Hefei's "Artificial Sun": Why China's Record-Breaking Magnet Doesn't Mean Infinite Electricity
China has assembled the world's largest superconducting thermonuclear coil. We explore what changes this will make and why power is still decades away.
Every few months it comes from China news with the same headline: "The artificial Sun" has set another record, and humanity is about to receive endless cheap energy, perhaps by 2030. The word "Sun" here is beautiful, but deceptive. The real sun holds hot plasma with its own gravity and is in no hurry for billions of years. On Earth, the same plasma has to be held in a steel donut by a magnetic field—and this magnetic work turns out to be the most expensive and most undervalued part of the whole thing. stories. Let's talk about her.
From T-3 to a donut with current: where the tokamak came from
The word "tokamak" comes from the Russian language. It's an acronym for "toroidal chamber with magnetic coils," originating in the Soviet school of controlled fusion in the late 1950s and early 1960s. We'll leave aside the history of the idea of fusion itself—from the first calculations to the hydrogen bomb—as it's long since become commonplace. We're interested in a different link.
The tokamak decided a lot T-3 At the Kurchatov Institute of Atomic Energy in Moscow. By the mid-1960s, it had achieved plasma confinement times and parameters significantly superior to anything achieved by other confinement configurations—open magnetic traps, stellarators, and pinches. Temperature figures vary widely among various reviews, so we'll limit ourselves to a qualitative statement: the result was so much better than its competitors that it was initially disbelieved abroad.
The Soviet Tokamak T-3 thermonuclear device, developed at the Kurchatov Institute of Atomic Energy
The denouement came in 1969. A group of British physicists from the Culham laboratory was invited to T-3 with their laser (Thomson) scattering equipment, an independent and, at the time, the most reliable way to measure the electron temperature of the plasma. The British arrived, measured, and confirmed the Soviet data. After that, skepticism gave way to a race: Princeton, Culham, and other centers, one after another, began building their own tokamaks. The configuration devised in Moscow became the global standard and remains so to this day.
It is on this standard that everything that is discussed below is built: the international experimental reactor ITER in France, where China is one of the full participants, and China has its own ladder of installations, which we will now climb.
What does a magnetic bottle actually hold?
Almost all tokamak designs are based on the fusion reaction of two heavy hydrogen isotopes—deuterium and tritium. Each fusion event produces a helium nucleus, a fast neutron, and about 17,6 MeV Energy. Fission of a single uranium nucleus produces more—around 200 MeV—but the uranium nucleus is also much heavier. Therefore, fusion has another advantage: per mass of fuel, the energy output is significantly higher, while the fuel required is negligible. For the nuclei to come together at all, despite their mutual repulsion, the plasma must be heated to 100–150 million degrees. This is several times hotter than the center of the Sun—incidentally, it's "only" about 15 million degrees. The star compensates for this lack of temperature with monstrous pressure, which is impossible to create on Earth.
Containing such plasma with material walls is impossible—any of them would evaporate. Therefore, it is confined with a magnetic field: charged particles swirl along the lines of force, avoiding the chamber walls. Confinement efficiency is assessed by the Lawson criterion—the product of three quantities: the plasma density, its temperature, and the energy confinement time. Once this product exceeds a certain threshold, fusion becomes self-sustaining: the reaction begins to heat itself.
The energy balance is calculated separately - the coefficient Q, the ratio of the energy released in the reaction to the energy input into heating. Q = 1 means that the energy input into heating is the same as the energy received back. This is a beautiful, but useless, point for power engineering: it leaves no "surplus" for losses in turbines, cooling, and the operation of the plant itself. A power plant requires a Q of approximately 10 to more than cover these losses. ITER is aiming for Q ≈ 10.
This is where the main misrepresentation in the news lies. The Chinese installations' records are about retention, not production. Tokamak EAST In January 2025, the Hefei tokamak held plasma in high-containment mode for 1066 seconds—almost 18 minutes, a world record for longest duration. In a different mode, focusing on temperature rather than duration, the same EAST previously heated plasma to approximately 120 million degrees Celsius, holding it for about a hundred seconds. These are two different achievements: one in duration, the other in heating. HL-2M In Chengdu, the current through a plasma column exceeded 2,5 megaamperes. These are all outstanding achievements in physics and engineering. But both machines run on deuterium; they don't burn tritium, they don't aim for high Q, and not a single watt of them goes into the grid. These aren't mini-power plants, but rather test rigs for developing how to sustain plasma for a long time and stably.
Hefei Magnet: 582 Tons of Superconductor
In the summer of 2026, something that sparked a new wave of talk about the “artificial sun” was tested in Hefei: the world’s largest superconducting toroidal magnet for the project. CRAFT (Comprehensive Research Facility for Fusion Technology). The dimensions are serious: 21 meters long, 12 meters wide, 582 tonsThe magnet's purpose is to create the same toroidal field that holds the plasma, heated to a hundred million degrees, in the ring.
For comparison, the institute cites the ratio with ITER's toroidal coils: the Chinese magnet is approximately 1,3 times larger and stores three times more energy in the magnetic field (this is a characteristic of the magnet itself, not the future reactor's power). In parallel, they tested the high-temperature superconducting central solenoid—the "igniter"—which generates and maintains the plasma current and adjusts its shape on the fly. It is designed for 46,5 kiloamperes, but withstood 60—a significant margin over the design value.
