HAPPY TUESDAY… Maybe 10 of you (<1%) will ever get behind the wheel of the “car” you see above, which will be built in a small run that’s a rounding error next to the abundant autonomy future its OEM is gearing up for. Still, you’ve got to love what has to be the best technology demo we’ve ever seen.
We’re talking, of course, about the Tesla Roadster reveal last Thursday, where we saw a supercar with a fan-powered suction system cling to a curved ramp until it hung upside down, then fire SpaceX-built cold-gas thrusters (10 of them, fed by 600 liters of compressed air) to launch off, flip midair, and land on its wheels. Make something so outlandish that people’s first reaction is This must be AI-generated, and you’ve done it right. The future should look like the future!
Today, we head to two corners of New England for a look at naval production at opposite ends of the barbell (big hulls ↔ subsea swarms), unpack a new model of making parts inside an old plant, and run through 10 numbers that suggest 2026 might be the space industry’s biggest year yet.
Forwarded this? Join Per Aspera today.
Every quarter, our friends at Space Capital (a PA Founding Sponsor) boil ~10,000 funding rounds down into a report called Space IQ, and every quarter we mine it for stats that can help us sound smart at space cocktail parties. Today we’re sharing the goods from the Q3 edition, which dropped last Tuesday. Fair warning: methodologically, Space Capital takes an expansive view of the space economy, all the way up to the application layer:
Infrastructure: hardware & software to build, launch, and operate space-based assets
Distribution: hardware & software to connect, process, and deliver what those assets produce
Applications: businesses that run on space-based assets but build no space hardware (so, by Space Capital’s count, Uber is a space company)
The 10 numbers that stood out:
Of the $518.9B in private space investment since 2009, 18% landed in the first nine months of 2026.
With SpaceX not taking new F9 rideshare bookings past 2028, and the satellite crowd starting to openly fret, pure-play launch startups have raised a record $7.8B this year (led by Stoke’s $1B Series E). Rockets reached orbit or landed boosters on three continents in one quarter (with China netting a Long March 10B booster at sea, Zhuque-3 sticking a landing, India’s Skyroot making orbit, and Isar flying from Norway).
Q3’s biggest infrastructure check, $1.4B, went to Hadrian. Manufacturing and components companies have raised $8.0B this year, more than double any prior year.
Golden Dome’s FY27 request stands at $17.5B, while the share funded in base authority under the CR is <$400M. Space Capital has a useful coda here: “Underwrite the requirement, not the appropriation.”
What we’ve felt for some time is showing up in the numbers: a renewed, continent-wide push for European space sovereignty. European startups had their best fundraising quarter on record, taking 28% of global space dollars (Europe’s highest share since 2019), led by Helsing’s $1.8B Series E.
Next stop for the IPO window: Shanghai??? Over in China, 10+ commercial space firms are lining up to list (at least five are launch providers).
Since Isaacman’s NASA pivoted from Gateway to a $20B permanent Moon Base, lunar investment has hit its highest level since 2021.
Buyers are busy, and stingy. Acquirers have spent $17.4B across 92 space deals this year at a median valuation step-up of 2.0x, down from 2.8x in 2021.
Space infrastructure startups’ survival odds are, maybe counterintuitively, slightly better than the rest of the pack:
Of 752 space infrastructure startups seeded since 2009, 51% reached Series A, 15% Series C, and 3.1% Series E.
A CB Insights study of 1,098 tech startups seeded in 2008–10 found 46%, 14%, and 2%.
And the outlier to end all outliers: SpaceX went public at roughly 171x the capital it raised. For comparison, Uber managed 5.5x at its 2019 IPO, Airbnb ~7x, Snowflake ~24x and Zoom, the poster child for capital efficiency, ~63x. Remember when hardware was supposed to be the capital-intensive category?
For more, check out Space Capital’s full report here & watch the Space IQ Briefing at Nasdaq.
Hadrian, the odds-on favorite to build the first dreadnought, has run up its footprint to nearly 3M sqft, raised $1.6B+ to date, and spread its wings across California, Arizona, and Alabama. Which leaves one glaring question: how is this thing not in Texas yet?!?
Well, it will be soon, through a clever side door. Hadrian is moving into Lockheed Martin’s missile plant in Grand Prairie, TX, which makes parts for the interceptors the Iran campaign ate through (#066). It sits 25 miles down I-30 from the big boy, Lockheed’s nearly mile-long Fort Worth plant, which rolled out B-24s in WWII and rolls out F-35s today. The Torrance-based startup will drop a 24-machine automated cell inside Lockheed’s building, running on Hadrian’s software and staffed by LM’s own machinists.
