Happy Hump Day, y’all. I’m in the Bay Area for work (s/o Array Labs), and you can feel the energy in the air. On one side, you’ve got all the giddy AI folks in SF pulling all-nighters with whatever Moltbot or Claude Code project has their undivided attention. Then, down the 101, final preparations are underway for Super Bowl LX at Levi’s Stadium, where my New England Patriots will take on the Seattle Seahawks this Sunday.
We’ve put together our own “Super Bowl Poll” that has nothing to do with football…so make sure to vote below.
As always, enjoy the issue that us mere mortals at Per Aspera have put together for you this week. Don’t be a stranger; let us know what you thought, what’s got you energized, what’s keeping you up at night, or anything in between.
Oh, and one more thing….go Pats!!!
IN THIS WEEK’S EDITION:
📊 The Supercycle’s 2nd-order winners (+ an inverse index)
🌕 Artemis II, and a commander’s words worth sitting with
🏈 The (non-football) Per Aspera Super Bowl Reader Poll
📰 News Parsed: Chips, capital markets, & SpaceX
Forwarded this email? Subscribe to Per Aspera here.
In our recurring The Supercycle’s Second-Order Winners series, we like to highlight non-obvious people, companies, and categories that are benefitting from the AI boom and capex buildout. A few updates on that front:
Caterpillar, an honoree in Issue #025, reported sales at an all-time-high and a backlog that surged 71% in 2025 to $51B. This is now the power company that just happens to also make yellow bulldozers. And, as this viral X post notes, Caterpillar has not had a year of single-digit returns since 2014!
Teradyne, a robotics and test equipment company, joins our list as a new inductee - after reporting a significant earnings beat this week, with “AI-related demand” driving 60+% of its Q4 revenue.
Lumentum, a photonics provider, is up 140%+ over the last year due to strong demand for its laser and optical components. Though we’re not counting this one as a “second-order” winner, given our longstanding photonics thesis.
Then, there’s the American Skilled Trades.
While the terminally online offer their takes and go back and forth on “the capability overhang”, perceived AI benefits by management vs. rank-and-file, and whether ROI on all of this capex will materialize…
…the skilled tradespeople, who aren’t sitting on their hands, and are using them to actually build out all of this infrastructure, are being handsomely rewarded. There are the 2nd-order winners we’ve previously discussed (electricians, welders, plumbers) and clear demand for specialized construction capabilities. But it’s time to add another category to the list: fiber-optic cable installation, which calls for drillers, splicers, linemen, and the like. Per the WSJ:
The industry needs 58,000 new workers by 2032. In that same period, 120,000 will retire, creating a combined gap of 178,000.
Since fiber has a less established training pipeline than other vocational trades, it must be built from scratch, with former Marines, rodeo riders, farmers, ex-inmates, and white-collar refugees who got sick of their desk jobs all flooding in.
Comp trends reflect the human capital scarcity, with median wages for telecom installers hitting $70,500 last year (42% above the national median), entry-level workers pulling in $60k with overtime, and site superintendents clearing six figures.
The kicker: “AI isn’t going to replace this,” one executive told the WSJ. “It will accelerate and support some of it, but it’s not going to go out and dig the hole.”
While we’re here…
…two additional notes:
The supercycle isn’t showing any signs of a slowdown. Meta has guided to spend $115-135B on capex in 2026 (vs. $72.2B in 2025), while Tesla estimates it will spend $20B this year (vs. $8.5B).
If you wanted to create an inverse index to the unofficial one we’ve been building above, you’d probably start with publicly traded software companies. Investor fears over AI/agentic advances from the frontier labs wiped out ~$300B in market value yesterday, across two S&P indexes that track SaaS, financial data, and exchange stocks. Software was the worst-performing S&P subsector this year — before this week.
What do you think? Agree, disagree? Something we missed above? Reply to this email and drop us a line.
As soon as next month, the Artemis II mission will launch for the Moon — the first humans to leave Earth orbit in more than half a century. As TIME writes, it’s been half a century since “the crew of Apollo 17 came home, the Apollo moon program was canceled, and the translunar trail went dark.” Now, as the stage sets for a space superpower showdown 250,000 miles from home later this decade, Washington, Beijing, and a nervy space community are closely following along.
On Deck…After a hydrogen leak during this week’s wet dress rehearsal pushed launch to NET March, Artemis II will conduct a ten-day lunar flyby, stress-testing Orion’s systems and implementing a revised reentry trajectory after Artemis I’s heat shield issues.
In the hole…Artemis III, nominally slated for next year, aims to put boots back on lunar regolith, with SpaceX contracted to land Starship and two Americans at the south pole.
