Fusion Power has the potential to rewrite trillion-dollar energy markets, but first, startups must prove their plans will work and not be too costly. Neither is easy, especially when you consider that the huge magnets and lasers used in many designs must be installed with millimeter or better precision.
Launch Fusion Thea Energy says the pixel-inspired reactor and its specialized control software should be able to generate power without requiring the same level of perfection.
“It doesn’t have to be that good to begin with,” Brian Berzin, co-founder and CEO of Thea Energy, told TechCrunch. “We have a way of smoothing out the imperfections in the back.” That margin for error could give Thea a leg up on the competition.
Fusion power plants promise to deliver gigawatts of clean energy to the grid, but material and construction costs threaten to make them uncompetitive with cheap solar and wind power. By first building a power plant and eliminating the kinks in the software, Thea could help reduce the cost of fusion power dramatically.
But first the company must build a working prototype. Today, Thea publishes the details of her design, including details of the physics behind it. The startup shared the paper exclusively with TechCrunch.
Thea builds a unique concept for the stellarator, a specific type of reactor that uses magnets to form the plasma fuel. Magnets are one of the two main ways that fusion scientists retain the heat of the plasma and confine the plasma until fusion reactions occur. The other, known as inertial confinement, uses a laser or some other force to compress small pellets of fuel.
Most stellarators are made with magnets that would look at home in a Salvador Dali painting. But Thea’s design uses a dozen larger magnets and hundreds of smaller ones to create what you might call a “virtual” astral.
Techcrunch event
San Francisco
|
13-15 October 2026
On a typical asteroid, the magnets are made to follow the contours of a shape meant to work with the quirks of the plasma, helping to confine it for longer using less power than tokamaks, which use an array of magnets of the same size and shape. However, stellarators have one major drawback: the irregular shape makes mass-producing magnets difficult.
Instead, Thea designed her reactor around small, identical superconducting magnets arranged in arrays. The startup will use software to control each magnet individually to create magnetic fields that can replicate the oscillating shape of a star.
The approach has several positives. First, it allowed Thea to quickly iterate on the magnet design. Over the past two years, the company has modified the design more than 60 times, Berzin said. “Most fusion companies, you’re dealing with magnets that are the size of a car or lasers the size of a car or a widget the size of a car. That unfortunately means one is $20 million and takes two years [to make]”, he said.
It also means the company can use software controls to overcome any anomalies in the way the magnets are manufactured or installed. To test her initial control system, Thea built a series of three-by-three magnets with sensor bonding. The controls, which were derived from the physics of electromagnetism, worked well. But the company also wanted to see how AI could handle the task, so it also trained a new one using reinforcement learning.
The team was amazed at how well everything worked.
“We intentionally threw curveballs in the lineup,” Berzin said. “We deliberately lowered a magnet by literally over a centimeter. You could see it was extremely out of line. It was really hard for us to build it that badly.” The team also tested superconducting material from five different manufacturers along with intentionally defective material. “Every time we did this, the control system, without us turning the knobs and intervening, was able to tune out these defects,” he said.
Thea’s reactor design, Helios, will use two types of magnets. On the outside, 12 large magnets of four different shapes will do the heavy lifting to keep the creature contained. They are similar to those found in a tokamak, the type of donut-shaped reactor made by competitor Commonwealth Fusion Systems. Inside the large coils, 324 smaller circular magnets will regulate the shape of the plasma.
The startup envisions Helios generating 1.1 gigawatts of heat, which a steam turbine will convert into 390 megawatts of electricity at a cost of less than $150 per megawatt hour for the first example. (Thea thinks it can bring the cost down to $60 per megawatt hour after the seventh or tenth power plant is built.) The reactor would have to shut down for an 84-day maintenance period once every two years. If all goes well, that means its capacity factor – a measure of how much power it produces in a given period of time – will be 88%. This is it much better than today’s gas-fired power plants and almost as good as today’s nuclear plants.
Helios is still in the conceptual phase. Thea must first build Eos, her initial fusion device meant to prove the science behind the concept. Berzin said the company will announce a website for Eos in 2026 with plans to activate it “around 2030.”
While building Eos, Thea plans to start work on Helios in parallel. It’s a similar approach to how Commonwealth Fusion Systems is proceeding with work on Arc, its first commercial power plant, while building Sparc, the demonstration plant.
For now, Berzin is eager to hear from the fusion community. “This is the release of the review paper. There will be a very significant body of work coming out through peer review and publication,” he said. “Now is the time to go and build the partnerships, the partnerships, engage the end users to go build the first one.”
Update 2:45 p.m. ET: Added information on expected electricity costs for Helios.
