Cotton as a Fabric

The machine consumption model

Cotton is the original proof case for a simple economic model: D = N × R. Demand for a commodity equals the number of machines consuming it, multiplied by each machine's consumption rate. When you know a machine's throughput, you can predict the upstream market expansion before it happens.

In 1793, a single cotton gin could clean 50 lbs of lint per day. A human hand-cleaner managed 1 lb. That 50x throughput multiplier, deployed across thousands of gins, predicted everything that followed: US cotton production went from 3,000 bales in 1790 to 4,000,000 bales in 1860. A 1,333x expansion in 70 years. The gin's consumption rate was the demand function for American cotton.

This is the same model that predicts GPU power consumption will drive US data center electricity demand from 61.8 GW (2025) to 134.4 GW (2030). The commodity changes. The machine changes. The math doesn't.

A common objection: this model is too simple. What about credit systems, land policy, imperial trade, labor regimes, transport infrastructure? Didn't those factors drive cotton's expansion as much as the gin did?

No. And that's the point. Once a machine begins consuming a commodity at scale, the machine's throughput becomes the gravitationally dominant force in that market. Everything else orbits it. The cotton gin didn't just participate in the cotton economy. It was the cotton economy. The gin's throughput created a demand signal so powerful that the American South built the institutional arrangements to satisfy it: slavery expansion, Indian Removal, credit networks from New York to New Orleans, shipping routes, and an entire political ideology to defend the system that fed the machine.

This is not unique to cotton. Bessemer's converter (1856) produced steel at ~700x the rate of puddling furnaces. US steel went from 2,000 tons to 11.2 million tons in 33 years — and the Mesabi Range, the Great Lakes shipping fleet, and Carnegie's vertical empire were all supply-side responses to converter-created demand. The Model T's fuel consumption rate multiplied by 23 million registered cars by 1930 predicted petroleum demand so accurately that an entire geopolitical order — the oil century — reorganized around it. Haber-Bosch predicted synthetic nitrogen fertilizer from zero to 100+ million tons per year, creating modern industrial agriculture as a supply-side response to a chemical reactor's throughput. In each case: know R, count N, and you have the demand function. The institutional arrangements follow.

To identify the causal chain is not to excuse the moral choices made within it. The gin created an incentive; slaveholders chose to fill it with slavery rather than alternatives that existed. The machine set the trajectory. Existing institutions shaped the path. D = N × R tells you the scale and direction of the demand signal. It does not tell you whether the supply response will be a regulated industry or a forced-labor camp — that depends on the institutional landscape the model deliberately excludes. For predicting the commodity market, those details are downstream. For the human beings inside them, they are everything.

Credit, land policy, imperial trade, labor regimes — these are real and historically important. But they are supply-side responses to machine-created demand. They are the economy reorganizing itself around a new throughput reality. If you want to predict the scale and direction of a commodity market, you need N and R. The rest follows.

The throughput table

Every major machine in cotton's 260-year industrial history created a step-change in consumption that predicted the upstream market expansion:

Between 1780 and 1820, the combined effect of these machines increased productivity in Britain's cotton factories 370 times.

Try it: the D = N × R calculator

The gin-slave-cotton triangle

The cotton gin is the clearest historical proof of machine consumption as demand function. The gin's throughput didn't just predict cotton volumes — it predicted slave prices, because the gin shifted the binding constraint from processing to growing.

By the 1850s, slave prices were rising even as cotton prices remained moderate. The gin had created so much processing capacity that the binding constraint was no longer ginning but labor supply for growing and picking. More gin capacity → more land under cultivation → more slaves needed → higher slave prices. The machine's throughput predicted the labor market.

Cotton represented ~60% of total US exports by 1860. The economic "value" of enslaved people as property exceeded the combined value of all US railroads and manufacturing capital at the time of the Civil War.

The adoption lag

Machine consumption models also predict timing. The average time from introduction to full adoption across cotton's machines is ~35 years:

The cotton gin was fastest (15 years) because it solved an immediate, obvious bottleneck. The power loom was slowest (75 years) because it required complementary innovations — improved yarn, factory infrastructure, worker training. The machines that solve binding constraints adopt fastest.

225 years of cotton prices

Cotton has one of the longest continuous price series of any commodity. In nominal terms:

In real (inflation-adjusted) terms, cotton has declined ~90% from 1800 to 2024. The 2011 record of $2.20/lb (March Cotton No. 2 futures) was, in real terms, only ~1/5th of the 1800 price. Cotton has exceeded $1.00/lb only three times in 230+ years: the Civil War blockade, the 2010-11 commodity spike, and the 2022 post-COVID inflation.

Explore: 225 years of cotton prices

Decade averages

The market today

The cost stack

A pound of US cotton lint costs $0.71-1.10 to produce (net of cottonseed credits). At the current farm price of $0.64/lb, most American cotton farmers are losing money.

Cost of production varies enormously by country:

The true cost gap

The market price of cotton ($0.65/lb in 2025) reflects neither the full cost of production nor the full cost of environmental damage. The gap:

True cost is 2.4x to 5.8x the market price. If Ogallala Aquifer water were priced at its true replacement cost, West Texas cotton alone would cost $4-5/lb more. The aquifer, once depleted, will take over 6,000 years to recharge naturally.

