🫡 Dan Goldin here. In the ’90s, I heard quiet rumblings that something important was happening on a small island in the Western Pacific. So I got on a plane to go see Morris Chang.
What I saw walking the early operations of a young Taiwan Semiconductor Manufacturing Company made my jaw drop. I came home and sounded the alarm, but those who had not seen it firsthand didn’t understand what was happening (they blew me off!).
Our beloved editor Ryan just made his own trip to Taiwan, spent time with some of its leaders, and returned with an excellent primer on what it has become and the predicaments it’s now facing.
Reading it gave me the same mix of excitement and unease I felt on my trips. My recommendation is that you go visit yourself. But in the meantime, reading Ryan’s piece will do.
IN THIS WEEK’S EDITION:
🛬 Confessions of a former Taiwan theorycrafter
✍ The Per Aspera readers have spoken on Pluto
🪱 And we’ve got a new one: Meatball or Worm?
3⃣ Three stats & the new Apple hardware bull case
🎬 Cerebras IPO, San Diego W, Brazilian heavy metal, & Top Gun!
Forwarded this email? Subscribe to Per Aspera here.
GREETINGS FROM TAIPEI…
Ryan here. The strangest thing about Taiwan is how often it appears in American discourse, and how rarely Americans appear there.
The island comes up whenever anyone wants to sound serious about semiconductors, the Indo-Pacific, or the future of AI. And yet precious few of us have been.
America’s ruling class thinks about Taiwan. Big Tech thinks about Taiwan. Frontier lab founders think about Taiwan. And so, naturally, do the professional situation-monitorers on LinkedIn, X dot com, or their geopolitics Substack (oh, the irony).
For half a decade, I was in that last boat, doing what every self-respecting geopolitics poster does. I theorycrafted about Taiwan from 8,000 miles away, fixated on things like “the silicon shield,” the “Davidson Window,” and so on, before finally admitting to myself, fine, guilty, I am one of them (an armchair analyst).
This month, when I had the chance to go and represent my company Array Labs on a delegation to Taiwan, I jumped at it. What I found was a place that’s humid, complicated, competent, contested, and living under a situation the whole world sweats about, but the Taiwanese talk about almost like weather.
My notes from Taipei move from its diplomatic contradiction to the industrial competence, from Dan’s “Cassandra” moment with Morris Chang to the silicon shield, the second shield, and the grey-zone pressure that has become an ambient way of life around the island.
Taiwan may be, pound for pound, the most economically consequential place on Earth. But from the ground, it cares less about our abstractions and more about the concrete steps a society can take to prepare for a season it cannot choose.
Keep reading…
LAST WEEK WE ASKED YOU: should Pluto be reinstated as the ninth planet? 74 of you voted and… reader… it was a landslide:
YES, restore the ninth won with 80% of the votes.
No, the IAU got it right earned 15%.
One of our readers, a distinguished astronomer, was in the room (literally):
“I was at the IAU that demoted Pluto. I feel then, as I feel now, it was politically manipulated."
Another reader, who earned the distinction of longest write-in response, had some procedural bones to pick. First: if Neptune is a "gas giant" and still a planet, why is Pluto a "dwarf planet" and not a planet? Second, as we noted last week: only ~420 IAU members were in the room for the vote — a fraction of the ~9,000-member body. Why wasn't it put to all of them by digital vote?
And for good measure, let’s have a few more readers weigh in:
"It orbits the sun, it's large enough to be a planet."
"Pluto has not lost its place in Astrology."
"Some of my best friends are from Pluto."
It is resolved: the Per Aspera community is decisively in support of restoring Pluto to its former glory!
IN REACTION… to this photo, Dan posted the full story of the worm's retirement this week:
I de-wormed NASA in the '90s. At the time, it felt right to bring NASA's meatball back from retirement to reinvigorate that Apollo spirit we had when we got man to moon in under a decade. We were in a time when the end of the Cold War was indicating the end of our robust space program and the Challenger tragedy caused immense angst across the nation. The team and I felt we needed a symbol that was reflective of that aggressive spirit we had in the 60s to explore and win.
The MEATBALL carries the full weight of Apollo: circular, blue, stars threading an orbit, and it captures the spirit of the agency when it put humans on the Moon inside a decade.
As for WORM, it was introduced in 1975 under the federal graphic design overhaul. It’s clean, modern, and legible at any scale — which is why SpaceX Dragon capsules and Artemis-era hardware have sported it since the logo’s 2020 unretirement.
Cards on the table, some of us are loudly "TEAM MEATBALL.” But it's only fair to also put this one to a vote…
Worm or Meatball?
001 / 1% → 99%… In 2022, Ukraine produced only ~3,000 drones and imported the ~99% of its system as finished kit from China. By 2026, this had completely inverted: with ~99% of final assembly done inside Ukraine and domestic production racing toward a ~5M annual run rate.
002 / $50,000 GPU… A single Blackwell GPU from Nvidia runs up to $50,000, which makes it more expensive than the average new car in the U.S.* An 8-GPU Blackwell server pulls ~14 kW at full tilt, ~2× an equivalent H100 and well above the average power draw of an American family home. Foundation model developers are often spending north of 80% of their funding on top-tier GPUs, because, well, they don’t really have a choice!