Still, ooh’ing and ahh’ing at American manufacturing is only half the job. The other half is reminding ourselves (and y’all) that the world is rarely as simple as we make it out to be.
Silicon Valley has been telling some version of Netflix vs. Blockbuster for 30 years: the new kid shows up, the incumbent bleeds share, and a decade later, its only presence on Earth is a novelty shop in Bend, OR. When venture (re)discovered defense, we brought this script along and recast it as neo-primes vs. primes, with you-know-who playing Blockbuster. That story has raised a lot of capital, escaped dual-use confinement, and spread to shipyards, steel, semiconductors, and so much more. Sometimes the shoe fits and the story really is simple-as: it’s a zero-sum cage match, and one goes home with nothing. Other times we want it to be true, because the primes earn the skepticism with blown programs, budget over-runs, and behavior that looks more like a lobbying-and-integration machine than a manufacturer.
But you can’t sort Hadrian into either box. It’s not a prime, nor neo-prime, but a third thing… like countless other suppliers, subs, and even venture-backed startups!
Where you stand depends on where you sit.
Which brings us to two more binaries we like to cram the world into:
Vertically integrate or you’re toast. The supply/value chains behind all manner of critical real-world industries would beg to differ.
Sovereignty and statesmanship cannot coexist. Sovereignty (building far more of what we need on our own soil) and statesmanship (working with Allies) get pitted as enemies. Also incomplete! (See: Pax Silica, and a growing roster of co-production deals). A timely example: Japan builds PAC-3s under Lockheed’s license and ships them to us.
Competition and cooperation turn on your place in the food chain, whether you’re a full-stack/finished systems company, which program/market you’re chasing, and the incentives of the moment. Sometimes they collide; sometimes they align. In munitions & missiles, the need is far bigger than any one narrative-driven protagonist can fill. Beware of overly simplified stories, however compelling. Because very little in this business is black and white.
It flew under the radar (sonar, really)… Anduril has had the kind of month that buries its own news, so you’d be forgiven for missing that on Oct. 2, it announced the acquisition of Popoto Modem, a small underwater-acoustics shop in Sandwich, Massachusetts. Back in June, Jeff & Ryan argued in The Leash that talking to a robot once it dives is the ocean’s great unsolved, and very investable, problem.
The crux: we mastered the fixed link on the seabed a long while ago, but the moving link still eludes us. Saltwater swallows radio, sound crawls along at kilobits, and blue-green lasers die off past ~200 m. Plus, transmitting at all gives your position away. So undersea autonomy runs on a de facto doctrine of fire-and-forget (program, launch, hope).
Anduril M&A: Its release reads like a sequel to our piece, calling sound “the only signal that travels a meaningful distance through water.” Acoustic gear, it says, “has historically been prohibitively expensive,” which has capped how big an undersea formation can get and restricted comms to point-to-point links between exquisite platforms.
Why Popoto: As the modem brand of Cape Cod’s delResearch, it’s been driving cost down since at least 2019. Its latest board weighs 170g, starts at $2,750, runs from 80 to 10,240 bits per second depending on range, and speaks the language (encrypted standards) a defense buyer cares about. The modems will be fitted into Anduril’s Copperhead family of small AUVs first, allowing a single operator to command “dozens, even hundreds” of vehicles with confidence they’ll stay “in contact and on task.”
The likely next step? An encrypted, allied-compatible acoustic mesh strung between vehicles, seabed sensors, and bigger hulls.
A second opinion: Will O’Brien, an Irishman and cofounder of Ulysses (which sells underwater drones for as little as $50,000), appeared on Cheeky Pint last week to claim the Guinness-per-podcasting-hour crown. Between pints, he said that corrosion and pressure can be engineered around: “I think the biggest problem, though, is comms underwater.”
CALL TO ACTION: We’d say IceNine, Jeff’s new early-stage fund, has been fishing in the right waters. If you’re working on this problem, his inbox is open.
001 / DANGER CLOSE… Between Oct. 15 and 19, Russia’s Cosmos 2612 closed to within 500 meters of ICEYE-X36, a radar satellite run by ICEYE’s U.S. arm, and spent days shuffling between positions ahead of and behind it. LeoLabs, which tracks this stuff for a living, calls it the most advanced uncooperative rendezvous it has ever observed. The approach was months in the making. Cosmos 2612 went up in April with five siblings, and in May five of the six made plane changes to match ICEYE’s orbit, a maneuver LEO birds almost never attempt because it costs so much propellant. ICEYE has supplied Ukraine with radar imagery since 2022, though which satellite does what has never been made public. Moscow has said commercial satellites aiding Ukraine are fair targets, and in 2020 tested a satellite that fires a projectile at close range. For all of the Star Trek-style marketing the space industry likes to wrap itself in, the reality on orbit looks a lot more like Star Wars: it’s contested, adversaries are already in position, and threats abound. (Dan covered the sledgehammer version in #065.)
