Helion Energy just did something no private fusion company had ever done. In January 2026 its seventh prototype, named Polaris, fired its first shot using deuterium-tritium fuel. The plasma temperature hit roughly 150 million degrees Celsius — about ten times the temperature at the core of the Sun — on the first try. By mid-February, Helion had announced measurable D-T fusion yield, becoming the first non-government machine to cross that line.
The headlines focused on the temperature. The temperature is not the story.
Every credible fusion approach reaches plasma temperatures above 100 million °C eventually. ITER will. SPARC will. The National Ignition Facility crossed it years ago using lasers. What separates Helion from the rest of the field is not how hot Polaris runs. It is the architecture of the machine — and the architecture is a direct bet that the entire steam-turbine-condenser infrastructure that has powered every conventional nuclear plant since 1957 can be skipped.
If Helion is right, fusion electricity costs collapse below anything the tokamak roadmap is pricing today. If Helion is wrong, the company still has tens of millions of dollars of capacitor banks to repurpose. Either way, the next 24 months are the most important in private fusion.
Why the temperature is not the metric that matters
The number every fusion engineer actually cares about is the Lawson triple product — plasma density times confinement time times temperature. You can hit 150 million °C with a desktop tokamak if you only need to hold it for a microsecond. The hard part is keeping the plasma dense and confined long enough to extract net energy. The triple product is what separates a science demonstrator from a power plant.
Tokamaks like Commonwealth Fusion Systems’ SPARC chase the triple product through extreme confinement: high-temperature superconducting REBCO magnets pull plasma into a torus and hold it steady for seconds at a time, then minutes, then continuously. SPARC’s design target is Q ≈ 11, meaning 50–100 MW of fusion power from 25 MW of input heating, with first plasma planned for late 2026 and net-energy operations in 2027. That is the conventional path, and it works.
Helion’s machine does not look like that. Polaris is a 19-meter cylinder with 50 megajoules of pulsed capacitor energy firing into two opposing field-reversed configuration (FRC) plasmas. The plasmas are formed at each end of the device, accelerated toward each other at high velocity, and slammed together in the middle inside a converging magnetic compression field. When the two FRCs merge and compress, the fuel reaches fusion conditions for a few microseconds. Then the cycle repeats.
This is magneto-inertial fusion — somewhere between the steady-state confinement of a tokamak and the single-shot implosion of laser inertial confinement. It is not the dominant paradigm. It has the cleanest economics if it works.
The architecture nobody talks about: direct energy conversion
Here is the part of the Helion design that quietly rewrites the cost curve.
Every fission reactor operating in the United States today, and every tokamak under construction, runs through the same energy conversion path: nuclear reaction → hot neutrons → heat exchanger → steam → turbine → generator → grid. That stack carries a hard ceiling. The Carnot efficiency of a steam Rankine cycle tops out around 35–40% in commercial practice. Two-thirds of the thermal energy a reactor produces never reaches the grid as electricity. It goes into cooling towers.
Helion’s FRC architecture targets a direct electromagnetic conversion path. When the merging plasmas compress against the external magnetic field, the expanding fusion plasma pushes back outward, inducing a current in the surrounding coils. That induced current is electricity — extracted directly from the plasma’s motion, no steam, no turbine, no thermal cycle.
The theoretical efficiency of magnetic direct conversion runs above 60%. Helion’s published architecture targets roughly 40–60% end-to-end, depending on the fuel cycle and pulse rate. If they hit even the low end of that range at commercial scale, fusion electricity bypasses one of the structural cost drivers that has held back conventional nuclear for fifty years.
That is the bet. Not \”fusion works.\” Specifically: the magnetic recovery loop is real enough to skip the entire thermal plant.
What Polaris actually proved in January
The D-T shots in early 2026 are not a demonstration of commercial direct energy conversion. Polaris is a physics machine. What it demonstrated, with measurable yield:
- Plasma formation and acceleration in the FRC geometry at production-relevant velocities
- Compression heating correlated with magnetic field strength in the way the FRC scaling laws predict
- Temperature above the fusion threshold for D-T, with the cross-section peaking near 100 million °C
- Operation under an NRC tritium license — Helion is the only private company in the United States authorized to possess and use tritium for fusion R&D
What Polaris has not yet demonstrated, but is designed to test through 2026: net electricity production, pulse repetition at commercial cadence, and the full direct energy conversion loop at meaningful power.
Why Microsoft signed the PPA
In May 2023 Helion announced a power purchase agreement with Microsoft to deliver 50 megawatts of fusion electricity to the grid by 2028, with a ramp window of approximately one year. Constellation Energy is the marketer of record. This is a real PPA — with financial obligations on both sides — for an electricity source that did not exist when the contract was signed.
Microsoft did not sign this because they believe fusion is finished. They signed it because their AI data center load growth curve is steeper than any utility’s interconnection queue can serve, and a 50 MW firm clean baseload contract has option value far above what they pay if Helion misses the date.
