pursuing rival ideas for integrating coolers into the chips themselves. They aim to chill with superconducting traps for hot electrons or lights that radiate heat away—dispensing with the giant chandeliers and elaborate plumbing.

A full-blown helium-3 crisis is unlikely in the next decade. But some quantum researchers say they are already feeling the pinch of rising prices. “So many people can’t afford this kind of science any longer, and that’s so sad,” says Silke Bühler-Paschen, a physicist at the Vienna University of Technology. “One has to come up with alternatives that make all this feasible again.”

In 1951, at a conference at the University of Oxford, German physicist Heinz London set all of modern cryogenic cooling in motion with a casual suggestion. Liquid helium was already a common refrigerant at the time, good for reaching temperatures of about 4 K, its boiling point. But London suggested the cooling could be boosted by mixing normal helium (helium-4) with helium-3, an isotope with one less neutron. The idea was laughed off as fantastical. Colleagues questioned where London planned to source the gas. Helium itself, with an abundance of just 5 parts per million in Earth’s atmosphere, is already a precious commodity. Helium-3, which constitutes 0.0001% of that helium, is extraordinarily rare.

But sure enough, a few years later, an unlikely source of helium-3 appeared. In 1955, at the height of the Cold War, the U.S. government began to develop hydrogen bombs, thermonuclear weapons whose destructive power is juiced by a dollop of tritium. With a half-life of 12.5 years, tritium happens to decay slowly into helium-3, which can be captured and sold. A decade later, a group in the Netherlands completed the first modern-day dilution refrigerator, capable of reaching temperatures as low as 0.22 K.

In principle, the cooling process London proposed works the same as sweating or dropping an ice cube in a drink. In both cases, water absorbs heat from its surroundings not by rising in temperature, but by going through a phase transition—from a liquid to a gas in the case of sweat, or from a solid to a liquid for the ice cube.

Similarly, a mixture of helium-3 and helium-4 absorbs heat during a special kind of phase transition. Below a temperature of roughly 1 K, the mixture naturally separates, like oil and water, into two layers: concentrated helium-3 sitting on top of a dilute mixture of the two species. If some helium-3 atoms are pumped out of the dilute layer, replacements will diffuse down from the concentrated layer. Like evaporation, this phase change draws heat from the environment. The helium-3 that’s pumped away can be captured and recycled back into the concentrated pool, allowing the cooling process to continue indefinitely—theoretically all the way down to a few millikelvins, thousandths of a degree above absolute zero. In practice, dilution fridges today operate below 20 millikelvins—hundreds of times colder than interstellar space.

Just as dilution fridges entered the scene, Finnish physicist Olli Lounasmaa, sensing the potential to corner a research market, established the Low Temperature Laboratory at what would later become Aalto University. Lounasmaa drew public attention—and funding—by regularly breaking his own world-record coldest temperatures, cementing Helsinki as a global hot spot for low-temperature physics.

Despite their success, the early dilution fridges were laborious to operate. They featured two main stages. First, a regularly replenished cold bath of pure helium-4 cools to 4 K as it boils off. Then, helium-3 is introduced and compressed, bringing the mixture down to the critical temperature where the separation occurs and kicks off further cooling.

By 2005, Rob Blaauwgeers, then a postdoctoral researcher at the Low Temperature Laboratory, had grown tired of coming in on weekends to refill helium tanks. He teamed up with his college friend Pieter Vorselman to streamline the first stage of the process by using a sequence of closed pipes to capture and recycle the helium-4 as it boils off. After constructing a prototype—which is still in operation, connected to a beer keg of helium-3—they went into business. In 2008, Blaauwgeers and Vorselman founded Bluefors—a combination of their last names. The private company has sold more than 1700 fridges to date and in 2024 grossed more than $200 million.

