👋 Hello, Dan Goldin here. A few weeks ago, I saw that after 30 years of operation, the first Milstar spacecraft was being retired. I wanted to commemorate its extraordinary run in my favorite publication — Per Aspera 😛!
When we were tasked to build Milstar, the requirements sounded like science fiction:
Build a communications payload that could survive a nuclear detonation in geostationary orbit and continue delivering secure, jam-resistant Extremely High Frequency communications to the U.S. high command.
👇 Here’s the backstory on how we got it done.
IN THIS WEEK’S EDITION:
✍🏻 Reflecting on MILSTAR…
📊 Guess who’s SPAC?
🏭 Tacomas in TX, Blue
🥊 Catch these hands
Forwarded this? Subscribe to Per Aspera
Back in the 80s, when I was a senior executive at TRW (now Northrop), I was asked to lead the design/development/production of the Milstar Antijam Communication System.
THE MISSION: Our objective for Milstar (Military Strategic and Tactical Relay) was to provide the President, Secretary of Defense, and our armed forces with a trusted and secure, jam-resistant, nuclear-survivable global communications system operating in geostationary orbit. At that time, the Soviets posed a serious threat to our nation’s survival, and reliable strategic communications were an essential part of the deterrence equation. 52 teams contributed to the effort — Lockheed, agencies, academic and corporate labs, advisors, and contractors — and we brought the orchestra together in person often, with 100+ mavericks and misfits traveling to Manhattan Beach, CA every six weeks. Within a year, we had a working system spec, and every contributor knew and owned their requirements. We were ready to make the system perform as intended! (So much of the job is conducting the orchestra!!)
ANTIJAM: Milstar’s most demanding requirement was anti-jam. We bet on a relatively new dichroic filter technology, proven by MIT Lincoln Lab, to build a lightweight, single-reflector, dual-feed antenna. The challenge was getting from a strong concept to reproducible hardware, with highly demanding performance, cost, and schedule constraints. The traditional “cut-and-try” approach was much too risky — technically and programmatically — so we instead went digital, big time.
MODEL, MEASURE, MAKE: We built simulation tools that rapidly and accurately predicted antenna patterns for specific feed geometries — which worked; converged on a design that met our needs; and then verified it in a new indoor, near-field facility instead of the slow, costly outdoor ranges of the day. It was a “digital twin,” decades before the concept would be rebranded into this term, working superbly in 1983 — so well that we extended it far beyond electromagnetic propagation. And finally, since we’re already deep down memory lane — I’ve said it before and I’ll say it again — this was when self-reliance was fully feasible. Instead of reaching for a semiconductor catalogue and buying our chip, we could draw on TRW’s own 100,000 sqft. experimental fab to produce the CMOS and gallium arsenide devices our designs demanded.
A TRW fab diffusion run, 1985. Custom chips built in-house for Milstar.
If we wished to enter uncharted territory in seeking devices with as yet available performance characteristics, the TRW fab (hey, in fact sounds close to: Terafab 🙂) was generally up to the challenge. On numerous occasions, I would literally design a spacecraft system from the semiconductor up to gain an edge.
Try that feat in today’s “fabless manufacturing” culture. If we choose to use the same chips available to our enemies and competitors, will we continue to be truly ahead? We run the risk of falling into the center of the pack!
Our team produced 6 satellites. 5 made it to orbit — sadly, one was lost because of an upper-stage failure during launch (sound familiar?).
‘THE SAVIOR MACHINE’
The constellation operated flawlessly for 30 years, serving the National Command Authority and helping preserve strategic stability. I remain enormously proud of our part in it, alongside a Lockheed Martin team led by the brilliant Sam Araki and a truly talented Air Force Program Manager, Col Eric Crabtree.
Some have called Milstar the “Doomsday Machine.” I call it the Savior Machine, because an adversary who knows America has it will think twice before doing anything strategically foolish.
