Unhedged announced markets tracking GPU rental prices on November 1. Source. Graphic by CantonNews.
How much will it cost to rent the chips used to run AI workloads? Unhedged has made that question the basis of a prediction market. On November 1, the project announced monthly and quarterly markets on GPU rental prices. When CantonNews checked its public Compute page on November 2 at 00:28 UTC, the page showed zero open markets.
What would decide the outcome?
A GPU is a processor that can perform many calculations at once. Renting time on one gives a customer computing capacity. An Unhedged contract instead lets a participant take a position on the future rental price. The reference is Flopmark, an index that tracks rental rates across seven GPU products and updates about every two minutes, according to Unhedged’s market guide.
Each contract has a target set in advance. At expiry, the index has to finish strictly higher than that target for Above to win. A reading equal to the target settles Below, just as a lower reading does.

Illustrative index values show how a tie settles under the published Compute rules. The example uses an invented target. Graphic by CantonNews.
For example, suppose the target were 100. The index could rise from 97 to 100 during the contract and the Below side would still win. What matters is where the index finishes relative to the agreed target, even if the rental rate for one particular chip moves in the other direction.
What the public listings showed
Unhedged had previously set November 1 as the target date. Its announcement that day described the markets as live. The next morning’s checks established what readers could see at that moment. Alongside the empty Compute page, the public feed returned 33 active entries across the platform. None identified a GPU contract.
There was therefore no public listing in those checks from which to quote a current target, expiry or price. The announcement establishes when Unhedged said the product went live. The rules and any open contracts can be checked on its Compute page.



