From The Early Signal by gmoney (@gmoneyNFT). Jevons Paradox, Standard Oil, and why the market keeps getting AI wrong.
Leopold Aschenbrenner called the shot on where AI was heading before most people were paying attention.
Former OpenAI researcher. Wrote Situational Awareness in 2024. He's been early. He's been right.
Google published TurboQuant. Markets sold off infrastructure stocks 19.5% in five sessions. Everyone called it peak capex. And all I could think was: we've seen this exact argument before. Every single time it's been wrong.
Kerosene. That was it. Kerosene lamps were how America lit its homes. Standard Oil built an empire around that single use case. One of the most powerful companies in history, built on one product.
Then the light bulb was invented.
Cheaper. Cleaner. Safer. The kerosene lamp was dead. By every reasonable analysis, Standard Oil should have been finished.
The opposite happened.
New applications emerged. Cars. Trucks. Industrial machinery. Plastics. They consumed vastly more oil than kerosene lamps ever did. The thing that was supposed to kill demand turned out to be a rounding error.
More oil consumed by cars and industry than kerosene lamps ever demanded.
When you make a resource more efficient to use, you don't use less of it. You use more. Because efficiency unlocks use cases that weren't possible before. Lower cost expands the market beyond what anyone could model.
Compresses the memory footprint of AI inference by 6 to 8 times. A 16GB MacBook can now run 100K-token conversations that previously required cloud APIs. Middle-out compression, but for AI memory.
Micron dropped 19.5% in five sessions. Thesis: AI needs less memory. Less compute. Peak capex. Sell the infrastructure stocks.
Jevons says they have it backwards.
When inference gets 6x cheaper, you don't run the same amount of inference for less money. You run inference everywhere.
On every device. In every app. On every agent. In every workflow. Total demand doesn't shrink. It explodes. Just like oil after the light bulb.
"Efficiency means we need less infrastructure." Every efficiency gain in the history of technology created more demand, not less. Every single time.
Chatbots. Coding assistants. Search. These are the kerosene lamps.
They're useful. They're real. But they are not what AI is ultimately for. Aschenbrenner's point is even bigger than the efficiency argument.
The car hasn't been invented yet.
We're building an empire around lamps and haven't even imagined what the car looks like. Everybody's arguing about the price of kerosene while the car is being built in someone's garage.
Aschenbrenner also laid out the case for a trillion-dollar nationalised compute cluster, the 2027 AGI timeline, the CCP espionage risk at AI labs, the parallels to the Manhattan Project. The nuances matter. Go listen to the interview. But the frame underneath all of it is still Jevons: the use cases we can imagine today are a rounding error against what comes next.
The bears are making the light bulb argument. And it has been wrong every single time in the history of technology.
I'm not making a stock call. I'm saying the analytical frame is broken. Efficiency unlocks demand -- that's a 160-year-old pattern. The people selling infrastructure stocks on an efficiency thesis are the same people who would have shorted oil refineries in 1890 because the light bulb made kerosene lamps obsolete.
Not the efficiency gains. Every time compute got cheaper, we used more of it. This time is not different.
g.money / @gmoneyNFT · curated by mikelitman.me