TradFi vs DeFi Leverage: 2x by Law, 200x by Code

Three ways to create leverage

Innovations in the crypto space have introduced new ways to create leverage outside of the traditional broker-borrowing relationship. While all three structures achieve the same leveraged exposure to an asset, they each take a structurally different path which ultimately dictates its competitiveness.

Margin and the collateral loop share the same shape whereby you borrow against posted capital and only pay interest on the borrowed amount. A perpetual never actually touches the underlying asset, hence fees are charged against the total value of the position. That's the shape each mechanism takes. What it actually costs, across real venues, is next.

Same $10,000, wildly different costs

Every ceiling answers to someone different

How the leverage is created determines the maximum leverage available. Traditional brokerages are required to follow a margin ceiling set by regulators. Perpetual exchanges still have the flexibility to set their own ceiling according to the venue's own risk appetite. DeFi collateral loops are limited by collateralization ratios set in code via protocol governance.

The American stock market provides the deepest liquidity, but brokerages must abide by a regulatory ceiling set in 1974. Regulation T sets initial margin at 50% of a position's value, which caps leverage at 2x. You can borrow up to 50% of your position's value from the brokerage. This 52-year-old rule was a reaction to the 1929 crash, when retail investors could borrow up to 90% of a position's value, 10x leverage.

Except for China, which recently eliminated leverage on new trades, other regulators looking to grow their own domestic markets have taken a less conservative approach. Japan sets a flat 30% margin, allowing up to 3.3x leverage. South Korea scales leverage to each stock's own volatility instead, with an official tiered table running from 20% up to 100%, allowing up to 5x on the loosest names. Hong Kong leaves margin lending to each broker's own risk judgment rather than a published table, with blue-chip max leverage usually capped at 3.3x.

A perpetual's ceiling is always the venue's own choice, no matter how much oversight sits behind it. Kraken's BTC perpetual runs through Bitnomial, a CFTC-regulated exchange, yet Bitnomial still sets and self-certifies its own 100x cap rather than the CFTC dictating it. Kraken's xStocks perpetual runs through Payward Digital Solutions, a separate, Bermuda-licensed Kraken entity, and caps at 20x on the same self-determined basis. Hyperliquid's S&P 500 market caps at 50x, Ostium's at 200x, with no regulator in the picture at all. Four markets, four numbers, none of them handed down by law.

DeFi collateral loops remain unregulated, with leverage caps enforced by protocol governance. Aave governance token holders vote directly on each risk parameter, but turnout for major proposals consistently sits below 10%, with the top three voters controlling over 58% of the vote. Morpho governance instead pre-approves a fixed list of collateral ratios that anyone can use when permissionlessly deploying a new market. Market isolation enables Morpho to provide 7.1x leverage against Aave's 3.7x. Morpho markets allow only one collateral per debt asset, which significantly reduces cross-market risk compared to Aave's multi-collateral, multi-debt pools. What Aave's single shared market enables instead is deeper liquidity, which provides greater safety against sudden price spikes and more efficient pricing.

What it actually costs

The structure also dictates the cost of leverage, and with it, each venue's competitiveness. In addition to the costs of maintaining a position via borrow/funding rates, some venues charge a one-time fee for opening and closing a position. Some charge fees on the whole position value while others only charge for the amount borrowed. Given all these combinations, the cost for holding the same leveraged price exposure for a month can vary by nearly 7x.

Brokerage margin sits in the middle and is the simplest structure whereby interest only accrues on the amount borrowed from the brokerage. Brokerages set their own rates, benchmarked to a base rate (~5–9.5% annualized in this comparison) plus their own markup. This cost is billed monthly on your brokerage account and never at the level of an individual trade. While this account debit arrangement provides more flexibility day-to-day, it can also bury the cost per position on the trading interface.

Perpetual venues bill you three times against the total position size: a fee on opening, a fee on closing, and an ongoing funding rate. The one-time fee is paid to the venue directly; funding is paid peer-to-peer between longs and shorts instead. Exchanges set their own open/close fees (~0.06–0.5% round-trip in this comparison) for facilitating the trade, while funding is market-driven, based on how far the perp's price deviates from the spot market (~23–34% annualized here), so as to incentivize opposing trades.

DeFi collateral loops are the cheapest, with interest charged only on the borrowed amount. Borrow interest is determined via a market-specific borrowing curve that scales the borrowing costs according to a target utilization rate. Competition across these venues usually results in the rates clustering and moving in tandem (~4–5% annualized). The underlying collateral can also earn a lending interest if the market allows and there is demand for it (0% for WBTC), which then offsets your borrowing costs. These rates work both ways, since each loop increases both your total collateral and your total debt.

