Helion
When the grift hits the (atom) accelerator
Way back in 2019, there was WeWork. For those blessedly too young to remember this progenitor to our current moment of crypto schemes, political insider trading, fusion valuations, and assorted other grifter bullshit; This was a real estate sub-leasing company that pretended, badly, to be a SaaS business, and kept the party going to the point of filing an S-1 to go public at a 47x earnings valuation. Amazon, at the time, traded around 4-5x earnings. WeWork, led by all-time grifter Adam Neumann, started to fall apart after analysts and journalists noted the hilarious phrase “Community adjusted EBITDA” in the aforementioned S-1. The company’s offered definition of this metric is too incoherent to bother placing here, but it essentially came down to, “What if EBITDA, but just the earnings?” The Apple TV series is very good, should you ever need a binge watch. Anyways…
We thought a lot about WeWork when looking at the largest climate tech deal of the week, and one of the largest of the year. The news that fusion company Helion has raised $465m at a $15.5bn valuation, which is meant to see the company through construction of its in-progress Orion power plant, and delivery of energy to launch customer Microsoft in 2028. We’re starting to get the vibe that those words mean something slightly different than their carefully phrased public updates might lead you to believe. Let’s dive in.
Polaris
Helion’s CEO, Dr. David Kirtley, has many times said variations of “we don’t want to engage in long discussions of theory, we want to go build prototypes for various aspects of fusion technology, gather data, and iterate fast.” This would seem to make sense, except… Does it? Fusion was first achieved in a laboratory setting in 1958, using a deuterium-fueled magnetic pinch process. It’s not a particularly new concept, and there are many demonstrated methods of using magnets, lasers, super-ovens, or other technologies to mash two atoms together at small scale. What’s proven elusive, in all the decades since, is net-gain electricity.
All of fusion comes from the insight, first published by British physicist Francis Aston in 1920, that the mass of four hydrogen atoms is greater than that of one he-4 (helium) atom. This meant that energy was expelled in the process of combination, which could then be captured and turned into electricity. The problem is that it takes an immense amount of inputted energy to effect the combination of atoms in the first place, and that in all-but-one of the fusion demonstrations since, the amount of input has been larger than the capture on the other side. In short, making electricity from fusion has been possible for decades, gaining electricity has been done once. It’s no coincidence that we think the only company in the space with a prayer of actually pulling off grid-interconnected fusion employs the team that did the once.
Back to Helion. Their seventh generation, largest prototype is called Polaris. Please enjoy this entry in the FAQ on the Helion website in regards to the Polaris project.
“Is Polaris a net electricity machine?
In 2021, we used “net electricity” when discussing the goals of Polaris. In hindsight, we realized that the term “net” didn’t align with the commercial-driven goals of our 7th fusion prototype, and also wasn’t as clear as we intended. Today, we prefer to direct focus to the more concrete goal of Polaris: to demonstrate electricity from fusion, where we show we made fusion energy and converted a portion of it to electricity on the capacitor bank. This achievement will help us move toward commercial electricity production for the grid.”
If the term “net electricity” doesn’t suffice, may we suggest “community adjusted energy” instead?
Orion
Helion broke ground on Orion, their commercial installation for Microsoft, in July 2025 in eastern Washington State. We have a question; What the fuck are they actually building? Let’s posit that the agreement with Microsoft is to start delivering power in Q3 2028, exactly three years after groundbreaking. Let’s further assume that the first three months of construction are preparing and pouring the concrete pad for the facility, after which it will need to settle for 6-12 months before being loaded with significant weight. That eats up a year. The two remaining years would be considered very fast to build a standard nat-gas plant, never mind a first-of-its-kind fusion reactor.
A further complication; Helion, in its preferred ready-fire-aim style, announced the results from the Polaris test campaign in February of 2026. We’re not nuclear physicists and honestly can’t tell you if these figures represent promising milestones on the way to net-gain fusion or not, except that, well, it’s still not net-gain fusion and we would expect the company to be able to do that before building the commercial plant. Interestingly, so did they! The original announcement of the Series E in 2021 cited a projected small amount of net gain from Polaris by 2024. It’s now June 2026, and the company hasn’t announced any such achievement.
Perhaps they already hit the milestone and simply aren’t discussing it publicly. We, as current stealth founders ourselves, get the idea of moving in silence, but come on. The company that can say “we have verifiably achieved net-gain fusion, are publishing our data in a legitimate journal, and know how to do it bigger” is worth $100bn overnight and can have a blank check funding round from any investor they point at. If Helion hasn’t said it, they don’t have it.
Even if we leave all that aside, which we probably shouldn’t, the timeline still has us scratching our heads. Fusion is largely a game of energy and fuel inputs, temperature management, and energy capture and storage. Tactically, this translates into things like the design of various pipelines, control spaces, walls, and chambers in the commercial plant. All of which need to be in a locked construction plan before pouring concrete.
Here’s another weird thing. We’ve spent the past couple of days reading every word we can find online about Orion and the Microsoft deal, and can’t find any references to Orion being a net-gain fusion power plant. There is plenty and sundry about the excitement of delivering electricity from fusion, but zero about this being more energy than is used to trigger the fusion in the first place.
Physics
We’ve been both incredibly hard, and perhaps not hard enough, over the past year or so, on fusion as a space. At this point, we largely think it’s an overfunded science project and distraction from the real work of renewables and batteries. Helion’s similarly-funded competitors, Zap Energy and Commonwealth Fusion Systems, are 3-4 years behind their initially projected technical milestones on the path to fusion. We know this because both of those companies regularly publish in peer-reviewed journals, which Helion does not.
