Denim and silicon

On the last page of McTeague, Frank Norris’s San Francisco novel, an illiterate dentist dies handcuffed to his pursuer in Death Valley, wearing gold he will never spend. Erich von Stroheim filmed it as Greed.

Half a century earlier Levi Strauss landed in San Francisco in 1853. He never touched a gold vein: he sold cloth and tents. The textbook line — the rush enriches those who sell picks — is true, with a complication.

The third character intro courses skip is Josiah Royce, born in 1855 in Grass Valley. In 1886 he published California: A Study of American Character, asking what that fever had done to a society’s moral character.

His answer was that the absence of settled institutions — no stable courts, no secure property, no communal memory — had unleashed the worst as much as the best: unchecked individualism, summary justice, intolerance of anyone foreign to the camp.

Royce condemned the illusion that energy alone produces moral order. You need institutions, not only enthusiasm. AI’s story is also a story of institutions that are not yet there: a mining camp without a sheriff.

isaiprofitable.com tracks the gap between AI investment and revenue: about $1.4 trillion spent against $613 billion booked. Amazon leads spending. The only clear black line is Nvidia — which does not build AI; it sells the hardware everyone else burns cash on.

The dot-com comparison, also in Carlota Perez’s work, helps: that bubble burned capital but left fibre and data centres. Perez distinguishes “installation” (bubble that builds infrastructure) from “deployment” (when the technology becomes fact and fortunes often go to others).

Thousands of kilometres of fibre laid and never lit in the dot-com rush became, twenty years later, the highway Netflix and YouTube travel on. Those who paid for those cables were, in most cases, long bankrupt.

The $1.4 trillion burned can be read as the cost — usually paid by those who will not benefit — of compute and data centres someone else will use in ten years. Today as in 1849, the safest seat looks like the seller of tools. Nvidia plays Levi Strauss.

But Nvidia does not fabricate chips: TSMC does. Every accelerator that matters in 2026 passes through Taiwanese plants. Under the pick seller sits a steel seller, and under that a silicon mine. Nvidia’s biggest customers design their own chips; its inference share may fall sharply by 2028.

The tool seller who seemed outside the fray turns out to be inside it like everyone else: they only bet on a different table.

Strauss did not invent jeans: Jacob Davis in Reno did in 1872, and partnered with Strauss for the patent money. They filed together in 1873. Strauss had been inside the fever for twenty years; his fortune came from a less volatile bet taken with someone who had the idea but not the means.

Who is Davis and who Strauss today? Labs that burn cash, or those who own infrastructure and can wait? Royce would add that without institutions able to tell productive energy from pure fever, even the richest field risks ending like Death Valley: gold that no longer serves anyone.