What grain, metals and oil teach us about pricing intelligence.

Every great commodity eventually needed a trusted price. The pattern is remarkably consistent.

Carry Research7 min read

Grain had Chicago. Metals had London. Oil had New York.

Each became a commodity the world could plan around only after three things fell into place: an agreed way to describe what was being traded, a price people trusted, and a market where buyers and sellers could meet at that price. The order rarely changed.

The stories are worth retelling, because the market for intelligence is at the start of the same journey. The details differ, but the problems these markets solved are recognizably the problems of AI inference and compute today.

Chicago and the problem of the harvest

In the mid-nineteenth century, grain poured into Chicago from the Midwest. Railroads and canals brought it to the city faster than ever before, but the rhythm of farming had not changed. Farmers brought their crops to market at the same time of year, prices collapsed at harvest and recovered months later, and both farmers and merchants carried large, unpredictable risks. Some years, grain was dumped because buyers could not be found at any reasonable price.

The Chicago Board of Trade, founded in 1848, gave the trade a central place to agree prices in advance through forward contracts. A merchant could agree in winter to buy grain for delivery in spring, at a price fixed today. That alone reduced the risk of the harvest glut. But each forward contract was still a private agreement about a particular lot of grain, and its value depended on what exactly would be delivered.

Over the following years the exchange did something less visible but more important: it standardized grain into defined grades. Grain from many farms was graded and stored together in elevators, and a receipt for a given grade meant the same thing to everyone. By the 1860s, standardized contracts were trading on the exchange, the forerunners of modern futures.

Standard grades made grain interchangeable. Interchangeable grain could have a single quoted price. A single quoted price, carried across the country by telegraph, let a farmer, a miller and a shipper agree on value without inspecting every sack.

London and the metals of an industrial age

Metals followed a similar path. As industry demanded copper and tin from mines on the other side of the world, in Chile and Malaya, merchants needed a reliable way to price cargoes that would arrive months later. A ship's cargo could lose much of its value at sea if prices fell before it docked.

The London Metal Exchange, founded in 1877, gave the trade standard contracts and official prices. Its three-month contract is often said to reflect the time it took to bring copper from Chile and tin from Malaya to London by sea, so a merchant could sell a cargo for delivery on arrival and fix its value on departure. The exchange's trading ring, where dealers sat in a circle and traded by open outcry, became one of the best-known institutions in London's financial life.

Those official prices became references that producers and consumers far from London wrote into their own supply agreements. A mine in one continent and a factory in another could agree to trade at the London price, adjusted for quality and delivery, without either of them trading on the exchange.

The lesson from London is that a benchmark travels. Once a price is trusted, it is used well beyond the trading floor that sets it, in physical contracts between parties who may never trade on the exchange at all.

New York and the price of a barrel

Oil is the most recent of the three. For much of the twentieth century, the price of crude was set largely through long-term arrangements and posted prices, decided by the largest producers and, from the 1970s, increasingly by producing countries. The oil shocks of that decade showed how quickly the price could move, and how exposed refiners, airlines, utilities and whole economies were to it.

In March 1983 the New York Mercantile Exchange launched a futures contract on light sweet crude, known as WTI, delivered at Cushing, Oklahoma. It started slowly. As oil prices became more volatile, producers and consumers began to use the contract to protect themselves, and volume grew rapidly through the decade. A few years later, a futures contract on Brent crude from the North Sea gave the rest of the world a second reference.

Today WTI and Brent are among the most widely watched prices in the world. Crude from fields everywhere is bought and sold with reference to them, at a premium or discount that reflects its own quality and location.

The pattern is consistent: define the thing, trust the price, then build the market around it.

The same three steps

Look across the three stories and the same sequence appears.

  • A standard. Grain grades, metal specifications and crude qualities turned varied physical goods into something that could be described in a contract.
  • A trusted price. Exchanges published prices that both sides accepted as fair, because the rules for setting them were open and consistently applied.
  • A market at that price. With a standard and a trusted price in place, liquidity followed, and with it the ability to hedge, finance and plan.

Volatility tends to accelerate the last step. Markets grow fastest when the people exposed to a price most need protection from it. Grain futures grew out of harvest gluts, metals contracts out of long and risky voyages, and oil futures out of the price shocks of the 1970s.

What the histories also show

Three quieter lessons sit beside the main pattern.

  • Standards evolve. Grain grades, metal specifications and crude benchmarks have all been revised as the physical market changed. A standard that never changes eventually describes something nobody trades.
  • Trust has to be maintained. A benchmark that loses the confidence of its users loses its market. Rules, oversight and transparency are what keep a trusted price trusted.
  • Benchmarks are sticky. Once contracts, systems and habits are built around a reference price, moving away from it is slow and costly. The choices made early in a market's life tend to last.

All three matter for a new market. They argue for getting the standard and the governance right from the start, rather than fixing them once contracts depend on them.

Intelligence is next

AI inference and the compute behind it are becoming the commodities of this century. Businesses now buy intelligence the way they buy energy: continuously, at scale, and as a growing share of their costs. Data centers finance years of capacity against prices that can change quickly.

The market for intelligence today looks much like grain before standard grades. There are many prices, quoted in different ways, for products that change every few months. There is no agreed unit and no reference that both buyers and sellers accept.

Intelligence also differs from the commodities that came before it, in ways that shape what its standard must look like.

  • The product improves. A bushel of wheat is the same each year; a model released this year can be far more capable than last year's. The standard has to hold capability constant, not a physical grade.
  • It is delivered digitally. There is no ship at sea or grain in an elevator. Inference is produced and consumed in the same moment, wherever the buyer is.
  • It is global from the start. Buyers and sellers are spread across the world from day one, rather than gathering in one city and spreading out from there.
  • It has two linked layers. Compute is the raw input, rented by the hour; inference is the finished product, bought by the token. Both need a price.

For inference, that standard is a fixed level of capability rather than a physical grade, so that the price compares like with like as models improve. Measuring price at constant capability explains why.

Building in the same order

Carry is following the order history suggests. Carry Indices publishes the reference prices: CICI for the price of inference at constant capability, and CRTI for the realized cost the market pays. Carry Exchange is building the market around them, where participants across AI inference, compute and infrastructure can meet at a trusted price.

Grain had Chicago. Metals had London. Oil had New York. Intelligence will have Carry.

This article is for information only and is not investment advice.

Newsletter

New research, in your inbox.