Perpetual Futures' 10% Annual Drain: A Funding-Rate Autopsy of The Economist's Warning
The data shows that a long position in a perpetual futures contract on a major centralized exchange pays approximately 10.95% of notional value per year before price moves. That is not a liquidation. That is not slippage. That is the fixed anchor rate inside the funding mechanism: 0.01% per eight-hour settlement, three times per day, 365 days per year. The Economist finally quantified this, and Crypto Briefing carried the warning. The number is real. I have spent years reading protocol accounting at machine level, from EVM opcode traces after The DAO to zero-knowledge circuit constraint audits. Code doesn't lie; audits do. The funding rate formula is simple enough to verify. The consequence is not.
Perpetual futures are the largest product in digital asset derivatives, representing roughly 80 to 90 percent of all derivatives volume. They have no expiration date. They are anchored to spot prices through a funding rate. The design dates to BitMEX in 2016. It has been running for eight years. This is not a nascent, untested mechanism. It is mature infrastructure with a mature cost structure.
The funding rate is a simple formula: anchor rate plus a premium or discount component. The anchor is often 0.01% per interval. At three intervals per day, the anchor alone creates a 10.95% annual cost. If the perpetual contract trades above the spot index, long positions pay short positions. If it trades below, shorts pay longs. The mechanism is designed to push the price back toward the index. Most explanations stop there. They treat funding as a zero-sum transfer between longs and shorts. The transfer is zero-sum. The structural outcome is not.
Over eight years, the market has seen long periods of positive funding. The long side is the payer. This is the point The Economist's calculation captures. The 0.01% anchor alone equals 10.95% annualized. In a balanced market, with no premium or discount, the long side still pays the anchor. In a bull market, when the perp trades above spot, the premium component is added to the anchor, and longs pay even more. In a bear market, shorts may pay, but retail traders are overwhelmingly long-biased. They are the natural payer.
The exchange does not collect the funding payment. This is one of the most misunderstood facts about perpetuals. Funding is paid by one side of the market to the other. The exchange earns trading fees. That distinction matters because it changes the incentive structure. An exchange that earns fees from open and close volume has no reason to care whether funding is fair. It has a reason to maximize turnover. High funding rates create discomfort, which creates closing transactions, which create fees. Protocol and user interests are not aligned. Trust is a bug, not a feature.
In 2017, I spent six months decomposing the EVM after the DAO hack. I disassembled thousands of lines of contract bytecode to find where high-level Solidity abstractions had hidden memory safety failures. The lesson was direct: high-level explanations lie; machine-level accounting does not. The same discipline applies to funding rates. The user interface says a trader can open a long position with leverage. The machine says there is a recurring carry obligation. The machine is right.
The 10% figure from The Economist is a lower bound, not a ceiling. It excludes trading fees. It excludes slippage. It excludes liquidation losses. On a major venue, the taker fee is 0.02 to 0.06 percent per side. For a trader who closes fifty positions per year, that is another 2 to 6 percent in fees. Maker fees may be zero or negative, but a maker must provide liquidity and accept adverse selection. Slippage is a function of order size and book depth. In less liquid perp venues, slippage can exceed 1 percent. Liquidation is the largest uncontrolled cost. A 10x position is liquidated after roughly a 10 percent adverse price move before maintenance margin. The funding cost reduces the buffer over time, making liquidation more likely. The combined annual carry on a long position ranges from 15 to 50 percent. The 10% number is the conservative floor, not the expected average.
Leverage converts notional cost into equity cost. A 10x long with an account of 10,000 US dollars controls 100,000 in notional value. If the funding rate is 10.95 percent annualized, the notional pays 10,950 per year. The equity is 10,000. The funding charge exceeds the entire account. The trader must add margin, reduce leverage, or close the position. This is not a theoretical outcome. It is the mathematical consequence of a recurring charge applied to notional, not equity.
From a capital allocation perspective, a perpetual long is a negative-carry position. A staked token can earn positive yield. A basis trade captures a positive spread. A perpetual long pays a repeatable financing charge. The expected return of a perpetual long is the asset's spot return minus carry. The historical spot return is positive over long intervals, but it is not smooth. The carry cost is smooth. In a bear market or a long consolidation, the carry cost runs while the spot return does not. The breakeven return is high.
The timing of funding costs is not evenly distributed across cycles. In a bull market, funding remains positive for weeks or months. Longs pay the anchor plus the positive premium. The actual annualized rate can exceed 30 percent. In a bear market, funding can flip negative. Longs receive payments, but the long is also losing from price depreciation. In a sideways market, funding hovers near the anchor. The Economist's estimate is a long-run steady-state cost, not a period measure. A trader who enters at a bull-market peak pays more. A trader who enters during a capitulation event may receive funding. The warning is about the long-run average, not the next hour.
The cost compounding has another consequence. A 10 percent annual charge is not linear after it is paid. Since the perpetual contract's position quantity is fixed by notional, the funding charge is deducted from the account's cash balance. If the price is flat, the trader must periodically add collateral or reduce leverage. At 1x leverage, the account loses 10 percent of equity per year. At 10x, the account loses 100 percent of equity per year unless the trader adds margin. The first liquidation is not a function of direction. It is a function of carry and volatility. This is a deterministic drain. It is not a tail risk.