Today, a new generation of EHF birds carry that mission forward. And to all of you helping ordinary Americans like my wife and me sleep safely at night — I salute you. 🇺🇸🇺🇸🇺🇸
Duffy here. Many of you know I’m a reformed newsman — who, in a past life, started & scaled hard tech media brands as their founding writer, then editor. It feels like eons ago and also like yesterday, where, during one stretch of the go-go times, every other week I’d find myself on the horn with the founder/CEO of a growth-stage startup reaching for the public markets through a side door: the Special Purpose Acquisition Company, or SPAC. All told, I interviewed the management teams behind at least two dozen of them.
From where they sat, the case for a SPAC was some flavor of it’s a ruthlessly efficient way to capitalize my business. They could raise more than another private round would allow, with more certainty than a traditional IPO: a negotiated valuation, a committed PIPE, and permission to pitch investors on forward potential rather than trailing financials. Bear in mind who was pitching: capital-hungry, unprofitable, sometimes pre-product (!) firms with factories to build, hardware to qualify, and revenue marooned somewhere in the out-years.
SPACs soon became as controversial as financial instruments get, perhaps the most side-eyed Wall Street acronym since MBS & CDO. Boom-era de-SPACs became a marker of the era’s excess: financing their fair share of vaporware and retail-left-holding-the-bag stories, averaging a 61% loss over three years (vs. IPOs: -8%), and disproportionately declaring bankruptcy. But SPACs also produced power-law exceptions, such as the space names that assembled fanatically dedicated retail communities and became genuine multibaggers (e.g. Rocket Lab and AST SpaceMobile).
WHAT’S OLD IS NEW… and SPACs are back, with defense, robotics, space, and energy companies once again taking a look at tip-toeing through the side door. By late June, 116 blank-check IPOs (SPACs) had raised $22.7B, making this the busiest start to a year since the 2021 peak. And the incoming class sits squarely inside PA’s wheelhouse:
Agility Robotics agreed to merge with Churchill Capital XI at a $2.5B pre-money valuation, which would make it the first U.S.-listed, pure-play humanoid developer with robots already deployed commercially (in China, Unitree’s STAR Market IPO is pending, and two peers trade in Hong Kong).
Autonomous aviation company Merlin completed its transaction in March. Elroy Air, led by friend of PA Dave Merrill, has agreed to go public at an $800M pre-money valuation, backed by more than $165M in committed PIPE capital.
The dealmaking continues apace: Quantum Space signed a transaction valuing the combined company at ~$1.2B; General Fusion agreed to merge at a $1B pro-forma equity value; and advanced-reactor developer newcleo followed with a $2.4B pre-money deal.
With early-2020s scars still fresh, the rules are tougher this time. The SEC now requires full target liability; sponsors face heavier disclosure requirements around compensation, conflicts, dilution, and deal structure; and de-SPAC projections have lost the federal safe harbor that helped make moonshot forecasts such a defining feature of the last boom. From mgmt’s POV, the pitch now is much like it was then: a more flexible, efficient way to capitalize the company.
So, we’re dying to know… What do you think?
SPACs: Are they...
001 / MEGAROUNDS FOR FOREIGN COs TOO... Korea’s SK Hynix has raised ~$27B in a 7x oversubscribed American Depositary Receipts (ADRs) offering — the largest foreign ADR listing on a U.S. exchange. SK Hynix is Nvidia’s dominant supplier of high‑bandwidth memory:
Rough beginning. Starting around 2010, SK Hynix ploughed R&D capital into something called “high‑bandwidth memory” — a then-niche, expensive technology built on TSV stacking and advanced packaging — inviting years of skepticism that it was over‑investing in a side market instead of chasing mainstream DRAM and NAND.
Then, market leader. That “niche” HBM is now the chokepoint of the AI era. SK Hynix holds ~60% of the market and has reportedly secured ~70% of Nvidia’s HBM4 orders. (And its 2026 DRAM/NAND/HBM capacity is sold out.)
