Sitting in the control room of Tri Alpha Energy's experimental fusion reactor, in front of computer screens labeled “plasma guns” and “shot control,” I feel slightly anxious as we make preparations to fire. This reactor is an early prototype for a power plant that would generate energy in a controlled version of what happens inside stars and H-bombs.

On the video feed overhead, I see workers out on the floor of this nondescript warehouse near Irvine, Calif., walking away from the large reactor toward the doors. The shiny, cylindrical vacuum chamber at the reactor's center, about as long as two school buses parked bumper to bumper, is encircled by two dozen ring-shaped electromagnets, each taller than I am and as thick as my leg. The temperature inside that chamber will rise, on my command, to around 10 million degrees Celsius—though only for an instant.

“Click that button,” the operator tells me. I do as I'm told.

In an adjacent building, four massive flywheels, spun up this morning using power from the local grid, release a 20-megawatt surge of electricity. The current energizes the ring magnets and charges up banks of beefy capacitors, preparing them for the huge zap to come. Within two minutes all the meters on my control screen have switched from “Preparing” to “Armed.”

The operator leans into a microphone. “Triggering,” he says over loudspeakers. Warning lights start flashing. I move the cursor to the button marked “Trigger.” Then, I click it.

The capacitors release their pent-up electricity in a microsecond. Clouds of hydrogen ions form at the opposing ends of the vacuum cylinder and are propelled toward the center at nearly a million kilometers an hour. There they collide and form a hot, spinning plasma shaped like a giant, hollow cigar.

It sounds dramatic, but in the control room there is no flash, no roar—just a faint “ping,” as if someone inside the reactor room had dropped a wrench onto the concrete floor. In an instant, the plasma blob has dissipated, and the computers have started processing the gigabyte of data streaming from dozens of sensors in the reactor. The warning lights switch off, and the workers return to their tasks.

Just another shot at fusion. When you fire as many as 100 shots a day, as Tri Alpha has been doing, one more is no big deal.

After 50,000 little pings in just two years, the C-2U test machine had by the time of my visit in February given the Tri Alpha team all the data it needed to move on. In April, Michl Binderbauer, the wiry, excitable physicist who is the company's chief technology officer, told his engineers to tear it down and cannibalize its parts for a more advanced reactor—dubbed C-2W—to be completed in mid-2017.

Tri Alpha's approach—build a prototype quickly, test it just enough and then trade up to a better one—is a striking departure from the norm in fusion research. For decades academic scientists have designed gargantuan machines intended to solve the mysterious behaviors in the fiery, pressurized plasmas that are supposed to create fusion reactions but often do not. Binderbauer, the son of a Viennese serial entrepreneur, exemplifies a new strain of fusioneer, driven by investors, an engineering mindset and an unwavering focus on building a practical power plant, not a monument to high-energy physics.

Several other start-ups, such as General Fusion outside Vancouver, are similarly betting that they can build a commercial machine without having to untangle every detail of the complex physics along the way. Such fusion power plants would run on fuels derived from ocean water or common minerals that are nearly inexhaustible and have no carbon. The plants would therefore produce almost no greenhouse gases. They also would pose virtually no radiation or weaponization risk and would generate enough electricity to run entire cities—all day, every day. All the new pioneers need to do is solve some of the hardest physics and engineering problems humans have ever tackled.

Right now the pragmatists have people's attention because the academics have hit practical dead ends: enormous reactors that have clarified some fusion science but are not on track to pump electricity into the grid by midcentury. One example is the National Ignition Facility (NIF) at Lawrence Livermore National Laboratory, a $4-billion machine that zaps tiny canisters of fuel with trillion-watt laser pulses. “NIF fires just a few hundred shots a year,” Binderbauer says in his Austrian lilt. A power plant would have to fire tens of thousands of times a day. The system has delivered useful weapons research (its primary purpose), but its energy output would have to increase almost 30,000-fold just to cover what is required to run the lasers—and many times beyond that to be commercially useful. Two years ago Livermore pulled the plug on designing a prototype power plant.

The second discouraging example is ITER, a 10-story-high machine under construction in France by a consortium of nations. It will rely on giant superconducting magnets to control a plasma burning at roughly 150 million degrees C for minutes at a time. Even if it succeeds, ITER will make no electricity.

The politicians who launched ITER in 2006 expected it to cost $11 billion and to be fully constructed this year. As of May, the cost had ballooned to $20 billion, with the U.S. on the hook for about $5 billion. Full operation will not come until 2035 at the earliest. Frustrated senators voted 90–8 to cut off U.S. funding. But after a subsequent though guarded vote of confidence by the U.S. Department of Energy, Congress was, at the time of this writing, poised to stay in the game, at least until next year.