The helium-3 that dilution fridges rely on comes mainly from the U.S. government. The Department of Energy (DOE) collects the isotope when servicing aging tritium reservoirs in the U.S. nuclear stockpile and sells it off to companies, which in turn distribute it to researchers and manufacturers. Supplies are subject to geopolitical drama. After the 9/11 terrorist attacks, for instance, the demand for helium-3 in weapons detection systems made the gas nearly impossible to buy for research, prompting intervention from Congress.

Paschen experienced those vagaries starting in 2008, when she ordered a new dilution fridge filled with 180 liters of helium-3. By the time the fridge was ready for delivery, the price of helium-3 had risen from about $100 per liter to $3000. Today, 1 liter of the gas—less than a teaspoon of liquid—runs for anywhere from $1000 to $20,000, depending on subsidies and discounts. Up to one-quarter of the total price of a dilution fridge—which can cost more than $600,000—can come from “the little droplet of helium-3 you have in there,” says Paschen, who uses her fridge to study phase transitions in exotic materials. “You cannot buy a perfume that’s so expensive, or even a diamond.”

A 2021 DOE document suggests the U.S. stockpile is roughly 90,000 liters, up from 50,000 liters during the crisis in 2010. “We’re in a better spot,” says Christopher Landers, director of DOE’s Isotope Program. But his team is bracing for a looming surge in demand from quantum computing. “If that demand hits, this is serious.”

The largest quantum processors today feature hundreds or thousands of qubits. A practically useful device may require millions. Cooling such a large quantum chip—along with all the wiring to control it—would require a dilution fridge even bigger and more powerful than Kide. With current technology, a single machine that size might require tens of thousands of liters of helium-3, Gunnarsson estimates—a sizable chunk of the estimated U.S. stockpile. The U.S. government is taking heed. In January, the Defense Advanced Research Projects Agency put out a request for cryocooling proposals that require no helium-3, citing “dramatic transformative potential in both defense and commercial domains.”

The Bluefors team remains sanguine about long-term helium-3 supplies. “Helium-3 is mainly a byproduct of nuclear weapons—if you see the world today, it’s not really going away,” Vorselman says. Gunnarsson believes growing demand will induce nations and companies to find new sources. In 2021, an energy company in Canada began to extract helium-3 from the tritium that collects in special heavy-water nuclear power plants, offering the first nonmilitary source of the gas. Meanwhile, some fusion companies want to use helium-3 as a fuel, and they say their reactors will be able to breed more than they need. “If we need more helium-3, it’s going to be produced,” Gunnarsson says.

But the company is hedging its bets, exploring ways to optimize its fridges to consume less helium-3. It’s also looking into a literal moonshot. The Moon’s surface could be rich in helium-3 delivered by the solar wind, which isn’t deflected by a magnetic field as on Earth. Last year, Bluefors struck a $300 million prospective deal with Interlune, a U.S. company, to extract it. (If this sounds like the plot of a bad sci-fi movie, that’s because it was.) The company plans to launch its first exploration mission around 2027, and has promised to deliver 10,000 liters annually to Bluefors over the following decade. “I don’t know how far they’ll come, but it’s worth a chance,” Vorselman says.

But other researchers say a radical shift is needed. “If quantum is becoming a major industry, I really don’t see how it could be built on the basis of dilution cooling,” says Alexander Regnat, founding CEO of kiutra, a German cryogenics company.

In graduate school at the Technical University of Munich, Regnat was determined to overthrow the dilution refrigerator. He set his sights on magnetic cooling, a technique that had been around since the 1930s but had fallen out of fashion. It relies on the magnetocaloric effect, which causes certain materials to heat up when placed in a magnetic field that aligns their atoms’ magnetic spins. If you dump the heat and turn off the field, the atomic spins revert to their natural disorder, drawing heat from the environment and cooling anything in the vicinity.

