Microsoft's big bet on helium-3 fusion explained
Magnets, not lasers, could be secret to endless power (if it actually works)
Feature Microsoft is making a big bet to cut its carbon emissions with a staple of science fiction: helium-3 fusion power plants.
Of course, that's assuming nuclear energy startup Helion can stick the landing and that's a rather big if. Despite decades of research and development and billions in investment, fusion power has thus far eluded the world's foremost scientists.
As we reported last week, Redmond signed the power purchase agreement with Helion to provide the software giant with a steady supply of clean energy starting in or about 2028. But before that can happen, Helion needs to commercialize its fusion reactor tech.
The basic principle behind fusion is well understood. Compress light gasses – mostly hydrogen – at high enough temperatures and they fuse together to form heavier atoms, like helium, and release massive quantities of energy in the process.
However, most of our successes around fusion have been from the standpoint of harnessing it to blow up cities, wipe humanity off the face of the planet, and other similarly gruesome ends. That's the nuclear fusion that occurs in thermonuclear weapon explosions, a highly destructive process that needs a fission-based component to get it going.
Recreating the physical processes that power the fusion engine in our Sun to generate electricity, however, has proven much, much harder to do.
To be clear, if your only goal is to fuse hydrogen, the technology has existed for decades – just look up Farnsworth fusor. The trickier proposition is extracting more power than you put into the reaction in the first place. In other words, achieving a sustained fusion reaction that generates net-positive power.
Last year, researchers at Lawrence Livermore National Laboratories (LLNL) made a major breakthrough achieving fusion ignition. The process involved firing 200 lasers at a tiny pellet of fusion fuel until it imploded, forming a dense super-hot plasma in which fusion occurs, releasing energy in the process.
However, those same scientists have been careful not to set unreasonable expectations. LLNL's National Ignition Facility generated more energy, 3.15 megajoules, than the 2.05 megajoules of energy used to kick off the reaction, it took 322 megajoules just to run the lasers, making it a net loss.
Also, 3.15 megajoules sounds like a lot but it amounts to 0.875 kilowatt-hours.
In an interview with Bloomberg following the announcement, Kimberly Budil, LLNL director, said the breakthrough could pave the way for a commercial fusion power plant within "several decades."
How Helion plans to harness the power of the Sun
Helion's bold claims have attracted the attention of investors, and it's certainly not hurting for cash. To date, the company has raised more than half a billion dollars, of which $375 million was reportedly provided by OpenAI CEO Sam Altman. Given what we know about the power requirements to train large language models (LLM) like GPT-4, Altman's investments are hardly surprising. This is especially so when you consider they were trained in Microsoft datacenters.
Helion has been working on fusion reactor tech for 10 years. But as chief business officer Scott Krisiloff told The Register, the research that led to the company's founding started five years earlier while much of the team was working on plasma physics technologies at MSNW.
Today, the company aims to build fusion power plants capable of delivering 50 megawatts or more of power. Compared to the big pressurized water reactors (PWR) we've had for more than half a century, that's nothing. PWRs often produce power on the scale of gigawatts. However, 50 megawatts should be more than adequate to power a fairly large datacenter campus – though that may change if Microsoft's appetite for GPUs goes unchecked.
The idea that fusion power is just around the corner is at odds with what we know about the technology. And yet Krisiloff says the company's seventh-gen prototype could start producing small quantities of power within the next 18 months.
"Our seventh prototype is under construction in Everett, Washington, north of Seattle. There's 160 team members today who are working on building that," he said. "We are targeting the ability to produce electricity in 2024… And then, based on that, we expect to be able to commercialize by 2028."
You might wish to take those claims with a healthy dose of salt. Diving through past coverage of the fusion upstart reveals the company has made some rather fantastic claims. In a 2018 article, the company claimed it would produce a 50MW reactor by 2021. And in 2014, the company was claiming commercial fusion by 2019.
- Since humans can't manage fusion, the US puts millions into AI-powered creation
- NASA's space nuclear power program is a hot mess
- Brit fusion magnets set for US gamma ray bombardment test
- America's nuclear fusion 'breakthrough' is super-hot ... yet far from practical
Helion's habit of insisting fusion power is almost here led Daniel Jassby, who ran the Princeton Plasma Physics Lab until 1999, to label the company's tech as voodoo fusion in a 2018 article [PDF].
Jassby defines voodoo fusion energy as "those plasma systems that have never produced any fusion neutrons, but whose promoters make the claim of near-term electric power generation."
Voodoo fusion or not, Helion's designs work quite a bit differently than the National Ignition Facility's reactor. Instead of lasers, their design uses strong electromagnetic fields to heat deuterium hydrogen and helium-3 gas into a doughnut-shaped plasma called a field reversed configuration (FRC) that generates its own electric field. These plasma are formed at each end of an hourglass or dumbbell shaped vessel and accelerated to "greater than one million miles per hour" before they're smashed together and further compressed in a central chamber. This process drives the plasma's temperature to more than 100 million degrees Celsius, releasing energy in the process as well as strengthening its own magnetic field.
"100 million degree temperatures are temperatures at which bulk fusion starts to happen, so this was an important signaling in terms of what the sixth prototype was capable of doing," Krisiloff said.
The design is also different in how Helion plans to extract the energy generated by the fusion reaction. Instead of using energy from the reaction to generate steam and spin a turbine, like most power plants, Helion plans to extract power from the reaction inductively.
This process is then repeated – pulsed – at varying intervals to control the output of the reactor. That's the idea anyway. Critically, Helion has yet to achieve ignition.
I thought helium-3 was rare
The idea of using helium-3 as a fusion fuel is by no means new. The problem is that while deuterium hydrogen is abundant on Earth, helium-3 is not.
Researchers have found evidence that the Moon may be a plentiful source of helium-3. But, unlike the fictional fusion power company Helios from Apple TV's For All Mankind, the real-world company isn't planning on setting up mining operations on the Moon. Not that it would help them on the ambitious timescales they're working with anyway.
This means that helium-3 will need to be sourced terrestrially, which is in itself problematic as the primary means of production involves the radioactive decay of tritium isotopes. And tritium – a form of hydrogen with one proton and two neutrons – is also pretty rare.
To get around this, Helion plans to use fusion to produce more helium-3.
"If you do deuterium-deuterium fusion, helium-3 is a byproduct of that fusion. So if you have a highly efficient machine at doing fusion you can actually do a deuterium-deuterium shot first and then extract the helium-3 from it, and then do deuterium-helium-3 for electricity production," said Krisiloff.
While it remains unclear whether or not Helion will prove the Department of Energy wrong, at least we won't have to wait long to find out.
Microsoft hedges its bets
Although the terms of the deal aren't clear – we don't know whether Microsoft has invested any cash in Helion on top of agreeing to buy power from the company – Microsoft is clearly keeping its options open.
From what Krisiloff tells us, Microsoft will pay close to market rate for fusion power, which means Helion has a pretty big incentive to figure it out. If Helion can't manage to make fusion more cost effective than alternative power types, the company could end up losing money on the deal.
It's worth emphasizing, however, that Microsoft is by no means putting all of its eggs in Helion's basket. The company has invested heavily in renewable energy and has even partnered with local utilities to serve as a load balancer for wind and solar at its datacenters in Dublin.
Microsoft founder Bill Gates also has his own nuclear power plans, with TerraPower building small modular nuclear reactors (SMR). As we have previously reported, a handful of SMRs could easily power a large datacenter campus. ®