Forewarned by the glacial progress of the giants, Binderbauer and the other mavericks are pinning their hopes on smaller machines that approach the problem from new angles. To deliver, they must compress a tiny amount of fuel densely enough, heated hot enough and confined that way long enough for atoms to fuse together, converting some of their minuscule mass into gobs of energy. NIF and ITER are at opposite ends of a spectrum of plausible designs that spans a huge range of plasma densities and energy-confinement times (a measure of how long heat stays inside the plasma). Most of the newcomers are searching for sweeter spots that lie in the less explored middle ground.

Equally important, the start-ups are designed to succeed or fail relatively quickly. Their reactors are “potentially 100 times less expensive than ITER, easier and faster to build, and lend themselves to faster research progress,” says Scott Hsu, a fusion physicist at Los Alamos National Laboratory who works with HyperV Technologies, yet another start-up. (In this design, hundreds of guns fire bursts of argon plasmas into the center of a spherical reactor, where they converge and compress hydrogen fuel.) Any show-stopping flaws in these schemes most likely will show up well before the stakes rise to billions of dollars and decades of time.

Click or tap to enlarge

Credit: Don Foley (reactors), Jen Christiansen (particle diagram)

That pleases their investors. General Fusion's $100-million bankroll has come in part from founder Jeff Bezos, the Canadian government and the sovereign wealth fund of Malaysia. Tri Alpha claims to have raised hundreds of millions of dollars from investors that include Goldman Sachs and Paul Allen, the co-founder of Microsoft. Another fast-moving group is Sandia National Laboratories, supported in part by the DOE's Advanced Research Projects Agency–Energy (ARPA-E), which funds long shots the way venture capitalists might do.

The backers are placing high-risk, high-payoff bets. Indeed, fusion research has been littered with cases where “nature says, ‘nice idea, but it doesn't work that way,’” quips Stephen A. Slutz, who is the senior theorist on the Sandia project.

Squeezing Fire

The challenge of stabilizing a furious plasma arises from the very nature of fusion itself. Two atomic nuclei, stripped of their electrons, can fuse only when they get close enough, for long enough, that the attraction of the strong nuclear force between them overcomes the electrostatic repulsion among the protons. When that happens, the ions merge to form a single nucleus of a heavier element that has less mass than the ingredients did. The missing matter transforms into bountiful energy, in the form of photons and fast-moving subatomic particles. Fission reactors, in contrast, extract energy from atoms such as uranium that are falling apart rather than joining together.

To get high rates of fusion, the ions in a plasma must be moving toward one another fast—but not too fast. That typically means a plasma temperature north of 100 million degrees C. A reactor must squeeze the superheated plasma into a relatively small space inside a vacuum chamber and hold the nuclei there until the reactions happen. As a rule of thumb, the product of the plasma's density and energy-confinement time has to be greater than about 1014 seconds per cubic centimeter. A wide combination of density, time and temperature can work.

ITER, a “tokamak” reactor design, will use a wispy plasma of about half a gram of neutron-rich isotopes of hydrogen known as deuterium and tritium, floating within a vacuum chamber the size of a small house. ITER aims for a low plasma density, with energy confined for seconds at a time.

NIF, in contrast, trains up to 500 trillion watts of laser blasts from 192 directions onto a tiny canister encasing a frozen speck of solid deuterium and tritium. The optics and electronics that create and direct the laser pulses fill a building 30 meters high that is big enough to cover three football fields. To achieve ignition—a state in which the fusing fuel releases enough energy to sustain ongoing fusion reactions with no outside help—NIF seeks an incredibly high plasma density, which it needs because the energy is confined by inertia alone, for just a fraction of a nanosecond.

A big opportunity, says Patrick McGrath, ARPA-E's program director, may lie in the less explored regime between these two extremes: moderate plasma density and moderate energy-confinement times. But no machine has mastered the gremlins of turbulence and instability that inevitably appear in such plasmas. Controlling a hot plasma while fusion roars inside it is like trying to squeeze a candle flame without touching it—but even harder because the ions in a plasma generate their own complex and disruptive electric currents and magnetic fields. “Even if you can get the candle lit,” says Dylan Brennan, a fusion scientist at the DOE's Princeton Plasma Physics Laboratory, “it blows itself out.”

Enter the Upstarts

Tri Alpha has shown the most progress among the start-ups in maintaining a consistent grip on a plasma. “Everything you see here was built in less than a year,” Binderbauer says proudly as we walk the 23-meter-length of the C-2U machine, tiny compared with NIF or ITER. Just three months after it turned on, it was creating up to 100 spinning blobs of hydrogen plasma a day, each having a density of about half the company's design goal of 1014 ions per cubic centimeter. The blobs remained stable and hot for five milliseconds.