002 / HOW TIMES CHANGE... In 1986, Washington had a big chip problem, and its name was Japan. Japanese chipmakers had just passed American ones in global market share, the two governments signed a pact to stop Japan from dumping memory chips below cost, and a year later Commerce and the Pentagon talked Fujitsu out of buying Fairchild Semiconductor. Forty years on, Tokyo is paying for American fabs and Commerce is the one taking equity. On Oct. 14, GlobalFoundries was picked to run two new 300mm fabs at its Malta, NY campus, funded with up to $21B of the $550B Japan pledged in last year’s tariff deal. The first fab is due in 2029. Somewhere in Tokyo, a retired MITI man is enjoying this.
003 / BATH IRON WORKS... Anyone who has driven Route 1 through Bath, Maine, knows the view: as you cross the Kennebec on the Sagadahoc Bridge, there’s no missing the giant destroyer yard on the riverbank. On Oct. 12, the Pentagon said that yard, General Dynamics’ Bath Iron Works, has the “capacity and stability” to build a second Arleigh Burke-class destroyer this year, reversing an FY27 request that funded just one. Your editor has made that drive 50+ times, mostly as a kid in the back seat lobbying his parents to pull off in Bath for the Maine Maritime Museum or Beale Street BBQ (now Kennebec Meat Company), where he could hear stories from the people building the big ships. There are 6,800 of them at Bath Iron Works, in a town of about 8,800, with 11 destroyers on contract. Congress still has to fund a second hull, and Deputy Secretary Steve Feinberg says the yard “must continue this momentum to fully resolve legacy production challenges.”
Bath Iron Works Shipyard on the Kennebec River, circa 2002. It’s one of two Arleigh Burke yards in the country (the other is HII’s Ingalls in Mississippi). Image: The U.S. National Archives
At its WWII peak, Bath delivered 67 destroyers in 1,156 days, more than any other U.S. yard. Today’s ships are several times the size and far more complex, so nobody expects that same pace, but more than a century of building destroyers buys a lot of know-how. Yes, legacy cuts both ways, and yes, Washington priorities can and will shift under the yard’s feet (the Navy once planned 32 Zumwalt-class destroyers, and Bath built all three that survived the cuts). Now the mandate is back, along with a need to modernize & pair know-how with new tools. Bath has out-built everyone before… count us as cautious optimists that it can do it again.
What have friends of Per Aspera gotten done recently?
001 / WOLF PACK, PRINTED... L3Harris is bringing in Divergent to help mass-produce its Red Wolf (strike) and Green Wolf (EW) launched effects, starting with 3D-printed parts for existing designs and working into new ones from there. The very same day we wrote that high-mix factories like Divergent’s were well-equipped to tackle our DIB’s production challenges (#066), a prime came knocking. Kudos, Lukas Czinger and co!
002 / ELEMENT, IN ORBIT... CesiumAstro spent years building a big business selling the phased arrays, radios, and processors that go in other people’s satellites and aircraft. On Oct. 1, CesiumAstro launched Element-1, its first full satellite, built from those same parts and designed to be reconfigured with software on the fly — the first, mighty step toward a planned constellation of 700+ spacecraft. 💪
003 / DRONES ON THE BATTLE NETWORK... On Oct. 8 the Air Force expanded SciTec's contract to pipe counter-drone sensor data into its Advanced Battle Management System, the same picture commanders use for everything else. Seems to us that one year into life under Firefly (following an $855M acquisition), all systems are nominal. Congrats to SciTec president David Simenc and team, and to Firefly, for the continued success!
DEPLOYING RIGHT NOW… SpaceX’s Bandwagon-5 lifted off this morning from Florida with 32 payloads, including Portal’s Starburst-1, Impulse’s LEO Express 4, six Xona navigation birds, and three HawkEye 360 RF-sensing spacecraft. As Aaron Burnett notes, up to 40-50% of the payload mass on today’s rocket is dedicated to vehicles coming back to Earth (Varda’s W-7 and SpaceX’s own Starfall capsule). Godspeed to all 32, round-trip and one-way alike!
|