The value chain: who captures what

Raw cotton at $0.64/lb becomes a $30 retail t-shirt. The farmer captures 1.1%. The value chain for a basic cotton t-shirt (160 GSM single jersey, made in Bangladesh):

The cotton farmer ($0.32) plus the garment worker ($0.50-1.00) together capture 2.7-4.4% of the retail price. The brand and retailer capture 53-100%. From academic research: the retailer value-add per kg is 92x the farmer's value-add per kg.

A $0.10/lb change in cotton prices adds $0.05 to a t-shirt's raw material cost, ~$0.03 to FOB, ~$0.01 to wholesale, and essentially $0.00 at retail. Cotton is 1-3% of the retail price. This is why retail prices don't track cotton prices — and why cotton price spikes trigger permanent substitution to polyester rather than retail price increases.

Try it: follow your t-shirt

The polyester substitution

Cotton's share of global fiber production fell from 68% (1960) to 19% (2024). Polyester grew 97x in the same period, from 0.8 MMT to 78 MMT. Cotton grew 2.5x.

Polyester crossed cotton in 2003. Substitution accelerates sharply when the cotton-to-polyester price ratio exceeds ~1.5:1. At 2:1 (as in 2011, when cotton hit $2.20/lb while polyester stayed near $1.00/lb), substitution becomes rapid and partly irreversible. The 2011 spike permanently accelerated the shift — cotton's share has fallen every year since.

This is itself a machine consumption story. Polyester is a petroleum derivative. The petrochemical refinery is a machine that consumes crude oil and outputs, among other things, polyester fiber. When that machine's output is cheaper than cotton, it captures cotton's market. The substitution is governed by relative machine economics.

Try it: the substitution threshold

Subsidy distortion

Global cotton subsidies total $6-8 billion annually. They depress world prices by an estimated 10-26% (ICAC, World Bank). The major subsidizers:

At peak US subsidy levels (2001-2005), the government was paying cotton farmers ~$0.30-0.40/lb — often exceeding the market price of cotton itself. China's domestic price target of 18,600 CNY/MT translates to ~$1.17/lb, roughly double the world price. West African farmers — among the world's most cost-competitive producers — receive 60% of the world price while subsidized competitors receive 130-170%.

The WTO Brazil-US cotton dispute (DS267) authorized $829 million/year in retaliation. The US settled for a $300 million one-time payment. The C-4 African countries (Benin, Burkina Faso, Chad, Mali) — third-party plaintiffs whose economies depend on cotton for 30-50% of export earnings — received $0.

Water: the unpriced input

Cotton requires ~10,000 liters of water per kilogram of lint. One t-shirt requires ~2,700 liters. Global cotton production consumes 256 billion cubic meters of water per year.

In most producing regions, water is priced far below replacement cost. The Ogallala Aquifer — which underlies 174,000 square miles across eight US states — has dropped 44 feet on average, with declines exceeding 300 feet in the most heavily pumped areas. Natural recharge is ~0.5 inches per year versus extraction of several feet per year. Cotton is increasingly the "crop of last resort" as aquifer decline makes corn irrigation uneconomical — cotton's 12-24 inches per acre versus corn's 24-30 extends the productive life of wells.

The Aral Sea — formerly the 4th largest lake in the world at 26,300 square miles — lost over 90% of its area after the Soviet Union diverted ~75% of river flow to irrigate cotton. 40,000-60,000 fishermen lost their livelihoods. It is considered one of the worst human-caused ecological disasters in history, driven almost entirely by cotton production quotas.

Production and trade

Brazil surpassed the United States as the world's largest cotton exporter in 2023 (14.3M bales vs 11.5M). The dominant trade flow is Americas-to-Asia: US Gulf ports and Brazilian ports ship to mills in Bangladesh, Vietnam, China, and Pakistan. Asia imports 69.5% of all globally traded cotton.

Modern machine consumption

A single modern mega-mill with 500,000 ring spindles consumes ~225,000 kg of cotton per day — approximately 365,000 bales per year, or about 2.5% of global production. The machine's consumption rate directly predicts the raw cotton market.

Spinning costs vary by country primarily because of labor. Vietnam ($1.19/kg) is 2.4x cheaper than Italy ($2.85/kg). Labor drives the gap: Italy's spinning labor cost alone ($0.97/kg) is 48x Bangladesh's ($0.02/kg).

The parallel: GPUs as cotton gins

The structural parallel between the cotton gin and the GPU rack:

The IEA projects global data center electricity consumption will double by 2030 to 945 TWh, growing at 15% annually — four times faster than all other electricity consumption growth. This is the modern version of British cotton imports growing at 10.8% annually from 1780-1800.

The cotton gin created a 1,333x expansion over 70 years. The GPU is creating a 2-3x expansion over 5-7 years in energy demand — smaller magnitude but much faster, compressed by modern capital deployment speed. The model is the same: D = N × R. Know the machine's consumption rate, know the number of machines, and you can predict the commodity market upstream.

Home