Shot 5,319.
That is the experiment number Energy Singularity attached to the result it announced on January 7: a 120-second steady-state long-pulse plasma run inside Honghuang 70, described in the company's release as the world's first fully high-temperature superconducting tokamak.
Last week this column followed China's superconducting magnet work and the engineering path it implies — cable-in-conduit conductors, radiation-resistant joints, millisecond quench detection, cryogenic thermal management. This week the story moved from hardware to operations: a long plasma shot from a commercial device, and, separately, anomalous heat and gas signals from a lattice-confinement reactor tested at Texas A&M.
Those two items are not the same kind of claim. Bundling them under one word — progress — is the week's real error.
Those two items are not the same kind of claim. Bundling them under one word — progress — is the week's real error.
Start with what Honghuang 70 actually shows. A tokamak that holds a plasma for 120 seconds is a tokamak whose magnets, power supplies, vacuum systems, and control loops all function together for two minutes without a disruption. That is not trivial. Energy Singularity, founded in 2021 and backed by miHoYo, says the run validates the engineering reliability and operating stability of the high-temperature superconducting route, and that it will next deploy AI-based real-time plasma control.
Read the announcement carefully, though, and it contains no energy gain claim, no neutron budget, no tritium accounting, no heat balance. The plasma current was sustained. That is a control milestone, and control milestones matter — they are the difference between a physics idea and a machine that runs. They are not evidence that fusion produced net power, and the company does not claim they are.
For scale, this is a private machine joining a public field's long-pulse work. China's EAST holds the record at 1,066 seconds of steady-state high-confinement operation at 100 million degrees Celsius, set in January 2025 and celebrated by the Institute of Plasma Physics in Hefei. The international CICLOP database, coordinated by the IEA and IAEA, now compiles long-duration data from ten tokamaks and two stellarators — ASDEX Upgrade, DIII-D, EAST, JET, KSTAR, TCV, WEST, W7-X and others — precisely because duration and fusion performance must be raised together, not one at a time.
Now the second item, which belongs on a different shelf entirely. Hylenr Inc., a Michigan company, completed Phase 1 of an independent study of its BRT-NiUCS-2 reactor and multiple hydrogen-loaded nickel-palladium catalyst samples at Texas A&M University.
The study detected elevated helium, argon, and neon signals in the active reactor relative to background — helium and argon roughly two to three orders of magnitude above it. Researchers also found temperature differences between the active reactor and a calibration device operating under comparable input power, using thermocouples and calibrated infrared imaging. Post-run SEM/EDX showed morphological and compositional changes in the catalyst.
Residual gas analysis ran on an SRS RGA 100 under high vacuum. The helium and argon rise came with no corresponding nitrogen increase, which the study says argues against atmospheric leakage as the sole explanation.
Then the part that matters most. Geiger-Müller and neutron detectors found no detectable gamma or X-ray emissions, and neutron counts stayed statistically indistinguishable from background across roughly five days of monitoring.
That is the crux. Deuterium-deuterium fusion produces neutrons. Helium appearing without a commensurate neutron signal is a puzzle, not a demonstration. It could be gettering, desorption, or contamination that the current diagnostics cannot exclude. The company itself says the results do not establish commercially useful fusion energy.
"We do not establish" is doing honest work in that sentence, and so is the Phase 2 plan: multiple independent reactors, quantitative calorimetry, hydrogen-loading parameter studies, and isotopic-ratio measurements, with secondary ion mass spectrometry and inductively coupled plasma mass spectrometry on the materials.
Isotopic ratios are the right target. Excess heat claims live or die on whether the fuel's isotope distribution shifted in a way chemistry cannot produce. "Reproducing the reported signals across multiple reactors would provide a stronger basis," said Prof. Lin Shao of Texas A&M, which is exactly the standard the field should hold.
Meanwhile the money keeps arriving. Private fusion developers raised $1.8 billion in targeted funding over the past two quarters, largely to build high-field magnet testbeds, with capital shifting from theoretical modeling toward industrial-scale prototype validation. The sector still faces tritium breeding ratios and material durability under continuous neutron flux, and utilities are watching trial parameters to judge baseload viability against fission timelines.
That capital is real and useful. It is also ambient credibility, and ambient credibility is the laundering mechanism. A 120-second shot makes a headline; record fundraising makes a narrative; an anomalous-heat press release then rides the same wave without ever being asked for the two things that would settle it.
The field should name the tiers. Tier one is plasma operations: repeatable shots, independent diagnostics, published shot logs, machine parameters, disruption history. Honghuang 70's run belongs there, and it earns its place.
Tier two is heat. It requires isotopic ratios showing a nuclear, not chemical, origin; quantitative calorimetry that closes the energy balance; and replication across independent reactors by groups that did not build the device. Until then, temperature differences and gas signals are hypotheses with interesting data attached.
One reader of the Chinese coverage put it more bluntly than any analyst: 120 seconds is nothing, since it is not two months of continuous operation. That commenter understood the gap better than some press releases did.
Seven decades of "thirty years away" did not come from a shortage of milestones. It came from treating each milestone as if it were the next one. A long pulse is engineering control. Anomalous heat is an unverified claim. Keep them apart, and the week's news is genuinely good — a commercial superconducting tokamak ran a plasma for two minutes, and a company submitted its reactor to outside labs and published what the detectors did and did not see.
Let the record capital buy magnets. Do not let it buy conclusions.