If you’re watching a marathon between three runners of roughly historical similar abilities and training, with two projecting to finish in 2:30 and the third in 1:45, you might start to wonder if the fastest runner jumped on a bike for a few miles while nobody was looking. It’s just physics. Again, we’re just two curious dipshits with no advanced degrees rambling on the internet, but it’s worth noting that the scientific consensus seems to be that Helion’s approach is the most potentially efficient if it works, but also the most technologically difficult of the three, and especially at large scale.
Microsoft and Altman
Here’s where we get to the awkward part.
We think Microsoft is an incredibly competent company, led by the best CEO in big tech. They are more than capable of hiring a few PhDs from the University of Washington to scrub through Helion’s data and designs, and come to determinations about how much of this is viable or not. We’re sure they did. But… Do they care?
Five pieces of perhaps-relevant data.
Helion was originally funded by Sam Altman, CEO of OpenAI. He invested $375m of personal capital into the company, served as chairman of the board until earlier this year, and is a material equity stakeholder.
Altman famously has no equity in OpenAI, and just as famously uses his position there to capitalize and monetize his numerous other startup investments.
Microsoft, Helion’s launch customer, owns 27% of OpenAI. A revision to their partnership in early 2026 means that OpenAI is now permitted to use other cloud hosting providers for their services.
Microsoft is investing $190bn just this year in standing up new AI hosting capacity, and needs that space filled before it turns into an unholy waste of money.
Anthropic, the other big frontier lab who could theoretically use the data centers, is primarily on AWS, Google, and xAI hosting.
So, yeah.
No financial details have been released regarding the Helion & Microsoft deal. We’d be astonished if they ever are. Let’s say that the contract is for $100m annually, which would be an egregious overpayment for 50MW, but whatever. If Microsoft was to turn a blind eye to Helion providing them what is technically fusion electricity, even if at 65% efficiency from the nearby hydroelectric dam on Lake Chelan, and make up any shortfalls by subsidizing an outright purchase from the grid, then do a carefully worded celebratory press release about it….
Well, that might be enough for Helion to sprint into a hugely lucrative IPO before they need to plug in a second plant, and thereby make several billion liquid dollars for Sam Altman. Who might, we then imagine, find himself extolling the virtues of Microsoft’s Azure data centers right in time for the training of ChatGPT11.
About that second plant. Helion’s second announced customer is the steel giant Nucor, with whom they claim they’ll be deploying a 500MW plant by 2030. This is obviously outdated, as a plant of that size would need to have a site selected and to start the permitting process at least 6 months ago, and there seems to have been no mention whatsoever of the project since the flurry of press releases in late 2023. Oh, wait, there was one mention, at least if that’s what you call Helion and Nucor cosponsoring legislation to have the manufacture of parts for fusion energy being included in a lucrative federal tax credit. We wonder if steel giant Nucor might be interested in some of those cash-raining manufacturing deals, purely to be a good partner.
Closing Time
This is the most cynical interpretation of Helion, what they’re up to, and their growth to date. It also may be wrong. There are plenty of reasons for a fusion company to keep their endlessly valuable IP secret, and plenty of companies that have missed early development milestones on their way to resounding success. If Helion has, genuinely, retired all the remaining technical questions to make their approach to fusion work, then it’s possible that they’re in a position where more money will make things move faster to the point that 2028 becomes a feasible goal. They just raised an awful lot of it.
If anyone from Helion sees this and wants to fact check us, on the record and with data, we’ll be happy to publish it with an apology.
We have written before that, as a base case, one should always assume the purpose of a company is to generate returns for shareholders, no more and no less. Incentives are gravity. Here we have a company, whose single largest shareholder is also the most important strategic customer of the customer keeping them afloat, a lot of shifted goalposts and missed timelines, and tech that many smart people say is too good to be true. Maybe we’re wrong, the tech is real and scalable, and we’ll look like absolute morons in a year or two.
You may also be wondering whether any of this actually matters. After all, it’s just a few big companies moving money in circles, nothing to do with any of us. We’d argue it matters immensely, for three reasons.
All of the risk eventually gets dumped onto retail. Microsoft is the 4th largest company in the world by market capitalization, with their stock price largely dependent on whether data centers get filled in a timely fashion. OpenAI just filed to go public later this year, despite the fact that they continue to set money on fire. Both are and will be in most of the index funds that prop up 401ks and retirement plans. Helion will have the option to do the same if Orion opens on time and as advertised. Taking money from VCs, who are professional investors and aware of the risks, is one thing. Lying to the public in order to grab cash and walk away from bad tech is quite another.
The US fusion sector has raised around $8bn to date, despite the fact that precisely zero of the startups in question have produced net electricity. This is where a material part of the capital targeted to climate solutions is going and, we would argue, being wasted on grifts and clownery. The tag line on Helion’s website is “Fusion to power our world.” Even if we take the representations of their tech at face value, their first two customers are a data center and a steel plant. Whose world, exactly?
We could fund so much other great tech if even a quarter of the capital spent on fusion was run through an actual process of scientific validation and impact analysis, rather than VC return models. Do you know how many solar panels you can produce for a billion dollars?
We don’t regret much that we’ve written in these posts. We get plenty wrong, and try to be honest and aware about it, but that’s just a function of learning loudly and in public. We do regret writing, almost three years ago, that “the answer is probably just to get fusion working, plug in massive DAC plants next to the reactors, and move on with our lives.” That was bad, and ignorant. The real answer is, always has been, and always will be renewables, batteries, and the hard work of systematic sector-by-sector electrification.
Maybe the real fusion was the friends we made along the way.
Coral Carbon is building the last greentech enterprise. Please follow, send to your smartest friends, and consider taking out a modestly priced subscription to this publication if you have the means. Thanks for being here.
https://www.coralcarbon.io/