The Right Stuff
Artemis II crew — Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen/NASA
At a NASA press conference last September, Artemis II Commander Reid Wiseman said something worth sitting with:
When I look at the future, when we talk about what is our legacy, I don’t want to look five or ten years in the future. I want to look 100 or 200 years in the future. I hope we are forgotten. If we’re forgotten, then Artemis has achieved its goals. We’ll have humans on Mars and on the moons of Saturn, expanding our presence in the solar system. Perhaps we inspired some children, and that might be our footnote — encouraging someone like Susie or Johnny to pursue their dreams. That, to me, would be truly magical.
A commander readies his crew to ride a controlled explosion a quarter-million miles into the void, and takes aim at the most humble victory imaginable: to be forgotten. Because if Wiseman becomes a footnote, it means we kept going. It means that in 2126, Susie looks up and knows the stars belong to her.
Friends, this is The Right Stuff. Our astronauts remind us that greatness requires running toward hard, unglamorous work — or as Shackleton’s apocryphal ad put it — toward hazardous journeys, small wages, bitter cold, long months of complete darkness, and constant danger.
A nation rises to the size of the problems it chooses — in space, and right back here on Earth.
Godspeed, Artemis II.
Time for our Super Bowl Reader Poll. In light of the big game this weekend, we wanted to ask you: which tech/industrial matchup do you think is most important this year? 👇
What's the biggest matchup this year you care most about?
Like their American counterparts, the Chinese hyperscalers — Alibaba, Baidu, Tencent — have spent years building in-house silicon programs as internal R&D departments, with the hopes of turning these cost centers into captive suppliers for their own clouds. These hyperscaler spinouts are now planning to go public, on the heels of a few independent chipmakers that have gotten out of the gate:
The IPO Pipeline
Company | Backer | Status | Exchange | Timing |
Moore Threads | Independent | Listed | STAR Market | Dec 2025 |
MetaX | Independent | Listed | STAR Market | Dec 2025 |
Biren | Independent | Listed | HKEx | Jan 2026 |
Kunlunxin | Baidu | Filed | HKEx | Jan 2026 |
Enflame | Tencent | Accepted | STAR Market | Jan 2026 |
T-Head | Alibaba | Reported | TBD | Jan 2026 |
America has its own, arguably better, AI chip “startups,” yet they don’t appear to be looking for paths to public markets. Cerebras, SambaNova, Tenstorrent, Groq, Lightmatter, d‑Matrix, Etched, and Ampere Computing together have raised well over $7B in private capital to build inference chips that compete head-to-head with giants like Nvidia and AMD. None have gone public, and several have now spent close to a decade as private companies. When an exit happens at all, it’s typically via M&A with a hyperscaler or inference titan (e.g. Groq x Nvidia licensing/acqui-hire deal).
Once upon a time, public markets were the primary tool America used to capitalize its hardware and industrial corporations. Today, thanks to megafunds and mega-rounds, companies can stay private for longer (or forever, if so inclined). Hardware companies that do go public face litigation exposure, disclosure requirements, and increased governance overhead — challenges magnified if you’re a complex engineering company not easily legible to investors; you’re unprofitable; and definitely if you’re pre-product. There are real tradeoffs.
But! There are glimmers that public market access matters for our renaissance. MP Materials, today a $12 billion rare earth producer, went public in 2020 through a SPAC merger. USA Rare Earth (mentioned last week) listed on Nasdaq last March the same way. SPACs have developed quite…a reputation…for themselves. But consider it a silver lining that at least a couple companies doing the hardest physical work of the American industrial renaissance — mining, refining, manufacturing — were able to tap the vehicle for the capitalization path they needed.
And then there’s SpaceX (Est. 2002), which has long resisted going public — for fear that shareholders wouldn’t put up with a “making life multiplanetary” mission, and that public market investors’ short-termist tendencies and fiduciary pressures could compromise this charter. Maybe the cash cow (Starlink) could spin out, analysts mused, but the mothership seemed destined to stay private. All of that’s changed over the last half year, as SpaceX (now with xAI and X, the artist formerly known as Twitter, tucked under its wing) is targeting a mid-June IPO at a $1.5T valuation — potentially the largest public offering in history.
This should provide hope to ambitious hardware founders and their teams, and including the “young“ startups that have been around for a decade. There’s a path to go public, and in the right situation, it can be the best-engineered balance of scale, speed, and safety. We want the Renaissance to flourish, and our best and boldest taking big swings in compute, hardware, and all manner of other complex engineering endeavors. This requires finding a better path for teams into the public markets, not as end goals, but tools to build 🇺🇸🏭.
|