Why Perps?
From crypto's favorite derivative to the financial primitive of the AI economy
What do the dot-com bubble, the 2008 housing bubble, and crypto’s most successful market structure have in common?
Robert Shiller.
Shiller is best known as the economist who warned about both the late-1990s tech bubble and the mid-2000s housing bubble. Less well known is that in 1993, he proposed a new financial derivative, one that would take a few decades to realize its full potential: the perpetual futures contract.
Ordinary futures have expiration dates. You buy the March contract, then the June contract, then the September contract, rolling your positions as time passes. Shiller saw an opportunity to simplify this, and introduced the perpetual future, which eliminates expiration dates and keeps a single contract alive indefinitely. His proposal received some academic attention but no immediate market adoption.
Today, the adoption has arrived, to the tune of $93 trillion in crypto trading volume this year alone. That is more than three-quarters of all crypto trading activity, and several times the size of the spot market.
To be fair, Shiller’s original idea was not “let’s help people lever up 50x on Bitcoin.” It was closer to the opposite. He was thinking about assets that mattered enormously but were hard to hedge cleanly: home prices, GDP, labor income, and other forms of diffuse value with weak or imperfect price discovery. As he put it in 1993:
“The greatest components of world wealth are not hedgeable at all.”
That observation is becoming more relevant in the age of AI, and if AI really does reshape the world, it will soon apply to the wider economy too. Compute costs, frontier-model performance, and private AI lab valuations already drive hundreds of billions of dollars in capital allocation decisions, but there is still no straightforward way to hedge those risks.
Perps in brief
A perpetual future or perp is a derivative contract that tracks the price of an asset or index. If you buy one perpetual contract on Bitcoin, you have long exposure to Bitcoin’s price. If Bitcoin goes up, you make money. If it goes down, you lose money. So far, so good.
What makes a perpetual unique is how the contract stays anchored to the underlying asset, through what is known as the funding rate.
The funding rate works like this: when the perp price trades above spot, longs pay shorts. When the perp price drops below spot, shorts pay longs. These periodic payments create an incentive for traders to push the perp back toward the underlying price. If the perp drifts too high, it becomes expensive to stay long, which encourages selling or arbitrage. If it drifts too low, the reverse happens.
That’s it. No expiry. No rolling from one contract month into the next. Just one live market and a built-in mechanism that penalizes it for wandering too far from reality.
Why this is such a big deal
There are at least three ways to think about a perpetual future.
The first is as a futures contract without the annoying parts. Traditional futures fragment liquidity across a March contract, a June contract, a September contract, and so on. Traders have to choose which one to trade, manage basis, and roll positions forward before expiry. Professionals handle this just fine, but it is operationally messy and cognitively expensive. A perpetual future collapses all of that into a single market.
The second is as a leveraged spot substitute. Many traders do not want to buy and custody the underlying asset. They want liquid exposure, leverage, and the ability to go both long and short. Perps offer that in a particularly clean package. Instead of borrowing the asset, borrowing dollars, worrying about recalls, and managing a tangle of financing plumbing, you trade one instrument and let the funding mechanism handle the economics.
The third — and probably most important — is as a way to concentrate liquidity. Markets tend to work better when everyone is looking at the same screen. Instead of ten semi-liquid markets, you get one very liquid one. That alone is often enough to create a self-reinforcing advantage. This is why perps did not merely coexist with dated futures in crypto; they mostly ate them.
The irony
Shiller imagined perpetual futures as a solution for assets that were systemically important but difficult to observe or hedge. Crypto turned them into the dominant derivative for one of the easiest assets in the world to observe. Bitcoin has thousands of market participants, continuous price feeds, deep spot liquidity, and instant global dissemination of price information. It is almost the opposite of the kind of underlying Shiller had in mind.
He envisioned perpetuals as the natural derivative for assets with imperfect price discovery. In practice, they succeeded first in assets with excellent price discovery. That is not an accident. A perpetual contract does not need a perfect spot market underneath it, but it does need something solid to anchor to. In practice, that usually means at least one of three things: a credible spot market that traders can observe and trade against; a credible index or settlement process that everyone accepts as close to reality; or enough arbitrage and informed trading that deviations get corrected quickly. Without at least one of those, the funding mechanism has nothing to anchor to. It can penalize divergence, but it cannot define the price it is supposed to converge to.
Why this matters for AI
The AI economy is full of risks that are large and hard to hedge. Unlike a few years ago, some of them are also starting to become legible enough to index. Semianalysis’s H100 price feed, for example, is the kind of reference process that can turn “GPU scarcity” from a vague industry complaint into something a market can actually settle against.
Consider the different sides of the market that already exist. Neoclouds financing large GPU buildouts want to hedge their future compute exposure. If they are committing real capital to racks of H100s or B200s, they are implicitly long on GPU pricing and utilization, and will want to go short. On the other side, a company training a model is exposed to increases in spot compute prices and will want to go long. An investor in a private model company may have a strong view on relative benchmark progress but no clean way to express it. A startup building on top of a single model provider may be taking concentrated platform risk without an obvious hedge. These are balance-sheet risks that remain mostly unmanaged because the market structures for them barely exist.
If something matters enough, eventually people want to insure against it, speculate on it, finance against it, and use it as an input into planning. But that only happens once there is a contract simple enough to trade and a reference process good enough to anchor it.
From crypto’s favorite derivative to a general market primitive
The big picture is that perpetual futures have already crossed the line from “clever market hack” to “general-purpose financial primitive.”
Crypto proved that a non-expiring, funding-anchored contract can scale to enormous volume. The question now is: where next? Shiller’s answer, three decades ago, was: wherever the exposures matter most and are hardest to hedge. The AI economy now has variables that look a lot like what he had in mind: economically central, difficult to package, important for capital allocation, and increasingly backed by legible indices. If perps became crypto’s dominant market structure by making one volatile asset easier to trade, they stand to be even more important when applied to entirely new classes of exposure that are not yet tradable at all.
Intrigued? Catch a glimpse of the future of perp trading at testnet.mnx.fi.