Build2 distinct publishers3 min readPublished
Ground broke on August 25 on a $1 billion, 156-module pulsed machine aiming at net facility gain by 2030. The commercial gap is five times the output and 86,400 shots a day.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
A part count comes before an energy balance. Each of the 156 modules is 32 circular stages of 10 bricks, and each brick holds two capacitors and a switch [10]. That works out to 49,920 bricks in the finished driver [1], carrying 99,840 capacitors and 49,920 switches [2]. The one-third-scale pulser test that cleared the way to construction [13] demonstrated what a single unit can do; it did not demonstrate fifty thousand switches closing in step, which is what the design's requirement for precisely synchronised pulses actually asks for [24].
Divide the stored energy of one shot across the modules and each pulser owes about 513 kilojoules [3]. The same two figures Nature reported put the facility's gain at 1.25 [4], which is a thin margin to be building a billion dollars of hardware around, and is exactly why the 2030 date is a gate rather than a product.
Now the commercial number. Pacific Fusion's own framing, as reported by mezha.net, is that an industrial reactor must produce roughly five times the demonstration machine's energy to cover the auxiliary systems a real plant carries [23]. Five times 100 megajoules is 500 megajoules a shot [5], and 500 megajoules once a second is 500 megawatts of fusion heat [6]. What neither account settles is whether the driver grows with it. If stored energy scales at the same factor, a plant firing at one hertz has to push 400 megajoules back into its capacitors every second; if the driver stays at 80 megajoules, the recharge burden is 80 megawatts [8]. Those two readings describe different businesses.
The rate is the harder half. One shot per second is 86,400 shots a day, something like four orders of magnitude above the handful the demonstrator will manage [7], and the Albuquerque machine is only designed for the handful [14]. So nothing built on this site tests capacitor lifetime, switch wear or thermal recovery at duty. What it does test is whether the modular driver holds together and whether the target burns, and the architecture is arranged to answer those separately: the driver can be exercised without the chamber and fuel target in the loop [20], and the modules are steel, aluminum, plastic, oil and water, built away from site and serviced like industrial equipment [19].
The financing story has a loose thread. RuntimeWire, citing TechCrunch, describes a milestone-gated Series A of more than $900 million led by General Catalyst, structured so the founders hold committed capital without receiving all of it before the hardware works [7], and notes that a later total above $1 billion does not reconcile with that figure [8]. Mezha.net, working from the same TechCrunch reporting, states the milestone-gated Series A itself exceeded $1 billion [9]. The gating is the whole discipline claim here, so which of those is right determines how much money is still hostage to the 2030 result.
Meanwhile the site has a second buyer built in: high-yield experiments tied to national security [18], with chief technical officer Keith LeChien telling TechCrunch that China is building this class of facility and that Pacific Fusion means to finish first [17]. That is a hedge worth having when four other companies are pointing at the same middle of the 2030s for grid connection [21].
Ranked by verification strength, evidence, and original report placement.
Pacific Fusion co-founders Carrie von Muench and Keith LeChien broke ground in Albuquerque on August 25 on a fusion machine built around 156 identical pulser modules.
The $1 billion demonstration project is designed to produce more fusion energy than the electricity stored in its capacitors.
A commercial plant would still need about five times the output of the demonstration machine and a firing rate measured in shots per second instead of shots per day.
Pacific Fusion is targeting net facility gain by 2030, a milestone that accounts for the energy initially stored to drive the reaction and goes beyond experiments producing more energy than reaches the fuel, including the National Ignition Facility's 2022 ignition result.
Nature reported that Pacific Fusion expects each shot to use about 80 megajoules of stored electrical energy and produce about 100 megajoules of fusion energy.
Von Muench told TechCrunch: "It's a huge step forward for commercial fusion power."
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Concrete hardware numbers, one primary source
The physical and financial specifics are unusually granular and consistent across both publishers — 156 modules, 32 stages of 10 bricks, a 440-gigawatt one-third-scale test with 1,100 qualification shots, and Nature-sourced 80/100 megajoule shot figures. But both cluster items derive from the same TechCrunch primary reporting, the energy figures are company expectations rather than measured facility output, and the funding total is internally unreconciled.
Concrete poured, no fusion output yet
Real-world traction is limited to construction start, a subscale prototype campaign, and organizational scale-up. No fusion energy has been produced at the facility, no production-scale module has been built, no shot has been fired at the demonstration machine, and no grid connection exists anywhere in the field.
Milestone framing outruns the physics gap
Executive framing — 'a huge step forward for commercial fusion power', ten times NIF, nothing like it in the world, intent to beat China — sits against a demonstration design whose implied facility gain is only about 1.25, needs roughly five times more output and about four orders of magnitude more shots per day to become a plant, and whose one-hertz driver recharge power is never disclosed. Both publishers do carry the gap explicitly, which keeps the overstatement moderate rather than severe.
Milestone-gated capital and named executive sourcing
Nearly all substantive material originates with the company or its backers: co-founder quotes to TechCrunch, the company's own June technical update and expansion announcement, and General Catalyst's investment account. A milestone-gated round gives the company direct incentive to publicize hardware milestones on schedule, and a national-security experiment role adds a government-facing interest. Both publishers relay this material largely as reported.
Facts firm, single primary lineage
The hardware and schedule facts are stated consistently by both items and the arithmetic gaps are straightforward to check, supporting moderate confidence. Confidence is capped by the shared TechCrunch lineage, the absence of independent measurement of any energy claim, and the unreconciled funding totals.
product
Inertia's fusion claim is a factory claim: fuel pellets in hours, not a week1 distinct publisher
invest
General Intuition's reported $6B ask reprices physical AI in weeks, not quarters1 distinct publisher
product
OpenAI's plan to hand everyone a coding agent leaves the hard part to the model1 distinct publisher
invest
Tiny corp wants Etched's numbers. Jane Street led $700M at $21B without publishing any1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
mezha.net
1 article · August 25, 2026
runtimewire.com
1 article · August 25, 2026