Yet, still valued below Micron. Despite that position, SK Hynix traded in Korea at about 5.5× forward earnings versus Micron's ~6.66× in the U.S., with Micron often enjoying a richer “AI premium” thanks to U.S. domicile and benchmark inclusion — hence SK’s Nasdaq listing. At ~$27B, this surpasses Alibaba’s $25B extravaganza in 2014. Cornerstone investors include funds of Coatue Management, Situational Awareness, and Baillie Gifford.
End of day, the U.S. continues to be the center of gravity for global AI capital — congrats and welcome to 🇺🇸, SK Hynix. NASDAQ: SKHY began trading today.
Blue Origin rolls New Glenn's first stage to the pad, Cape Canaveral Space Force Station — Oct. 8, 2025. (Image: Blue Origin)
002 / ‘BOOTSTRAPPED’ NO MORE… Jeff Bezos has finally invited others onto the cap table, with our second Coatue‑anchored investment of the day and what looks like a cross‑fund affair. The NYT reported Wednesday that Blue Origin is raising its first outside capital, in a $10B on $130B round. For ~25 years, Blue Origin was “bootstrapped” 😅, with Bezos bankrolling the operation himself and selling ~$1B Amazon shares a year to do so. Opening up to outside investors will change the operating physics for Blue Origin, whose motto is Gradatim Ferociter (step by step, ferociously), and perhaps add some useful pressure for the company as it gears up for New Glenn RTF, a growing cadence, missions to the Moon, and more. 25 years bought Bezos + Blue the freedom to determine their destiny, but now, it would seem that it’s time to scale.
003 / THE TACO’S COMIN’ BACK… If you live in Utah, Colorado, Texas, Northern Cali and similar… and you're a Per Aspera reader…. and you work in hardware… we’ll give it a 1-in-3 chance you drive a Taco. The Tacoma is Toyota’s best-selling truck and one of 🇺🇸’s most popular midsized pickups.
The Taco: for drivers in Lake Havasu City, Parker, Bullhead City, Kingman and the like.
Toyota’s spending $3.6B to double its San Antonio manufacturing footprint and add a second assembly line for the Tacoma. Some fun lore: The truck began its (American) life in 1995 (same year as Ryan!) at the NUMMI plant in Fremont, CA — the GM-Toyota JV where GM spent a decade trying to reproduce “The Toyota Way,” at the very same factory that Tesla would later take over. In the ensuing years, Taco production bounced between Texas and Mexico before consolidating south of the border during Covid.
Keep in mind that Toyota is famously a conservative, cost-obsessed company. A $3.6B, long-duration commitment to 🇺🇸 manufacturing was not a decision it took lightly, nor something done just for optics or vibes.
Venus Aerospace's RDRE — first high-thrust rotating detonation engine to fly.
BONUS / GO VENUS GO… Congratulations to Venus Aerospace and PA friend, CEO Sassie Duggleby, on their $91M raise to turn their flight-proven rotating detonation rocket engine (RDRE) into full propulsion systems. The engine delivers more thrust per unit of fuel in a smaller, simpler package — opening a path to lighter, cheaper, higher-performance propulsion for hypersonics, launch, and eventually high(er)-speed transport.
This past week 1X showed off NEO’s new hands, featuring 25 degrees of freedom (DoF) per hand, 22 fully actuated joints across the fingers and palm, and three at the wrist. This puts it surprisingly close to the 27 DoF generally attributed to a human hand. Every joint is force-controlled and backdrivable, while motors in the forearm pull proprietary tendons through the wrist. 1X says hundreds have already come off its production line, with capacity for 10,000 this year.
We have said that dexterity is the keystone capability for general-purpose robots, but more dexterous devices only get you to the starting line. The task at hand is building a scalable, real-world platform — an “Internet of physical interactions” — where your fleet operates in messy homes and generates the failures, training data, and edge cases needed to break the loop of no deployment → no data → no improvement → no deployment.
We wrote all about this in: The Last Hardware Problem.
So, ICYMI, now you can catch up! Per Aspera to you all.
|