Hey it’s your editor Ryan. With apologies to all visitors here for the World Cup, we're going to use an analogy from real football 🏈 for today’s top story…
IN THIS WEEK’S EDITION:
🏈 Fourth quarter: 2 critical, 1 to go
⚛️ Antimatter Moment
🗳️ Your two cents: three by July 4?
🛰️ Dan Goldin on Sophia Space
🌶 Jalapeño
Forwarded this? Subscribe to Per Aspera
They Had Us In The First Half
Let’s imagine the last year in SMR-land as a football game and start where we’re at now.
Fourth quarter. Last Thursday, Ward 250 went critical — the moment a reactor’s chain reaction becomes self-sustaining — making Valar Atomics the 2nd of the DOE’s 11 pilot reactor projects to fire — two weeks behind (PA friend) Julia DeWahl’s Antares, whose Mark-0 microreactor hit criticality on June 4.
First quarter. A ~year ago, with the ink still drying on EO 14301, this felt like a longshot… or, if you talked to someone in the nuclear establishment, a pipe dream. When we put it to a reader vote in April (#049), a plurality of you voted the field would come up short — which, quite honestly, was a fair (and still in-the-money) bet:
Until recently, there was very little cause for hope in nuclear, given American public opinion, the decadal $10M+ licensing pathway, Vogtle’s cautionary tale, and more.
America’s nuclear stagnation didn’t inspire confidence that anything involving fission could move fast or on budget. Critics called out all of the ways that novel form factors and the new licensing pathway/pilot program wouldn’t work.
VCs weren’t as hot, initially, either — but not for the reason you might think. Last fall (#022) we mentioned Isaiah Taylor’s account of fundraising in the early days, where the Valar cofounder attributed initial VC no’s to one fear: that a nuclear startup could never keep raising the ever-bigger rounds a scale-pilled megafund treats as the price of entry. We love a founder with a chip on their shoulder.
Left: Valar’s Ward 250 in transit. Right: Antares’ Mark-0 reactor hardware.
What now?
With two down, one to go, and nine days til July 4, we’re in true nail-biter territory now. Two we’re watching for that third slot are Aalo-X at Idaho National Lab and Oklo’s reactor down in Prototown, Texas. Hopefully we get a clean sweep by the Fourth, but regardless, a few things are already clear to us:
For the ones out front, the hard part is still ahead. Going critical proves the physics, but production — at volume and on budget — is a separate beast. (“The factory is the product” is a hard tech cliche for a reason!)
And, mind you, these are micro- and small modular reactors (SMRs). They have a long, safe track record in the Navy’s subs and carriers, where the customer will pay a handsome premium for megawatts and doesn’t need the unit economics to pencil against a merchant power market. Commercial buyers won’t be that forgiving.
But in some sense, what we see from this ‘race’ is that we’ve all already won. Competition and innovation are a great thing, especially in a corner of energy that has gone 40+ years without ‘em.
YOUR TWO CENTS… We need an update on how you’re all thinking, given the developments since our first poll:
Will we get three critical by July 4?
In 1928, Paul Dirac wrote down an equation that predicted the existence of a particle identical to the electron but with opposite charge. Carl Anderson found it in a cloud chamber four years later — the positron or positively charged electron, the first confirmed form of antimatter. The antiproton or negatively charged proton, a little less than 2000 times the positron mass was discovered in 1955 by researchers Owen Chamberlain, Emilio Segrè, Clyde Wiegand, and Thomas Ypsilantis at the University of California, Berkeley. For nearly a century, the physics has been settled: when matter meets its mirror image, 100% of the combined mass converts directly to energy.
Can we make it, and can we hold it?
Last Thursday, that question, dormant in the popular imagination for decades — came roaring back.
Elon posted on X: "In the future, a trillion times a trillion dollars will be spent on making antimatter to travel to other star systems."
To which Jared Isaacman replied in four words: "I support antimatter propulsion."
And Casey Handmer resurfaced his August 2024 engineering case with three words: "An incredible discovery."