Given the differences in rates, the holding period for the leveraged position becomes a major factor if cost-optimizing between the different venues. DeFi collateral loops are generally the cheapest long term option while perpetual fees generally incentivize shorter term trading. The rate itself also moves at different speeds, ranging from real time in DeFi collateral loops, hourly to every 8 hours for perps, to only at the brokerage's discretion. This adds an additional consideration as effective rates can spike when updated more frequently which also increases liquidation risks.

That $0 line for Ostium reflects a real design choice, where all your assets are forfeited once a position crosses its liquidation threshold. Forced liquidations often cost more than just losses from leveraged price exposure (e.g. a 10% price drop is >20% loss for a 2x position). Brokerages are the gentlest, with no additional liquidation penalty on top of that price loss, only the interest paid on the loan while the position was open. DeFi collateral loops charge an additional ~5% liquidation penalty on the total borrowed amount, which scales actual dollar losses with leverage. Perpetual venues vary, with Hyperliquid only charging a penalty if market depth was too shallow to close the position cleanly, and Kraken not publicly documenting an equivalent fee for this product. Ostium takes your whole position once it goes under. A low-leverage position's safety comes from surviving longer, not from losing less once it finally breaks.

Who actually pays

Beyond the immediate trading costs, there are also structural risks unique to each market. From exchange-specific failures to dependence on external infrastructure, positions can become inaccessible, or be wiped out entirely. What matters more than the millions lost is who actually ends up paying for these failures, and whether users even have a reliable path to recourse.

Brokerages' reliability derives from established regulations around consumer harm, but getting a payout can still take years. Robinhood was ordered by FINRA to pay back $12.6M to customers affected by its own outages and misleading account information. Part of that figure was based on estimated harm from users being unable to trade. The rest covered losses from inaccurate account and margin data. Robinhood also paid a separate $57M penalty to FINRA as part of the same broader settlement. IBKR's risk engine failed to handle negative prices when nobody wanted to take physical delivery of oil futures contracts during the COVID-induced oil crash. IBKR covered $104M from its own funds to compensate customers before any regulator acted. A CFTC order over a year later made $82.57M of that restitution formal, credited entirely against what IBKR had already paid, and added a separate $1.75M penalty.

DeFi-based perpetual venues sit outside regulation, with recourse depending on the venue's own decisions. Ostium raised $27.8M across investors including Jump Crypto and General Catalyst, with over $50B in cumulative volume and ~1M trades. On 15 July 2026, an attacker exploited Ostium's own pricing mechanism to drain $18–24M from the shared Ostium liquidity pool that functions as the counterparty to every trade. Liquidity providers absorbed the exploit in full, with no compensation plan announced. Hyperliquid's backstop design was exploited too. An attacker manipulated an illiquid JELLY market to force Hyperliquid's own liquidation engine to hand off their liquidated short position to the HLP pool, then kept pumping the price further, inflating HLP's paper loss to ~$13.5M. Validators voted off-chain to delist the contract within minutes and force-settled it at the attacker's own entry price, closing HLP's position net positive. Hyperliquid's Foundation separately reimbursed nearly every other affected user with its own funds. A slower vote, or a less generous Foundation, could have resulted in real losses instead.

DeFi collateral loops require stacking multiple protocols on top of each other, each with their own unique risks. An attacker exploited a weak verification setting in Kelp DAO's cross-chain bridge to forge a message and mint rsETH with no backing. This rsETH was then used as collateral on Aave and Morpho to borrow tokens with real value. Aave's pooled risk model meant non-rsETH tokens got drained from the same shared pool. A rescue raise driven by Aave raised about $160M to cover losses that can range from $124–230M, depending on whether Kelp DAO decides to socialize the losses across all networks or isolate them to just Layer 2 networks. Any gap that doesn't close falls to Aave's Umbrella backstop stakers, who signed up for exactly this risk in exchange for yield. Morpho took an isolated market approach instead, where one collateral is paired with one debt, so only the rsETH markets were affected, limiting the damage to just 2 of roughly 500 markets, a total loss of $1M.

When leverage is more than a number

Demand for leverage is only growing across all three structures. Brokerage margin debt just hit a record, perpetuals account for the majority of crypto's trading volume, DeFi money markets are growing as looping becomes more widely adopted. Behind this growth sits the uncomfortable fact that the majority of people trading on leverage will lose money doing it, and this is only when looking at the PnL.

Leverage was never just a promise to multiply your gains. On the surface, leverage is a multiplier that amplifies both gains and losses, but costs vary by structure and risk isn't linear either. Instead, how leverage is structured sets a ceiling for how far leverage can go and what it actually costs. Who sits behind the number when something goes wrong determines if you end up bearing the costs. Understanding the structure matters because a number that just reads as a profit multiplier encourages more trades, and exchanges earn on that volume either way.