🫡 Hello my fellow builders (of the next American Century). Dan Goldin here. A few weeks ago, we had a conversation with Caleb Boyd of Molten Industries that filled my soul.
I'll let Ryan get into the substance below, but today we're launching Hard Reflections (told ya we had more in store!): an interview series to capture tribal knowledge of past and present with the founders, chemists, operators, and obsessives doing the hard, unglamorous work Per Aspera exists to cover, convene, and celebrate — Caleb being our first.
Those who’ve met me know I just love smart, strong-willed, young people (yes <85 is young) who are doing hard things for our country, so with that I’d like to let Ryan take it away.
PS. Speaking of meeting: I'll be at Reindustrialize next week. Reply to this email if you'd like to get together!
IN THIS WEEK’S EDITION:
🧪 America’s Next Dow Chemical
🏰 Friction Stir Welding II
🙌 Antares goes critical
📰 China’s AI bet, fusion, tradespeople
🏋 DARPA’s Lift Challenge
🇺🇸 From a German’s POV
Forwarded this? Subscribe to Per Aspera
EP 001 / Ryan Duffy here. Back in #041, we wrote that chemicals are the underrated backbone of American reindustrialization, and put out a call for those of you working in the sector to drop us a line. Caleb Boyd — cofounder of Molten Industries and a reformed VC, who crossed paths with our very own Jeff Crusey in a past life — answered the call! Below, a summary of Caleb's conversation with the PA team (Dan, Joy and yours truly).
On Caleb Boyd's desk sits a batch of original Acheson graphite — a reminder of Edward Acheson, the American who worked out how to turn carbon into synthetic graphite in 1890 and put the US at the front of the industry. Today, graphite sits at the anode of nearly every lithium-ion battery on Earth. It's a fitting paperweight for a founder trying to do a very American thing all over again, more than a century after the country that invented the process let it go.
The graphite dilemma
~99% of the world's battery-grade graphite is made in China
China owns the refining knowhow, the equipment to refine it, and the equipment to make that equipment
US demand is ~200,000 tons today, trending toward ~1M tons by 2030 — almost all currently sourced from one country (you know who)
As Caleb tells us: You can buy graphite, you can maybe buy the equipment. But the bottom of the stack is the hardest part — the knowhow, experience, and ‘equipment that makes the equipment’ stuff, which the West largely, willingly walked away from (classic). Which means there are things we have forgotten how to do!
You reap what you sow. Tell an entire industrial culture to avoid a long list of "messy" chemicals at all costs, for decades, and eventually you forget how to do the thing — and someone else remembers for you.
The journey to split the gas
Molten Industries is an Oakland-based startup founded by two Stanford materials-science PhDs (both came out of thin-film solar) who spent the pandemic tinkering in a garage on Stanford's campus, running an early rig off the landlord's EV charger. A true Silicon Valley garage startup story!
They set out to make the cheapest hydrogen on the planet via methane pyrolysis — natural gas in, hydrogen out, no oxygen, no toxic post-processing — and partway through started making high-quality graphite instead. Kevin Bush, Caleb's cofounder, first thought it was diamond. As Boyd puts it, Molten "became a graphite company with a hydrogen byproduct, instead of a hydrogen company with a carbon byproduct."
Today, they have a pilot unit making ~3 tons of graphite and ~1 ton of hydrogen a day — and a whole lot more on the way. Molten is targeting an astonishingly low production cost — $3,500 a ton (vs. China’s ~$5,000 and the $8,000–9,000 it costs producers anywhere else).
Central Debates
There are spicy parts of Molten’s plan, such as making graphite in America and exporting it back to China, still the ~80% battery market for the next decade (the idea is to treat it like LNG, upgrading cheap gas into a commodity and shipping it wherever there's demand).
The model Caleb reaches for is Dow, which got its start pulling bromine from Michigan brine while German incumbents tried to strangle it — today, nobody thinks of Dow as a bromine company.
And then, of course, there’s the standing irony: the NIMBY impulse that pushed messy chemical manufacturing out of the US is the same one that hollowed out our ability to make this material.
A Hard Reflection
Nothing in this business is clean or cut-and-dried — least of all bringing American graphite back from a near-total dependency on China — making Caleb a perfect first guest for Hard Reflections. Big thanks to you, Caleb, for reaching out and letting the PA community dig into the country-shaping work you've taken on. 👇️ Read the full interview online.
Dan Goldin again. I never thought we'd get such an active response from you all on the Friction Stir Welding piece we did — on how it saved the ISS. So I want to share a vignette from the other side of the ledger: a time FSW failed me, and why that failure is just as relevant for you all today.
The program was Eclipse Aviation. Vern Raburn — Microsoft's 18th employee, a pilot, and a true believer — set out in the late 1990s to build a twin-engine jet for under $1M, against the $15-20M business jets of the day.
Vern Raburn!
The key nut to crack was the engine.
Through NASA's General Aviation Propulsion program, we put ~$39M behind Sam Williams (founder of Williams International and essentially the inventor of the small gas turbine engine) to build the FJX-2, commercialized as the EJ22, the lightest, most efficient small jet engine ever built. Lightest is the operative word: that engine was sized for a featherweight airplane, and we at NASA worked with Vern on exactly that: an ultralight graphite-composite airframe to match.
Composite shell + tiny engine = <$1M price tag. Or, at least, that was the working theory.
Keep reading for why FSW didn’t work in this case and how SpaceX faced a similar design fork (and chose correctly).