The real advantage is specific impulse — how efficiently an engine turns propellant into thrust. An advanced antimatter drive could reach millions of seconds. For comparison, the best chemical rockets top out around 450 seconds, while even nuclear thermal rockets only reach just under 1,000.
Two hard problems remain:
Production: CERN synthesizes roughly 1 nanogram of antiprotons a year — at that yield, fueling a notional spacecraft is a geological-timescale problem
Storage: antimatter cannot touch any normal matter, requiring magnetic and electric traps — a bottle made purely of fields, with no walls
This is a manufacturing and precision-engineering problem of the first order.
Left: The first known antimatter particle; Right: AMS on the ISS
Dan had a note directly for Elon and Jared:
Hey Elon and Jared, why wait for the future to expand the thought process for making antimatter propulsion a reality. During the 90’s I worked with Sam Ting to develop the Advanced Magnetic Spectrometer (AMS) technology so it could probe the heavens for antimatter and other elemental particles. We flew a development unit (AMS-01) on the Shuttle in 1998 and the great NASA spaceflight team installed Sam Ting’s brilliant operational (AMS-02) on the International Space Station in 2011. In a few months the NASA spaceflight team will upgrade AMS-02 on orbit to refresh its capability.
In a call yesterday, Sam told me it is still yielding exciting results. In parallel, we’re beginning to see the initiation of serious theoretical work in antimatter propulsion. Research is advancing from broad speculation to detailed architectural studies. It will take time but under NASA’s leadership we need to intensify the American focus onto antimatter propulsion research. It’s impactful time is coming!
Dan Goldin, here. As a Cold Warrior, my career has unfolded largely behind the scenes, in work of proprietary nature. Today, however, I'm excited to share something I can talk about — Sophia Space!
I've been an advisor to Leon Alkalai since he founded Mandala Space Ventures, and helped spin its orbital datacenter startup, Sophia, out of the studio. We were early: it began as a 2-yr incubation back in ‘21 drawing on Mandala, JPL's Blue Sky Studies program, and Caltech's Space Solar Power Project. (This is your reminder that you have partners/resources in your backyard!!)
Sophia pairs commercial accelerators (Nvidia, Qualcomm, AMD) with rad-hardened controllers, and is building orbital compute for the harshest computing environment there is.
They’re attacking what we at PA see as the “wedge” for space-based datacenter technology — crunching data on orbit, instead of beaming raw bits to Earth with ultra-high bandwidth data links, large/$$$ ground antennas, and associated latencies.
Leon calls it “space computing as a commodity, not a luxury” — and he’s right.
This month gave us reason to celebrate: Sophia’s seed round closed 4x oversubscribed, the startup was named “one to watch,” and it has signed with Apex to fly hardware on their Nova spacecraft bus. 🎉 🎉 🎉 Now let me show you how a rocket scientist and engineer celebrate:
This past weekend, we climbed 2,000 feet of elevation gain — five miles up, five down — to beautiful view of the Santa Monica mountains as they meet the surf of the Pacific Ocean. Up there, we bumped into three recent Purdue engineering grads who just arrived in LA only last week, to start their careers at SpaceX!
We got to talking, and I shared how I myself ended up in LA. In ‘67, Jerry Elverum, the genius at TRW behind the Pintle engine, whose 10:1 throttling ratio landed the Apollo crews on the Moon — hired me away from NASA Glenn. Over the next five years he turned me into an engineer who could take a cutting-edge idea and make it work, fast and cheap. Years later, Jerry did the same for a brilliant young Tom Mueller, also a friend of PA. Tom would go on to become SpaceX Employee #1, lead critical propulsion development for Elon Musk, and today he’s cofounder and CEO of Impulse. Three generations of propulsion people from one man… the story blew the minds of the newly minted SpaceX employees 😃.
People always ask: how do you keep going? But when you work alongside people like Leon, Tom, and Jerry — who never, ever give up on expanding human capacity — the real question becomes: how could you stop?
Sophia flies next year, and I couldn’t be prouder. Ad astra per aspera!!! 💫⭐
P.S. Not paid. Just a patriot excited about building the future.