That is a long way from the company's vision of a plasma that rotates quietly in place for days or weeks. But the tests were limited by the external power supplies. “Nothing says it can't go as long as we want it to,” Binderbauer claims, in a full-scale reactor that powers itself as well as the grid. The upgraded C-2W will add digital feedback to counteract the tendency of the blob to wobble or drift.

Hsu, who has no affiliation with Tri Alpha, says the company has achieved “tremendous progress. They have essentially solved the stability problem.” But demonstrating longer containment times—and at far higher temperatures, while pumping in a steady stream of fuel—will be crucial because the reactor must run continuously to generate power.

General Fusion's reactor, in contrast, works in pulses. The spherical, steel reaction chamber, erected inside a suburban warehouse, is a meter across and bristles with pistons a third of a meter wide, each nearly as long as Michel Laberge, the company's tall, red-bearded founder and chief scientist. Laberge describes the steampunk-looking machine in a pronounced French-Canadian accent: “Compressed gas accelerates these pistons to 200 kilometers an hour, and then they hit anvils—bang!” he shouts as he claps his hands loudly. “The impacts all have to occur within five microseconds to create a shock wave” that collapses at the exact center of the chamber.

When all the machine's pieces are integrated, they will fire once every second, like a beating heart. With each beat, a smoke ring of plasma squirted into the sphere will compress and set off a brief but energetic cascade of fusion reactions. It is easier to manage turbulence using this pulsed approach, Laberge argues, because each little doughnut of plasma has to remain stable for only a millisecond or so.

Laberge says the injector system has already produced plasmas having the right preimplosion density, as well as the necessary temperature and magnetic field strength. But the plasmas lasted only 20 microseconds—50 times too short—before falling victim to instability. Laberge is confident that a new nozzle design, shaped more like the bell of a trumpet, will twist the magnetic field that the plasma itself creates by just the right amount to hold the fuel together long enough to fuse.

And yet “a lot of people in the field say that General Fusion's approach is never going to work,” notes Brennan, who is helping the company. The critics doubt that a small group of people in a start-up can master plasma problems that have frustrated academic researchers for years. “But scientifically, do we have the answer that tells us they can't do it?” Brennan asks. “No.”

Half a continent away in New Mexico, experiments with a technique called MagLIF at Sandia have accomplished what the start-ups have yet to do: create appreciable amounts of fusion. Like NIF, MagLIF aims for high ion densities—around 1024 ions per cubic centimeter—and energy-confinement times of mere nanoseconds. But the Sandia system is, at 34 meters across, far smaller and much less expensive than NIF because it uses a one-two punch to heat and pressurize fuel trapped inside a cylinder no bigger than the eraser on a pencil.

The so-called Z machine that feeds MagLIF unleashes a 19-million-amp electrical jolt that exerts a powerful magnetic pinch, crushing the cylinder. A brief trillion-watt laser blast ionizes the fuel as it starts to implode. The machine imposes a separate magnetic field to keep the resulting plasma from squirting out the ends of the cylinder. But the collapsing cylinder can develop instabilities that allow fuel to escape through the sides.

The amount of fusion generated in each MagLIF shot has soared 100-fold since tests began in late 2013. “MagLIF works pretty darn well already,” Hsu says. Daniel Sinars, the project leader, says he expects even better results from shots scheduled for late 2016.

If all goes well, the team plans to boost the electrical jolt to 25 million amps. That should generate around 1016 fusion reactions, enough to offset the energy absorbed by the fuel on its way to fusing, matching NIF's 2014 achievement at a mere fraction of the cost. “That would be very exciting,” Sinars says.

Sandia is already drawing plans to upgrade the Z machine. With 65 million amps and the addition of tritium to the deuterium fuel used thus far, the new Z800 could generate up to 100,000 times more energy per shot. Is that enough to achieve ignition, reaching self-sustaining fusion a decade or more before ITER will? Sandia researchers calculate that it might be.

Because Sandia is a national lab, Congress would have to approve any major upgrade, and it has not been in a spending mood. But competition could change that sentiment. According to Slutz, Chinese scientists have already constructed a smaller version of Z and are replicating Sandia's published experiments, and Russia is planning to build a similar 50-million-amp machine.

Turning Up the Heat

If any of these fusion schemes succeeds in reaching the necessary ion density and confinement time, it still must supply the third ingredient required for ignition: an incredibly high plasma temperature. Doing that is hard because light emissions, electron interactions and myriad other mechanisms can cool the plasma enough to snuff out fusion reactions soon after they start.

At Sandia, for example, Sinars and Slutz have been scratching their heads over why the laser has not been heating the fuel nearly as much as their models predict. The thin window that covers the open end of the fuel target may be scattering the light. But a laser may simply be the wrong tool for the job. For a commercial system, “you probably would want to heat the fuel some other way,” Sinars admits. The team is trying to improve laser heating, but if it cannot, the failure at least will have come early in the game.

