Sort the announcement into two piles and the diligence gets straightforward: what the hardware did, and what a spreadsheet said hardware would do. The measured pile holds one result, and the company's framing of it is the part worth reading. At a plasma beta of one, American Fusion equates that pressure to roughly 160 Tesla, generated dynamically during the pulse instead of by a massive static coil [5]. That equivalence is the whole basis for calling this a modular family with configurations from about 1 MW to 500 MW [2], since a machine that never needs a static coil can in principle be built small.
The other pile is where the megawatts live. Both chamber models assume the same plasma condition, 60 keV and one microsecond of confinement at the pressure already reached [6]. The gigawatt figure attached to the smaller design is 3.9 kilojoules divided by a millionth of a second [6]. Release any modest amount of energy in a microsecond and the arithmetic comes out in gigawatts.
What decides whether there is a product is the per-second number. At 3.9 kJ of gross thermal per shot, a 500 kW module has to fire about 128 times a second, which is 500,000 watts over 3,900 joules [1]. The 5 MW design needs about 153 shots a second at 32.6 kJ [2]. Both figures sit before conversion losses and before whatever the drive circuit itself draws. The release calls the compression repeatable and puts no rate on it [7].
The second chamber size is arithmetic rather than evidence. Both modeled designs run at the same ion density, about 4.2 x 10^17 ions per cubic centimetre, from 1.88 x 10^20 ions in 445 cubic centimetres and 1.57 x 10^21 in 3,720 [4]. The 8.4x jump in yield is 8.4x the volume [5]. That is honest modeling, and it tells a reader nothing the smaller case did not already tell them.
John Brandenburg, the chief technology officer who led the work, says fusion is determined by the complete plasma condition rather than by any single measurement [12]. Read that as an accurate description of what is still outstanding.
Two axes will carry any pulsed-fusion claim that lands on your desk. Measured against modeled, and per pulse against per second. Compression sits in measured and per pulse. Yield sits in modeled and per pulse. The quadrant where a hardware product actually exists, measured and per second, is empty here, which is a fine place for a physics program to be and a poor place for a purchase order. This result is for investors and plasma physicists; the person who would specify a 1 MW unit can wait for a shot rate.