Historically, the problem with magnetic cooling was that it only delivered a single pulse of cooling. Once the magnet was removed, there was no way to keep a sample from warming again. But Regnat and colleagues have developed ways to attach a sample to two magnetic fridges simultaneously, allowing one to recharge while the other cools. In addition to enabling indefinite cooling, this trick also means the magnetic fields can be about 100 times weaker, and less likely to interfere with sensitive quantum devices.

Anasua Chatterjee, a physicist at the Delft University of Technology, bought one of kiutra’s machines with a grant from the beer company Carlsberg. It’s not cold enough to operate a full-stack quantum computer, but it’s good enough to perform quality control tests on the quantum chips her lab fabricates. And it doesn’t involve helium-3—or worries about expensive accidental releases. “It can more or less do everything that the dilution fridge can do,” Chatterjee says. “It’s a bit more idiot-proof.”

Regnat says there’s still room for improvement. In March, kiutra unveiled an modular design that stacks together multiple magnetic fridges, to compound the cooling power they can deliver to a single device. By the end of the year, the company aims to release a fridge that can reach 20 millikelvins, with enough cooling power to put it on par with Bluefors machines. “I’m very confident this will be available as a real alternative and even outperform dilution refrigerators in the future,” Regnat says. “We’re getting there.”

At a café on the Aalto campus, Mika Prunnila blows on his fresh cup of coffee, the steam condensing on his sleek rectangular glasses. He belongs to a small cohort of researchers looking to cool quantum chips from the inside, instead of sticking them in a bulky refrigerator. For quantum computing, what really matters is not the ambient temperature of the room, but the heat, or average energy, of the electrons racing through the chip and its wires. Shoo away the fast, hot electrons—like blowing steam off a cup of coffee—and you’ve got a cold chip.

His technique traces back to a 1990s proposal to cool a metal by siphoning off the most energetic electrons into a superconducting trap. It relies on a junction between the metal and a superconductor, separated by a thin insulating barrier. Because superconductors have a gap in the energy states that electrons can occupy, they act as filters: Only the hottest electrons can tunnel across the barrier to an available state, carrying heat away.

Aalto physicist Jukka Pekola realized he could strengthen the effect by adding a superconducting junction to the other side of the metal and reversing the filter, so only the coldest electrons in that superconductor tunnel into the metal. The trick effectively doubled the cooling power.

In 1994, Pekola constructed the first double-junction sandwich out of copper and superconducting aluminum. Short on copper, he ran to a mechanics shop with a hack saw, cut off a piece of copper pipe, and brought it back to the lab, where he vaporized the metal and deposited it on his sandwich. “Things have changed a lot since then,” Pekola quips. To his surprise, the makeshift apparatus worked: It cooled electrons on the chip from 300 millikelvins to 100 millikelvins. Over the following decades, Pekola drafted a road map for microscopic, on-chip refrigerators and handed the project off to Prunnila at VTT Technical Research Centre of Finland, a state-owned research institute that shares an office building with Aalto. “I’m very happy that some of the things that we do seem to be some small seed for something practical,” Pekola says.

Magnetic cooling

Certain materials heat up when a magnet aligns their atomic spins. After that heat is dumped and the magnet is removed, the natural disordering of the spins can draw heat from a quantum chip.

Photonic cooling

If a light-emitting diode (LED) is powered with less electricity than it needs, the electrons inside it will steal heat from the atomic lattice to change energy levels and glow, drawing heat from a quantum chip.

Electronic cooling

Hot electrons jump over a barrier to high-energy states in a superconductor, leaving a cool chip. The same barrier also prevents vibrations in the material from warming the chip.

At VTT, Prunnila has been fighting another source of heat in a quantum chip: noisy vibrations in the crystal lattice called phonons. In 2020, his team managed to tame these vibrations using a specialized junction that allows hot electrons to escape from silicon while protecting it from phonons in the superconducting traps. They used this trick to cool an entire silicon chip by about 40%, from 244 millikelvins to 161 millikelvins.