Market structure reinforces this. Binance Futures holds roughly half of the centralized perpetual market. OKX and Bybit cover another 20 to 30 percent. Decentralized protocols such as dYdX, GMX, and Hyperliquid have smaller but growing shares. CME serves institutions with dated futures. The largest centralized venues rely on retail user growth. The Bank for International Settlements has estimated that retail accounts for more than 70 percent of crypto derivatives turnover. When a mainstream publication warns that the product drains long positions, it hits the exact user segment that supplies counterparty depth. The short-term result is not a crash. It is a risk-premium adjustment. Open interest falls, funding normalizes lower, and the spread between the perp and spot shrinks. The longer-term result is a shift in who uses leverage.
Decentralized perpetual protocols offer a different accounting layer. Funding rates, collateral, and liquidation prices are visible on-chain. For a user, this is materially better than a centralized black box. Zero knowledge, maximum proof. The correct standard for a financial primitive is not a blog post; it is a set of constraints that can be checked. On a decentralized exchange, a user can check. But transparency is not the same as fairness. In most decentralized perps, the parameters that determine funding, oracle constants, and leverage limits were set by a small founding team or by a governance process that tokenholders rarely use. Retail is the same passive counterparty. The venue moved, the structure did not.
Centralized venues can adjust funding parameters, change mark-price logic, or alter insurance fund rules. The user does not have a vote. This is the central governance failure. The DAO was a warning we ignored. It showed that the absence of a visible transfer does not mean value is not moving. Perpetual funding is not a reentrancy bug, but it is the same lesson: the accounting must be read before the position is opened. A market that cannot disclose its own cost structure is not a transparent market. It is a rent collection mechanism.
Perpetual exchanges also maintain insurance funds to absorb losses from liquidations. The insurance fund does not remove liquidation risk. It only socializes the systemic loss. When a cascade occurs, the insurance fund can be depleted and auto-deleveraging begins. This transfers value from profitable positions to cover insolvent positions. For a long-term long, auto-deleveraging is another unlabeled cost. It is not included in the 10 percent figure. It is a tail risk, but it is triggered more easily when funding costs are high and leverage is high.
The oracle and mark price are another failure surface. Centralized venues use their own index prices and determine the mark prices used for liquidation. Mark-price manipulation is less common than before, but the adjustment mechanism remains discretionary. In decentralized perps, the oracle is a separate failure surface. The oracle price is a constraint on the system. If the oracle fails, the entire position accounting fails. I have audited zero-knowledge verification where a public input mismatch allowed a false proof. The same class of error appears when the mark price is treated as a neutral input rather than an attack surface. The 10 percent cost calculation assumes the oracle is correct. That is not always true.
The Economist is not a regulator. It is a narrative-setting institution. The warning will be cited in policy papers. The United Kingdom already bans crypto derivatives for retail. The European Securities and Markets Authority has imposed leverage restrictions on CFDs. Singapore's central bank has capped retail crypto derivatives leverage. Australia has proposed restrictions. The 10 percent figure gives policymakers a clean number to quote. A cost-disclosure requirement is the most likely next step. In European retail investment products, a KID document must show risk and cost in monetary terms. If a similar requirement is applied to perpetual futures, centralized exchanges will have to publish funding history and expected carry. That will reduce volume. It will also reduce the information asymmetry that makes the product profitable for issuers.
The CFTC has already taken enforcement action against the derivatives arm of a major exchange. The Economist's narrative may accelerate that timeline. If the media story spreads from the Economist to the Financial Times, Reuters, and CNBC, the regulatory path is shorter. The product faces not one single ban, but a series of leverage caps, disclosure rules, and suitability limits. Each restriction removes a layer of retail participation.
The contrarian conclusion is that the warning can make the market less safe, not safer. Consumer-protection narratives assume that removing retail reduces harm. In derivatives, the opposite can happen. Retail is the main provider of counterparty depth. If retail leaves, order books become thinner, liquidation cascades become easier, and price gaps become larger. The market may become more institutional and more volatile in the tail. A reduction in leverage does not automatically reduce risk. It can reduce liquidity and increase the impact of a forced seller. We ignored the accounting lesson of The DAO once. The same lesson now applies to aggregate liquidity.
The more immediate paradox is that the warning is positive for the most regulated venues. A trader who leaves a high-leverage offshore perp is likely to move to spot, to dated futures, or to a regulated venue like the CME. The capital does not leave the asset class. It leaves the low-disclosure product. This is not a bearish signal for bitcoin. It is a bearish signal for the unregulated derivative layer. The product's market share will shrink before the underlying market does. This is the standard lifecycle of a financial product after a consumer-protection narrative solidifies.
An honest cost disclosure for a perpetual future would be simple to build. A protocol could display the trailing 30-day average funding rate, the historical maximum, the median, and the worst-case annualized cost. It could display the funding schedule in the trade ticket before entry. It could show the estimated annualized carry under three scenarios: neutral funding, positive funding, and negative funding. None of this is technically difficult. The data exists. The reason it is not published is not technical. It is commercial.
I have written this exact requirement into protocol review documents before. When I modeled dispute-game economics for optimistic rollups, I required the bond requirements and challenge windows to be visible in the order-flow simulation. The same standard should apply to perpetual futures. A product that cannot display its expected cost cannot claim to be designed for consumers. The cost is not a bug. It is a mandatory feature of the product architecture. But the absence of disclosure is a choice.
Takeaway: Perpetual futures are a trading instrument. They are not a holding vehicle. The Economist's 10 percent annual drain is a lower bound, not an upper bound. The true carry for a leveraged retail long is materially higher. The product will survive. The user base will migrate. The next phase will be regulatory cost disclosure, product redesign, and a shift toward venues where the cost lines are explicit. The question is not whether the warning is correct. It is what happens when the marginal long stops paying the carry. The market is about to find out.