ANTARES GOES CRITICAL. On June 4, fully one month ahead of Uncle Sam’s deadline, Antares Nuclear's Mark-0 went critical at Idaho National Laboratory — the first of the eleven reactors in the DOE's pilot program to do so, the first privately developed advanced reactor to go critical on American soil in over 40 years, the first new design to run at the site in 50+ years, and the first to ever get there under DOE’s new authorization pathway instead of an NRC license.
Wowza!
(Zero-power criticality means a chain reaction ran with zero usable watts. Mark-1, the machine that makes electricity, is due next year.)
Reminder: back in April we had you vote on whether America would hit its target of three test reactors critical by July 4, our 250th birthday. 35% of you said fewer than three, 32% said exactly three, and 32% more. The count is now at one, with 23 days to go! 🍿
Congrats to Julia DeWahl, a friend of PA and cofounder of Antares, and the rest of the team. And godspeed to the rest of you in the pilot program!
001A / CHECK-IN ON MADE-IN-CHINA AI… Beijing is said to be lining up a 5-yr, 2T¥ (~$295B) program to blanket the country with AI datacenters and stitch them into one national computing network by 2028. The government is steering it, of course, with state telcos like China Mobile and China Telecom doing the heavy lifting on the ground. At least 80% of the kit, AI chips included, is supposed to come from domestic suppliers like Huawei (Nvidia need not apply.) Now, we are not ones to soft-pedal the competitive threat from China across critical technology and industries, but let’s put that big scary number into context:
$295B sounds enormous in a vacuum, but divide it by five and you get ~$59B per year. Then size that up against ~$750B, the (upper-end) aggregate capex that U.S. hyperscalers have guided to for 2026.
Translation: one year of PRC AI spending in this new plan is equal to roughly one month of Big Tech/frontier lab AI spend.
And before anyone in accounting yells at us: yes, the 2T¥ is the state's share only, and yes, China has its own companies that will spend on top of this. But stack state plan and 'private' sector together and China's all-in number still looks to be on the order of a fifth or less of what American industry will spend this year.
001B / SO WHY THE GAP? We see a few compelling live theories: 1) China just may not be as AGI-pilled as the powers that be in America. 2) A frontier model that behaves probabilistically, even with deterministic harnessing, is still a tricky challenge for a censorship state. 3) Open weights (and/or model distillation) trail the frontier by months and run at a fraction of the price, so why shoulder the cost burden of being at the frontier? Let American labs front the bigger bill finding the ceiling, then fast-follow for pennies on the dollar. 4) It could simply be that China is placing its chips where it already has a strong hand: upstream (abundant cheap electrons, and reconstituting atoms in the real world at megaproject scale) and downstream (embodied AI).
002 / AMERICA’S FUSION PLAN... On Tuesday, the DOE released its Fusion Science and Technology Roadmap, the national plan for putting commercial fusion on the grid by the mid-2030s. This follows an October draft that was hardened by input from 800+ researchers, engineers, and industry hands. The not-so-subtle subtext is that fusion is no longer “just ten years away.” Washington is indicating it will build what it must: those things that no single company can justify (materials test facilities, neutron source, tritium testbeds), while the private sector — already $10B+ deep — builds the pilot plants, then the real thing. This is precisely the division of labor we called for in Fusion: Core Power, so we’re happy to see it stamped into federal policy!
003 / TRADESPEOPLE ARE IN… This week, Dina Powell McCormick and Mike Rowe launched America’s Workforce Academy, a new program that pays workers as they train and guarantees a job at the end building the datacenters, fiber, and power infrastructure that AI needs. Datacenter electricians are now pulling in north of $280K in some markets, and the construction industry needs ~349,000 new workers this year. A theme we keep coming back to here at PA (see #029, #037, etc)… the AI race depends on skilled tradespeople — welders, pipefitters, HVAC engineers — who build and maintain the real world.
Last November we told you about the open call for the DARPA Lift Challenge — a national competition awarding $6.5M for teams that can crack one of aviation’s harder problems: lifting more weight! The challenge specifically calls for a drone that lifts 4x its own weight, in an industry that mostly sits at 1:1 or worse.
This week, DARPA announced the first 72 teams down-selected from the open call to compete in an August fly-off.
480+ teams applied, 200+ passed gating milestones (final concept paper, flying prototype video, FAA Part 107 compliance), and 72 got the “you’re in” call, with DARPA saying more tranches will arrive in coming weeks up to ~150 teams total.
The next fly-off will take place Aug 2–9 at the National Museum of the U.S. Air Force in Dayton, Ohio. Teams will fly a 5-nautical-mile circuit, and the highest payload-to-weight ratios plus most novel designs will take home the prize money.
For any aviation junkies out there: the final four days are open to the public, though seats are limited! Register here.
Outside the laws of physics, there's no good reason to accept a long-held industry convention or technological compromise just because "that's how it's always been."
Challenging complacency, compromises, and convention is the spirit we love at PA. Here's hoping DARPA and the winners get it done. 💫
Over the last few days, German X user Freddy (@FreddyLA7) has amassed a huge, rabid American online following as he documents his experience here in the States, his road trip from Alabama down through Florida, and his willingness to fully soak up the local culture along the way.
While in Auburn, he attended a World Cup warm-up friendly (Argentina vs Iceland) and got the full SEC treatment — including seeing the War Eagle take flight (can you spot it?). The tradition traces back to Auburn’s first game against Georgia in 1892, when a Civil War veteran’s pet eagle broke loose and circled the field during Auburn’s game-winning drive.
Freddy has also hit 1 AM stops at Buc-ee's (IYKYK) and discovered Ella Langley 🎶. Overnight, he's won over Americans nationwide. And hopefully… we've won him over too ❤🫰🇺🇸.
As he put it: The European mind can’t comprehend!
|