Who Got Rich Off the Punch Cards?
5 min read

›Contents
Three kinds of people touched the punched card in its 150-year reign. They came out of it very differently, and the differences are about to matter again, because we are re-running the experiment.
The piqueurs#
In the Lyon silk trade, the piqueurs were the workers who punched the cards: thousands of holes, card after card, transcribing a designer's pattern into machine-readable form.1 It was skilled work, the kind that took training and real fluency in the encoding, and it paid a real wage.
And it stayed a wage. Piquage never became an asset. The piqueur could not accumulate anything from punching: each card belonged to the workshop, each pattern to the designer or the merchant. The skill was real but the skill was translation, and translation, however difficult, is work-for-hire by its economic nature. When card-punching machinery improved, the wage fell. When it automated, the trade evaporated. Nobody inherited a piquage fortune.
The piqueur's modern descendant is not the data-entry clerk. It is everyone whose job is transcribing intent into machine-executable form by hand. The piqueur is the programmer, the prompt-writer, the workflow-builder. Difficult translation is still translation.
The inventor#
Jacquard himself did meaningfully better, but look at the form his returns took: the city of Lyon granted him a pension and a royalty on machines sold in France.1 He was compensated by institutions, as a recognition of contribution, not enriched by ownership of the system his invention created. The loom attachment was copied across Europe within his lifetime, mostly beyond his reach. Invention without ownership of the ongoing flow earns honors, and honors are not a business model.
The man who owned the reader#
Herman Hollerith is the interesting case. Hollerith did punch cards; his company employed plenty of card-punch operators. But that is not what he owned. Hollerith owned three other things:
-
The format. The card layout, what each column and hole position meant, was his design. Every census form, every railroad waybill, every insurance record that wanted machine tabulation had to be expressed in his encoding.
-
The reader. The tabulating machines that could ingest the cards were his, and the masterstroke was that he leased them rather than selling them. The 1890 U.S. Census didn't buy tabulators; it rented counting capacity.2 Recurring revenue, tied to volume, from a machine only he could supply.
-
The count. This is the subtle one. What the census, the railroads, and the insurers were actually buying was not holes in cardboard. It was a number they could trust: a total that would be accepted as the total. Hollerith's machines were the mechanism, but his real product was the institutional acceptance of the result. When the count was disputed, it was re-run on his machines, in his format. He was the party everyone had agreed would say what the cards added up to.
The Tabulating Machine Company merged into what became IBM, and the model (own the format, lease the reader, charge on volume, be the accepted arbiter of the count) powered the largest information business of the twentieth century. The cards themselves were a consumable. IBM sold billions of them at commodity margins.3 The margin was never in the punch.
The lesson#
In any translation layer between human intent and machine execution, value settles in a predictable place, and it is never where the labor is:
- Punching is a wage. However skilled, however scarce today, the act of transcription commoditizes with the tooling. Its price trends toward zero and its practitioners hold nothing when it does.
- Invention is a pension. Building the clever mechanism earns a one-time reward unless it converts into ownership of an ongoing flow.
- The format, the reader, and the count are the business. Whoever defines the representation, controls the machinery of ingestion, and is trusted to pronounce the result captures the compounding value, and gets paid per unit of volume, forever, on a flow that only grows.
Running the experiment again#
Generation by AI is the new punching, except this time the piqueur costs nothing at all. The volume of machine-produced work is about to do what tabulated records did after 1890: grow by orders of magnitude, because the cost per unit collapsed. Every economic instinct trained on the last technology cycle says the money is in producing the output: better, faster, cheaper generation.
The punch-card century says otherwise. When production is abundant, the scarce positions are: the format in which intent gets expressed precisely enough to execute; the reader that ingests the flood and processes it at scale; and above all the count: being the party whose statement of "here is what this adds up to, and it is right" is accepted by people who have money at stake.
Watch where today's builders are crowding: into generation, the punch. The uncrowded ground is the other side. Somebody is going to be the accepted arbiter of whether the flood of machine-made work is actually correct, per unit, on volume, in a format they define. That position looks small now, the way a census-counting contraption looked small next to the industries it counted. It compounded into IBM.
The Operator's Move#
The free move this quarter: inventory your translation layers. List the places in your business where someone's job is converting intent into machine-usable form: reports built by hand from other reports, reconciliations, data re-keyed between systems. For each, ask Hollerith's three questions: Who defines the format? Who owns the machinery that reads it? And whose statement of the total do we actually accept? Where the answer to all three is "a vendor," that is a dependency worth knowing about. Where it is "nobody, really, we just trust whoever did it," that is where the next decade's leverage will come from, or be taken from you.
References#
-
James Essinger, Jacquard's Web: How a Hand-Loom Led to the Birth of the Information Age (Oxford University Press, 2004). ↩ ↩2
-
Geoffrey D. Austrian, Herman Hollerith: Forgotten Giant of Information Processing (Columbia University Press, 1982). ↩
-
Emerson W. Pugh, Building IBM: Shaping an Industry and Its Technology (MIT Press, 1995). ↩