In 2024, they showed the method also works at a higher temperature stage, where phonons are stronger and harder to block, cooling a chip from 2.4 K to 1.6 K using junctions made of niobium. They plan to stack chips and junctions on top of one another, building multiple cooling stages like a miniature dilution fridge. Simulations by his team suggest a cascading cooler should be able to get their chips from a few kelvins—where helium-4 can still reach—to tens of millikelvins, where many quantum computers must operate. “We now have all the pieces of the puzzle,” Prunnila says.

Prunnila sees on-chip cooling as a natural alternative to giant cooling systems that he says will become impractically cumbersome and expensive as quantum tech develops. Doing the majority of cooling on-chip would allow the entire apparatus—including the helium-4 precooling stage—to be shrunk to a fridge the size of a suitcase. “The whole field has been stuck on a track that didn’t lead to scalability,” he says. Paschen is intrigued and cautiously optimistic. “It’s just electrons moving around—it’s really elegant,” she says. “It would be a shame if it doesn’t come to a really big market application.”

Down the street at Aalto, Jani Oksanen has an even brighter idea for integrating refrigerators into chips. It grew out of frustration with his home movie projector, an old model that relied on hot, short-lived incandescent halogen lamps. He looked into projectors powered by light-emitting diodes (LEDs), which efficiently convert electricity to light. He never got around to upgrading his projector. But he did develop a fascination for LEDs.

LEDs work by applying an electric current to a special semiconductor, which forces electrons to fall to lower energy levels. As they relax, they give off their surplus energy as light. Normally, LEDs warm up, because some of the supplied electrical energy becomes heat. But if the voltage is a little lower than the electrons need to switch energy levels, the light can still glow. That’s because a fraction of the electrons will steal heat from the atomic lattice to make the jump—and the photons they produce will radiate the heat away. Starve the LED, and it starts to eat its own heat. “LEDs are not just simple electricity-to-light converters,” Oksanen says. “They are actually heat pumps.”

Physicists still haven’t managed a direct observation of this photonic cooling, mainly because the radiated heat is often deposited very close to the device, making the temperature drop difficult to detect. But in 2019, Oksanen’s group managed to show LED power conversion efficiencies exceeding 100%, which they took as indirect evidence of photonic cooling. In theory, you could also capture the light emitted by the LED with a solar cell and recycle that energy to help power the LED—allowing for massively efficient cooling, he says.

The technique may work best at higher temperatures, where there’s more energy to draw from a material. But his team is also exploring whether LEDs can cool efficiently at very low temperatures, as some have predicted. Meanwhile, other groups are pursuing a related cooling method that involves shining lasers at materials that re-emit the light at higher energies by drawing heat from the lattice. Such techniques could also find applications beyond quantum chips, cooling buildings as well as infrared cameras and electron microscopes.

Oksanen isn’t yet convinced LED cooling is practical for quantum cryogenics. “I don’t know that it’s possible,” he says. “But the people saying that it’s not possible, they don’t know that either.”

Over a customary Finnish dinner of reindeer and salmon soup, Gunnarsson, Prunnila, and Oksanen meet to plot out the future of cryogenic cooling. Although they are pursuing different paths toward absolute zero, they see each other as allies more than competitors. That’s in part because they all belong to the same lineage of researchers that hail from Lounasmaa’s laboratory and his “low-temp mafia,” as Oksanen calls it.

It’s also because none of them expects the dilution fridge to disappear any time soon. Its rivals, despite progress, still lack the power to get bulky samples cool enough. Yet Gunnarsson says it’s good for Bluefors to keep close tabs on the competition. The company’s objective is to cool by the most efficient means possible—even if that means one day shifting gears.

He imagines a future where hybrid cooling systems stack various platforms together at different temperature stages to “use the best technology where it is best,” he says. “If you draw on all the advantages these different technologies have, maybe you can also reduce the need for helium-3.”

Then they toast over shots of Jaloviina, a traditional Finnish brandy—which, like many things in life, is best served chilled.