The number to sit with is 0.04 nano-ohms.
That is the internal joint resistance the team at the Chinese Academy of Sciences' Institute of Plasma Physics drove into the superconducting cables of the world's largest fusion magnet—effectively zero loss at currents around 100 kiloamperes, holding up across more than a thousand electromagnetic and thermal cycles at 4.2 Kelvin while still beating design values.
That is what a hard engineering problem looks like once it has been beaten.
Meanwhile, in the Idaho desert, a squad of twenty-somethings from a two-year-old, twelve-person startup reportedly signed a reactor vessel—arranging their signatures into the shape of an American flag—after one such startup claimed, by its own account, to be "the first startup in history to generate nuclear electricity and power a chip."
Two milestones, both filed this week under "the nuclear race." They are not measuring the same thing. They are not even close.
Last week this newsletter tracked the fuel-cycle bottleneck—the tritium supply crunch, the quiet ITER schedule slip. This week the story pivots to hardware: the component that just cleared verification, and the licensing sprint that just cleared a deadline. It is worth being precise about what each one actually proved.
Start in Hefei. The toroidal-field magnet accepted on June 27 is a D-shaped structure 21 meters long, 12 meters wide, weighing 582 tonnes—1.3 times the volume and three times the stored energy of the comparable magnet at ITER, the international project in France.
Sixteen of these coils will eventually form a ring generating a 6.5-tesla field at the plasma center. The job, as ASIPP researcher Wu Yu put it, is to keep a 100-million-degree "fireball" suspended inside the vacuum chamber without touching the walls—"an invisible, extremely sturdy magnetic cage."
The magnet is part of CRAFT, the Comprehensive Research Facility for Fusion Technology. And here is the first thing the coverage tends to blur: CRAFT does not generate electricity. Not a watt. It is a component-proving facility—a bridge that verifies hardware survives real-world stress before it goes into the machines that will actually try to make power.
What got retired this week was a genuine materials-science bottleneck. During heat treatment of the niobium-tin superconductors, a temperature deviation of just a few degrees can cause performance to collapse—and keeping a furnace uniform across coils more than 10 meters wide is itself a frontier engineering feat.
Then there is the point Wu made about the joint resistance: "It's not about passing a single run, it's about staying reliable after being put through the wringer again and again." These magnets must operate for 60 years at minus 268.95°C under high current, radiation, and mechanical load.
Alongside the toroidal magnet, the team passed full-parameter tests on a high-temperature superconducting central solenoid—the "spark plug," in Qin Jinggang's phrase, that induces and drives the plasma current. Rated for 46.5 kiloamperes, it was, per Nuclear Engineering International, tested at 60.
The economics moved too. The cost of high-temperature superconducting tape fell from 400 yuan (about $56) per meter to 100 yuan, and the whole supply chain—tapes, cryogenic steel, insulation—is now 100 percent domestic, across 47 patents and 25 standards.
That is a real achievement, and it deserves to be called one without hedging.
But here is the caveat the triumphant coverage buries. Qin Jinggang, who led the work, says plainly that this is "just 80% of the journey." The remaining 20 percent is the hard part: installing the coil in the device and verifying its long-term stability over that 60-year life.
CRAFT feeds BEST—the Burning Plasma Experimental Superconducting Tokamak—slated to finish assembly by 2027 and designed to demonstrate actual net-energy "burning plasma," a fivefold energy gain where the reaction sustains its own heat. Grid electricity is targeted around 2030, with a demonstration reactor, CFETR, aiming for first grid power by 2035.
Read that timeline carefully. The thing everyone actually cares about—more energy out than in, at scale—has not happened. This week reportedly verified a piece of the cage, not the fire inside it.
The thing everyone actually cares about—more energy out than in, at scale—has not happened. This week reportedly verified a piece of the cage, not the fire inside it.
Now cross the Pacific.
Executive Order 14301 set a goal: three new reactors critical by July 4, 2026. "We delivered four," the celebrants wrote. "This is proof that, with the right mindset and smarter regulations, nuclear can move fast."
And that is true—about a specific thing. What EO 14301 demonstrated is regulatory velocity. A two-year-old company with twelve employees went critical and powered a chip. That is a licensing-and-engineering sprint, and licensing sprints are not nothing; permitting has strangled Western nuclear for decades.
But notice the unit of achievement. The startup "powered a chip." One chip. That is a symbolic demonstration—a flag-signing on Independence Day—not a megawatt figure feeding a grid.
The tell is in what nobody quotes. There is no tesla figure, no net-gain number, no sustained-plasma record attached to the July 4 blitz. What is celebrated is the speed of the paperwork and the audacity of the team, which is a story about institutions and spirit, not about physics.
This is the conflation to resist: the idea that because deregulation made American reactors move fast, America is winning a race whose finish line is net energy. Those are different measurements. One clock times permits. The other times the point at which fusion produces more than it consumes—and on that clock, neither country has stopped it.
Look at where the honest numbers are versus where the money-talk is. China's own scorecard is engineering: 0.04 nano-ohms, 6.5 tesla, 1,066 seconds of plasma at 100 million degrees, 60 kiloamperes on a coil rated for 46.5. Concrete, verifiable, and explicitly "80% of the journey."
Set against that, the market rhetoric: a projected 5.2-trillion-yuan fusion market by 2050, a "trillion-yuan track," brokerages describing a "triple resonance" of policy, breakthrough, and capital. The trillions are running far ahead of the teslas.
And the American celebration substitutes narrative—"the indomitable American spirit," "we build stuff, we dream bigger"—for the load-bearing quantity. Spirit does not close the net-energy gap. Neither does a faster clock at the licensing office.
So here is the scorecard, stated flatly.
China verified a 60-year-service-life component under brutal full-parameter stress. That retires a bottleneck. It does not produce electricity.
The United States compressed the permitting timeline enough to bring four reactors critical on a deadline. That proves regulation can move. It does not prove net gain either.
Both are real. Both are necessary. Neither is the thing.
The load-bearing question in fusion has not changed since last week or last decade: demonstrated net energy at scale, sustained, and delivered to a grid. That answer is not due from Hefei until BEST reports out around 2027, with power years after. And it did not arrive in Idaho on July 4, no matter how the vessel was signed.
A magnet that survives a thousand cycles and a chip that lights up on a holiday are both worth building. Just don't mistake either one for the finish line—or let the trillion-yuan track outrun the physics that hasn't crossed it yet.