The same calculus is now driving the rest of hyperscale procurement. Amazon’s Energy Northwest deal with X-energy for up to 12 Xe-100 SMRs. Meta’s January 2026 announcement of 6.6 GW of nuclear partnership commitments. Google’s Kairos Power agreement enabling up to 500 MW by 2035. Microsoft’s restart of Three Mile Island Unit 1 with Constellation. The pattern is unambiguous: AI infrastructure capex is now the marginal buyer of advanced nuclear and fusion electricity in the United States.
The fuel question — and the helium-3 endgame
Polaris runs D-T because D-T is the easiest D-anything reaction to ignite. The cross-section peaks near the temperature Polaris just hit. Every fusion machine that wants a near-term physics result runs D-T first.
Helion’s commercial fuel cycle is different. The company plans to operate its production machines on deuterium-helium-3 (D-He³). The advantage of D-He³ is that the primary reaction produces a high-energy proton — not a 14.1 MeV neutron — which the magnetic direct conversion system can capture far more efficiently than neutron heat. Less neutron load also means less material activation, less shielding, longer plant life, and far fewer waste disposal headaches.
The catch: helium-3 is rare. Helion’s plan is to produce its own helium-3 on-site by running deuterium-deuterium side reactions, harvesting the resulting tritium, allowing it to decay to helium-3, and feeding it back into the primary fuel stream. The chemistry is straightforward; the engineering of running it at industrial scale is not. It is the highest-risk piece of the commercial design.
Competitive landscape — honest read
Commonwealth Fusion Systems is the most advanced tokamak program. SPARC is 75% complete as of early 2026, with first plasma scheduled before year-end. CFS has raised roughly $3 billion to date. Their target is conventional: Q > 1 in 2027, then ARC as the first commercial machine in the early 2030s. SPARC will produce neutron heat for steam.
TAE Technologies, Zap Energy, Type One Energy, General Fusion are all pursuing alternate confinement geometries (FRC variants, sheared-flow Z-pinch, stellarator, magnetized target). All are smaller and younger than Helion in funding terms.
Helion sits at roughly $1 billion raised through Series F at a $5.425 billion valuation (January 2025), with Sam Altman as board chair and SoftBank Vision Fund 2 leading the recent round. Their public roadmap is the most aggressive: 50 MW on the grid in 2028 from a facility already under construction in Eastern Washington.
The fair read: SPARC will almost certainly produce more measured fusion energy than Polaris during 2027. Helion will almost certainly be the first private machine to put fusion-derived electricity on a real grid. Those are not the same race.
What this means for the AI power problem
The U.S. electricity demand growth curve has bent upward faster than any utility planner forecast as recently as 2022. EPRI’s most recent load growth analysis assumes data center demand grows from roughly 4% of U.S. electricity in 2023 to 9–12% by 2030, depending on AI training intensity. That growth is concentrated in five or six grid regions and it is happening faster than any combination of natural gas peakers and renewables can serve.
The structural answer is firm, dense, clean baseload close to load. Advanced fission (Xe-100, Natrium, Hermes) is the near-term answer for 2028–2032. Fusion is the answer for the 2030s if any of the leading designs commercialize on the timeline they publish.
Helion’s 50 MW is a small number against a multi-gigawatt problem. It is a very large number as proof of architecture. If Polaris produces measurable net electricity at any point in 2026, every hyperscaler procurement team in North America rewrites its 2030s baseload assumption.
What to watch next
- Polaris net electricity demonstration during the 2026 campaign — the inflection point for the entire FRC + direct conversion approach
- Commercial plant first plasma at the Eastern Washington site, currently targeted for 2028
- D-He³ closed fuel loop demonstrated at any scale — the long pole of the commercial design
Watch capacitor pulse rate, not peak temperature. Watch sustained yield over a campaign, not single-shot records. Watch the regulatory file at the NRC — every advanced fusion plant in the U.S. will need a path through Part 53 or its successor, and Helion’s tritium license is the leading-edge case.
For the Porritt Inc. audience
If you are running a process safety, engineering, or industrial AI program at a refinery, a chemical plant, or an advanced reactor developer, the Helion story matters for a specific reason. The next decade of U.S. industrial infrastructure — fission SMRs, fusion plants, hydrogen electrolyzers, GW-scale data centers — will be built and commissioned faster than the conventional PSM-compliance, hazard analysis, and operator training pipeline can absorb. The bottleneck is not capital. It is documented, defensible, operator-ready compliance.
Porritt Inc. builds AI infrastructure for that bottleneck. NORMEX Standards AI, NEXUS Compliance AI, and NEXUS CAD are designed to compress the gap between FEED-complete and operator-trained for next-generation industrial facilities. If your team is staring at an advanced-reactor or AI-data-center buildout calendar and the PSM and operator-readiness side has not been resourced, we should talk.
Schedule a 20-minute consultation →
Timothy Porritt is founder of Porritt Inc., building AI-powered tools for process safety, engineering compliance, and industrial operations. Based in Salt Lake City, Utah.