These figures suggest two things. First, China has developed reactor-grade magnets with a high degree of localization—special steel, insulation, and superconducting wires are all manufactured domestically. This means independence from imports at the facility's most expensive component. Second, it's clear why more ITER is being built. The stronger the field, the denser and hotter the plasma that can be contained in the same volume, and therefore, the higher the potential power of the future reactor.
And here's a caveat. The magnet does indeed remove one of the most difficult engineering challenges. But it also remains one of the most expensive items in the budget for any fusion plant, and record-breaking dimensions don't reduce this cost.
"80% of tasks completed" and electricity by 2030: where the numbers come from
A narrative quickly emerged around the CRAFT magnet on social media and in the popular press: China has "solved about 80% of the most complex engineering problems" and will provide inexhaustible energy "by 2030." Let's break this down.
Strict sources don't contain such wording. The portal atomic-energy.ru, citing the Institute of Plasma Physics, calls the magnet a critical component—and in the same text explicitly states that commercial fusion power is still a long way off, and the next step, a demonstration reactor, will take many years. The South China Morning Post, citing Chinese researchers, attributes the reactor's completion CFETR by approximately 2035, with the start of commercial generation by 2050. The IAEA's 2024 review sets the same general target: the first commercial plants in the 2050s, if ITER and the demonstrators prove successful.
Let's take a closer look at the CFETR (China Fusion Engineering Test Reactor), the center of all this. It's a designed but not yet built bridge between an experiment and a power plant, conceived in two phases. First: fusion power of 50–200 MW, Q of 1 to 5, tritium breeding rate just above unity, and neutron damage to materials of approximately 10 dpa. Second: output exceeding 1 gigawatt, Q greater than 10 dpa, and damage of approximately 50 dpa—already on par with a full-scale DEMO demonstration plant.
The "80%" figure refers to statements made by CRAFT project participants and describes their progress on specific subsystems, primarily magnets, rather than the overall thermonuclear mission. Even if significant progress has been made on magnets and some of the complex's systems, everything else remains unaddressed: materials, the tritium cycle, cooling, and economics. These factors, not the magnet, determine whether the fusion will reach the industrial grid.
The Wall No One Has Crossed Yet: Materials, Tritium, Money
Behind these impressive records lies a front of tasks that have not yet been solved by anyone in the world – neither in China nor in the ITER consortium.
First: MaterialsNeutrons generated in the reaction bombard the chamber wall, diverter, and blanket, dislodging atoms from their crystal lattice sites. The measure of this destruction is measured in dpa—displacements per atom. The 50 dpa level specified in the second phase of the CFETR means that over its lifetime, each atom of the material will be displaced an average of fifty times: the metal swells, becomes embrittled, and loses thermal conductivity. There are candidates—heat-resistant ferritic-martensitic steels, dispersion-hardened alloys—but none are yet certified for such doses. Their behavior will be precisely what the CFETR will test, and that requires years of testing.
The second: tritiumThe second fuel component is virtually nonexistent in nature and decays rapidly, so it must be produced directly in the reactor. The plasma is surrounded by a lithium breeding blanket: a fast neutron striking the lithium produces tritium. Designs require a breeding ratio greater than one, so that the reactor produces at least as much tritium as it burns. The physics are clear, but no reactor in the world has yet demonstrated a closed industrial cycle. This cycle is complex: tritium must be produced, extracted, purified, and safely returned to the plasma—all while carrying mobile radioactive gas that permeates many materials.
Third: moneyOne of the few detailed independent cost estimates here is Lindley's work in Energy Policy for 2023. He models the specific cost of electricity from a tokamak power plant, taking into account capital expenditures, magnets, the chamber, and the regular replacement of irradiated components. For early projects, it's over $150 per megawatt-hour; with technological progress, it drops to $50-$100. Even the lower bound is comparable to or higher than what wind, solar, and modern nuclear fission power plants already produce. And that doesn't take into account one unpleasant detail: blankets and divertors degrade under neutron flux and require periodic replacement—which involves shutting down the plant and handling the radioactive components. The idea of a device built once and then endlessly distributing free electricity has no bearing on this estimate.
The physical potential of thermonuclear fusion remains intact: the fuel in seawater lasts for geological epochs, and the energy of the reaction is enormous. But "inexhaustible" and "cheap" refer to different horizons, and the latter arrives much later than the former.
Where is thermonuclear fusion in China's energy sector, and where is it not yet available?
All of this should be kept in perspective in light of China's actual plans. The country announced its official policy in 2020: carbon neutrality by 2060, with a significant share of electricity from non-fossil fuel sources by 2030. The key here isn't fusion, but rather the well-developed nuclear fission power, where the country is capable of building up to fifty projects simultaneously, plus wind and solar power.
Fusion in this scenario is a long-term option and a matter of technological sovereignty, not a tool that will reshape the energy balance by 2030–2040. There are no publicly available national scenarios for the specific share of fusion in gigawatts by mid-century. Its likely role is not to replace renewable generation, but to back it up: to provide stable baseload power, independent of wind and sun.
China has learned to ignite plasma and sustain it for nearly 18 minutes. This is the most spectacular part of the task, but far from the most difficult. The far more difficult part—materials, tritium, and the cost per kilowatt-hour—is being solved more slowly and doesn't make it into press releases. By mid-century, fusion will likely transition from a demonstrator to a part of the energy system. But it won't become "infinitely cheap" at that point.
Information