Re: ODCs, catch up on Realities of Space-Based Compute.
001(A) / REMEMBER THE SANDWICH?… Last month, Principal Mineral acquired Cerberus’ Isola Group, a leading producer of copper-clad laminates and prepregs used in PCBs. The $280M financing was co-led by your friends at The New Industrial Corp. (& highlighted at REIND in Detroit last week), along w/ Overmatch Ventures. This deal follows Principal Mineral’s acquisition of Camden Copper, currently the only North American manufacturer of defense-grade electrodeposited copper foil — the thin conductive layer that gets laminated onto the insulating board and then etched into traces. (Making it one of the most critical raw materials for PCBs.)
American Made Copper (via Camden Copper)
In April 2025, with the Camden plant at risk of closing, Principal Mineral stepped in, acquired the assets, and breathed new life into the operation. Founded in ‘24 and led by CEO Adam Johnson — who, fittingly, was at MP Materials in a past life — Principal Mineral is targeting the long-neglected midstream of the U.S. electronics supply chain (remember our rant on the middle meat in the Great American Sandwich?) If only there were a generational, multi-sector demand boom for copper… oh wait, that’s right, there is.
001(B) / SANDWICH SPECULATORS… There’s a new class of industrialists and financiers whose thesis is simple, really: if you can buy a real-world industrial asset, in a category where America is going to need more capacity and not less, for less than it would cost to replace or rebuild from scratch... you buy it, feed it, and hold it. And wait for the market to return to its senses. We think this class — sitting between the mine & the missile, datacenter, satellite, robot, or factory — is going to do quite well.
002 / SCHRÖDINGER'S SHOP FLOOR…. It feels that American factories are booming 👆 and falling all at the same time. It’s something that’s hard to hold in our head all at once. But it’s true:
On the output side, S&P Global’s U.S. Manufacturing PMI rose to 55.7 in June — its highest in four years.
But mfg employment has declined by more than 80,000 jobs over the past year — marking one of the more sustained drops outside of the Covid era.
The leading explanation is obviously productivity. Technology → do more w/ fewer people → decouple output from headcount. But as with everything else in this space, it's worth resisting the urge to treat the problem monolithically.
The sectors driving most of the measured “strength” in output, especially computers and electronics also happen to be most capital-intensive (remember when capital intensive was taboo?). On the flip side: labor-intensive industries like apparel and textiles have seen both output and employment collapse — job losses on the order of 70–90%.
The result is an aggregate picture that looks stable but conceals the increasing concentration of manufacturing growth into high-productivity, high-capex, low-labor segments.
Perhaps indicative of what America is prioritizing — we can’t do it all… just yet.
FINALLY SOME SPICE… As Harvey Dent once said: “You either die a hero, or you live long enough to see yourself become an Application-Specific Integrated Circuit developer.”
This week OpenAI proved the timelessness of Dent’s wisdom when it announced Jalapeño alongside Broadcom: the lab’s first custom AI chip, a purpose-built inference processor taken from clean sheet to tape-out in nine months, in what may be one of the fastest ASIC cycles ever (helped along, of course, by ChatGPT and Codex). This puts OpenAI in league with Google (TPU), Amazon, Tesla, Meta, and countless others who have pursued custom silicon in search of greater efficiencies and more vertical integration.
The most important piece of this, for the culture, is the name. For years the frontier labs have named their products like printer drivers: GPT-40, o1, Opus 4.8, o3-mini, 4.1, 04-mini-high. The naming convention started innocuously enough — a quirk of the company culture, perhaps — but after this many repeat offenses, it feels like there are all-powerful alphanumeric committees who are doing this to us on purpose as a bit.
Imagine our surprise, then, when word came that a lab’s first custom silicon had turned up, and that it was called… Jalapeño! What a breath of fresh air. A tiny win for us humans, but a tactical one at that, because compute that runs inside the company isn’t technically a product. We won’t rest our case until something spicy, daring, and memorable shows up as a major model name.
|