Tri Alpha has to reach a far higher temperature than its competitors because it is using a fuel blend of protons and boron 11, which burns at 3.5 billion degrees C. That is more than 20 times hotter than needed for deuterium-tritium fuel.

Hotter plasmas tend to be harder to contain. But Binderbauer is betting that Tri Alpha's energy confinement will actually improve as the temperature soars. It has in experiments thus far, but even the new C-2W machine will heat plasmas to just a fraction of 1 percent of the needed temperature and hold them for just 30 milliseconds. Binderbauer concedes that he could lose this bet on physics but says “we don't have data in this regime. We have to go prove it.”

General Fusion must struggle with unproved physics, too—notably, how fast heat escapes from the plasma. “This cannot be calculated from first principles, so there's plenty of space to have a bad surprise—or a good one,” Laberge says. “If the [heat] losses are worse than expected, we can make the machine bigger. But if it grows to be the size of ITER, then we have a problem.”

From Prototype to Power Plant

Champagne corks will pop the day some reactor achieves ignition—and then a long slog of hard engineering work will begin to transform an experimental reactor into a power plant that generates both electricity and profits. To make a dent in the global electricity supply, which is forecast to grow 70 percent by 2040, fusion will have to compete on cost with other clean energy options.

Giant tokamaks like ITER probably will never succeed, says Dennis Whyte, who directs the Plasma Science and Fusion Center at the Massachusetts Institute of Technology, because they eat up too much of their own power to function. The start-ups have put more thinking up-front into engineering but will still face numerous practical challenges.

For the foreseeable future, for example, each MagLIF shot at Sandia will destroy part of the equipment. Deuterium-tritium fusion releases most of its energy as high-speed neutrons, which damage steel parts and gradually turn them radioactive. Any fusion plant that uses this kind of fuel will have to capture the fast neutrons and use their heat to spin turbines that generate electricity while minimizing the side effects. Scientists there are not dwelling yet on how to prevent the damage, and they have only rough and untested notions of how they might quicken the shot rate from several a week to several a minute. HyperV and Magneto-Inertial Fusion Technologies, a small company in Tustin, Calif., are using money from ARPA-E to explore related schemes that might solve some of these issues, but these efforts are not nearly as far along.

Tri Alpha is pursuing proton-boron fusion precisely to avoid the headaches that come with fast neutrons. Fusion with this fuel emits three helium nuclei, known as alpha particles—hence the company's name—and x-rays but hardly any neutrons. The downside: the x-rays carry over 80 percent of the energy produced.

In principle, Binderbauer says, photovoltaic cells lining the interior of the vessel could convert those photons into electricity. But that technology does not yet exist. So the company is exploring the idea of running coolant along the interior wall of the fusion chamber to extract heat deposited by the x-rays.

General Fusion is sticking with deuterium-tritium fuel, despite the neutron issue and the fact that tritium is mildly radioactive, exceedingly rare and very costly. Laberge plans to pump a swirling vortex of molten lead and lithium along the interior walls of the reaction chamber to capture the neutrons' energy. The neutrons will also split some of the lithium atoms into helium and tritium, which can then be recycled as fuel.

It is an elegant solution on a whiteboard, but no one has ever built such a system. The amount of tritium that would be bred is still speculative, Hsu says. And Laberge worries that as shock waves from the pistons pass through the lead-lithium mixture, some of the metal could spray into the plasma, squelching the fusion. “It would be like pouring water on fire,” he concedes.

Roads Less Traveled

Given the disappointing pace at ITER and NIF, Whyte says, “the time is ripe to take all of the science we've built up and look at other optimizations,” including riffs on tokamaks that make them smaller or twist them into odd shapes called stellerators. “I would love to see a race between a very compact tokamak, General Fusion's idea, a compact stellarator and a machine like Tri Alpha's. Let's see what works best.”

Currently that race in the U.S. is relying on the kindness of investors. Federal money for alternative paths to fusion has been dwindling year by year, Hsu notes. He and Stewart Prager, head of the plasma physics lab at Princeton University, have urged Congress to increase research funding to explore innovative fusion concepts, which could allow other ambitious start-ups to rise to the challenge. If any of the innovative concepts are successful, Hsu says, “fusion energy could possibly be developed for a few billion dollars in less than 20 years.”

Maybe, maybe not. As Binderbauer points out, “there's plenty of opportunity for the physics we don't yet know to bite us in the ass.”

But consider the potential payout: a new source of energy that doesn't rely on the whims of the wind or sun blocked by clouds, wouldn't require big changes to the existing electrical grid, doesn't raise concerns about nuclear weapons, can't melt down or irradiate surrounding communities, and might be no more expensive, after it gets going, than other forms of clean energy.

Is it worth taking a few more shots?