*Jensen would like to remind you that Nvidia does not make cars.
003 / $4.2 TRILLION… Tim Cook will step down as Apple CEO on Sept. 1, after 15 years at the helm and taking the company from $350B to nearly $4.2T in market cap at the peak. Congratulations, Tim Apple, on a hell of a run. Waiting in the wings is John Ternus, Apple’s longtime hardware boss 👇.
JOHN TERNUS is a mechanical engineer who joined the product design team in 2001, cut his teeth on the Cinema Display, and went on to oversee everything from iPad and AirPods to the Intel‑to‑Apple Silicon transition.
As the keys to the Cupertino kingdom are handed from a supply chain wizard (Cook) to a hardware maximalist (Ternus), now may be a good time to remind you that Apple has mostly sat out the AI Supercycle arms race.
With the hyperscalers guiding to $600B+ on capex this year, Apple intends to only spend a measly ~$10B. In an environment where Mr. Market rewards “who has the most GPUs,” this has been perceived an Apple weakness.
But… speaking hypothetically… in a world where:
on-device inference and local AI get more feasible as models get smaller, faster, and more power-efficient,
a bonafide hardware renaissance is arriving in Cupertino (which reportedly has six net-new hardware form factors in the works, ranging from ambient AI devices to lightweight AR glasses to home robots), and,
you already have a distribution base of 1B+ active devices and an ultra-loyal, locked-in consumer base, each already carrying a potential neural engine in their pocket,
and you’re sitting on ~$100B in annual free cash flow and $160B in cash…
…a John Ternus-led Apple may indeed have some interesting cards to play. 👀👀
P.S. As Jess Frazelle, CEO of Zoo (AI CAD platform), said:
The best thing Apple did over the last 10+ years was move off Intel to their own chips.
001 / SECOND TIME’S THE CHARM… Cerebras is back in the IPO lane, filing for a Nasdaq listing after yanking its first attempt in 2025 over national-security concerns tied to UAE-backed investor G42. The dinner-plate-sized wafer-scale AI maker booked $510M in 2025 revenue and $88M in net income, a huge swing from its $485M loss the year prior. With a 750 MW inference commitment from OpenAI and a binding deal to run inside AWS clusters, the chipmaker is coming to market with a much beefier S-1 this time around. P.S. back in #037, we talked about how a cohort of disruptive new American chipmakers had soaked up $7B+ in private capital without a single one making it to public markets, so we’re happy to see Cerebras pave the way to the big leagues. Godspeed, guys!
002 / STAY CLASSY, SAN DIEGO… San Diego now has so much drinking water that it’s shopping its surplus to drought-hit Colorado River states and neighboring utilities: a wild place to be, for a city that spent decades as the poster child for Southern California (+ downstream Colorado River) water scarcity. The region pushed ahead with the FOAK, ~$1B Carlsbad seawater desalination plant that came online in 2015, as everyone else kept arguing about the river compact. In one of our first Antimemos last year, The End of Thirst Traps, Per Aspera’s fourth-ever send (!), we argued for a far larger version of this: a Sonoran desert solar-plus-storage mega-desalination complex, throwing off millions of acre-feet a year into the Colorado River system. Carlsbad is a nice pilot-scale proof of concept for this megaproject (which would need better unit economics with scale). This is simply the best path: rather than let the American Southwest slide into degrowth and permanent water rationing, let’s ride the learning curves on solar and storage as hard as possible and build the big-a** desal plant (+ Rainmaker!) to manufacture water abundance in a region that desperately needs it.
003 / HEAVY METAL, AMERICAS EDITION… USA Rare Earth has a $2.8B deal in hand to buy Brazil's Serra Verde Group, owner of the Pela Ema rare earth mine in Goiás — the first large-scale operation outside Asia producing all four key magnet REEs (Nd, Pr, Dy, Tb) plus yttrium. Phase one is slated to hit ~6,400 tonnes of RE oxides annually by 2027, which would put at just over half of heavy rare earth output outside China on current forecasts. We’ve had our eyes on USAR since #036, when we looked at its $1.6B LOI with Washington, a huge federal bet on building a mine‑to‑magnet rare earth value chain on American soil. Huge kudos to USA Rare Earth CEO, and early PA reader, Barbara Humpton — it’s time to mine!
ONE FOR THE CULTURE… Great news: Top Gun 3 is officially in the works, with Tom Cruise back in the cockpit and Jerry Bruckheimer returning to produce.
Maverick’s opening sequence set a high bar that will be tough to follow. In that scene, Cruise straps into the fictional Darkstar and punches the jet past Mach 10 before the airframe comes apart in the upper atmosphere. It was a love letter to hypersonic flight, a technology that Hollywood picked because it still felt exhilarating, just on the edge of science fiction yet still credible, and distinctly American. Darkstar’s silhouette echoed the SR-71 Blackbird, a plane we built, actually fielded, and flew starting in the ‘60s, setting speed records that still stand.
We go into all of this, and the challenge of hypersonics, in Why Aren’t We Flying Faster? A preview chart from that piece: 👇
Count us among the series’ biggest fans. Top Gun remains the best mass-market franchise weaving together Americana, hard pursuits, and high performance in the perfect way.
|