The Signal and the Sale

10 Jul, 2026

There is a peculiar thing about prediction markets. When you buy a contract on Polymarket (say, "Will the Fed cut rates in June?"), the price you pay is not informed by an existing data feed. The price is the data. Every trade produces the public signal; there is no separate information product underneath it. The market for information and the information itself are the same transaction.

This feels like a radical inversion of how information commerce is supposed to work. But it is actually the endpoint of a 300-year process in which the boundary between information about markets and the market itself has been steadily dissolving. The signal and the sale have been merging for centuries.

Stage 1: Embedded

The coffee house era, c. 1688–1800

Edward Lloyd opened his coffee house on Tower Street in London around 1688. He sold coffee, but the real product was the room itself. Lloyd catered to sailors, merchants, and ship-owners by providing reliable shipping news, and the coffee house became the place where marine insurance was negotiated. You did not buy a data feed from Lloyd. You bought a seat and got access to a community whose collective knowledge (which ships had sailed, which had returned, whose cargo was reliable) was the most valuable information in maritime commerce.

This was information fully embedded in a social network. Its value came from exclusivity: who else was in the room, not how widely the news was distributed. There was no product to separate from the people who produced it.

The telegraph changed that. Once information could travel at the speed of light rather than the speed of a horse, the room was no longer the container. The signal could be separated from its source, and for the first time, information became something you could buy and sell on its own.

Stage 2: Separated

The news agency era, 1851–1870

Paul Julius Reuter established his news service in London in 1851, using the newly laid Dover-Calais telegraph cable to transmit stock prices between London and Paris. Before the cable was fully operational, he used homing pigeons between Brussels and Aachen. Reuters, along with Havas in Paris (1835) and Wolff in Berlin (1849), invented the first pure information commodity: text produced in one place, transmitted electrically, and sold simultaneously to multiple subscribers.

This was a clean separation. Reuters sold news about markets: closing prices, political developments, ship arrivals — to banks, brokerage houses, and newspapers. It did not trade. It did not execute. It produced and transmitted information, and charged for access. The signal had been detached from the social network and placed on the wire.

The value of this separated information was tied almost entirely to speed. A 24-hour head start on financial news from the Continent was worth a subscription fee. Delay meant the information was worthless. The business of selling information had become a business of selling time.

Stage 3: Streamed

The ticker era, 1867–1920

The stock ticker, invented by Edward Calahan in 1867 and perfected by Thomas Edison, was the first continuous real-time data feed. A relentless stream of every trade as it happened, transmitted over telegraph wires and printed on paper tape. The New York Stock Exchange licensed the data to the Gold and Stock Telegraph Company, which sold it to brokerages. The data feed as a subscription product was born.

Here is where the boundary between signal and sale first started to blur. Bucket shops (betting parlors that subscribed to the ticker and let people gamble on stock price movements without owning the underlying shares) were a direct derivative of the information feed. The NYSE sued to shut them down, arguing that ticker data was licensed for information purposes, not for creating competing markets. The courts eventually ruled against the exchanges. The signal had become a sale.

The ticker created a strange duality. The price feed was simultaneously:

Information about the market (a legitimate data subscription)
A market in itself (bucket shop bets on the feed)
The infrastructure of the market itself (no exchange could operate without it)

This duality would only deepen.

Stage 4: Bundled

The terminal era, 1981–present

Michael Bloomberg's Terminal, launched in 1981, was the first product that let you see a price, analyze it with models, read news about it, and execute a trade — all on one screen, on one keyboard. Earlier financial data systems like Reuters and Quotron were displays. Bloomberg was a trading tool.

The terminal collapsed several layers that had been separate: data provision, analytics, news distribution, and trade execution. A Bloomberg subscription is simultaneously a window onto the market, a tool within the market, and a door into the market. (The Bloomberg chat network is where a significant fraction of bond market deals are negotiated.) The signal and the transaction live in the same box.

At this stage, the "information product" has become practically indistinguishable from market participation itself. If the Bloomberg network goes down, trading stops. The signal has become the circulatory system of the market.

Stage 5: Collapsed

The prediction market era, 2020–present

Prediction markets complete the process that the ticker started and the terminal accelerated. On Polymarket or Kalshi, there is no separate data feed to subscribe to. There is no terminal to lease. Every trade produces the information. The price of a contract on a future event is the aggregated belief about that event, and the price itself is the visible record of that belief.

This is what the InfoFi (Information Finance) narrative describes: prediction markets as the world's fastest financial data transmission mechanism. In January 2026, the industry hit $5.23 billion in weekly trading volume. The boundary has fully dissolved. You cannot point to "the information" and "the market" as separate things. They are the same transaction, observed from different angles.

This creates an odd inversion of the historical pattern. In Stage 2, Reuters produced information and sold it to traders. In Stage 5, the traders produce the information and the platform (Polymarket, Kalshi) captures the value of the data they generate. The crowd replaces the news agency, but the intermediation remains.

Stage 6: Metamorphosed

The compute era, 2026–

In May 2026, the CME Group announced plans to launch futures contracts tied to the price of GPU computing power, developed in partnership with the data firm Silicon Data. By June, asset managers including ProShares and Rex Shares had filed proposals for exchange-traded funds tied to these contracts. AI compute (the raw processing capacity to train and run models) is becoming a standardized, exchange-traded commodity.

This is not a market for information. It is not a market about information. It is a market for the capacity to produce information from data. Compute is the substrate on which all signal extraction runs. A futures contract on H100 rental rates is a tradeable instrument on the thing that discovers the signal in the first place.

The arc closes:

Embedded: Information was in the room
Separated: Information was put on the wire
Streamed: Information became a continuous feed
Bundled: Information was fused with execution
Collapsed: Information and market became one
Metamorphosed: The capacity to process information itself becomes the tradeable asset

What the Dissolving Boundary Means

Three observations that survive the 300-year arc:

First, intermediaries never disappear; they just change form. The coffee house owner became the news agency became the exchange became the terminal provider became the platform. Each stage disintermediates the previous one and reintermediates in a new shape. The prediction market platform is the new Bloomberg terminal. The compute exchange is the new stock exchange. The role of the intermediary shifts, but someone always sits between the signal and the sale.

Second, the speed premium keeps growing, but the unit of speed changes. In 1851, a 24-hour head start on news from the Continent justified a subscription. By 1920, minutes mattered. Today, in high-frequency trading, microseconds matter, and the premium has shifted from who receives the information first to who processes it fastest. Compute futures are the logical endpoint: a market on the speed of processing itself.

Third, the three properties that started this research (information as discoverable, computable, tradeable) map directly onto this arc. The signal is discovered (coffee house gossip, news reporting), then computed (ticker data, terminal analytics), then traded (prediction markets), until finally the capacity for discovery itself becomes the tradeable asset (compute futures). The boundary between the signal, the computation, and the sale keeps dissolving until they become the same substance.

One second take-away

Edward Lloyd would recognize what is happening at Silicon Data. He knew that the most valuable thing you could sell was not information. It was access to the capacity to produce information. He sold a seat in a room where shipping news was born. Silicon Data sells a contract on the machine that extracts signal from noise. The room has become a futures contract, the shipping news has become a probability distribution, and the coffee has become a GPU rack. But the impulse (to discover the signal, to compute its meaning, to trade on what you find) is the same impulse that sent merchants to Tower Street in 1688.

The coffee house is still there. It has just changed shape.

#finance #history #information #markets #prediction-markets