If you believe that AI is the next great leap in the story of humanity, you may also recognise that progress is constrained by a surprisingly narrow set of technologies.
Grain-oriented electrical steel for transformers, extreme ultraviolet lithography for chips, single-crystal superalloys for turbine blades and vanes, and the super heavy-lift launch vehicle stack.
Chip production is limited by lithography. Powering chips on earth is limited by transformers and gas turbines. Getting them into space, for abundant solar, is limited by launch capacity.
Within those four categories are about 13 companies who are responsibile for all Western capability.
Beneath those 13 companies are about 40 major suppliers of critical materials, services or components.
It’s a remarkably small pool of esoteric knowledge.
These companies are not safe from competition, but they are many miles out on the frontier. Their technical moats have endured for decades, only growing deeper as the world was distracted by bits and pixels.
EUVL was pioneered in the 1980s (ASML). We’ve had single-crystal superalloys (Pratt & Whitney) and super heavy-lift capacity (Saturn V) since the 1960s , and GOES since the 1930s (Armco Steel). None of this is new.
Significantly, none of these technologies were invented with the goal of enabling artificial intelligence. The serendipity of human innovation and technological progress; magical works from the past, converging in the future.
“By comparison, ours is an age of stagnation — of slowing median wage growth, rising inequality, and decelerated scientific discovery”
Boom: Bubbles and the End of Stagnation, by Byrne Hobart and Tobias Huber (2024)
Contrast with pervasive attitude of today, where technology appears to be an impatient race for domination. Get to scale as quickly as possible, whatever sacrifices are required; all digital, low margin, low capex, low complexity, legible to capital.
As soon as companies have revenue, they are weighed and measured with financial ratios. The fattest are fois grased with power law inflows, as investors can no longer wait for power law outcomes. Get big or die trying.
“Winners” are determined by their ability to consume allocation like a prize bull in a feedlot, not by any intrinsic quality or productive value. The big get bigger.
Thus, the early game of discovery, and much of the midgame of testing and development, has been eliminated. The serendipity of opportuniy is pushed aside by the immediate need to engage in performative success.
It’s a shallow attempt to manufacture synthetic outcomes at scale. Salvation for a market of technology investors that have become almost entirely disconnected from real technology.
In the process, capital and talent have learned to abandon any project that appears to flounder, with no shortage of alternative paths to short-term wealth and status. No looking back; failure is natural. On to the next thing.
So, we have largely stopped creating new esoteric knowledge.
We pursue AI to replace the human brain, robots to replace the human body, and endless dopamine vending machines to distract us from our impending obsolescence.
There is very little that the future will thank us for, and no telling what problems we create for ourselves with the opportunity cost of complacency today.1
“Civilization is not inherited; it has to be learned and earned by each generation anew; if the transmission should be interrupted for one century, civilization would die, and we should be savages again.”
The Lessons of History, by Will Durant and Ariel Durant (1968)
- That biotech remains the last bastion of real innovation probably reflects nothing more than our persistent fear of death; no amount of stimulation or distraction can defeat the lizard brain. If only it was so obvious how existential innovation elsewhere might be. [↩]
Leave a Reply