A tokenized share of Nvidia trades on-chain on Saturday. Nvidia itself does not.
That gap is the whole problem with using tokenized stocks as margin. The token moves around the clock. The market that defines its value is open for about 32.5 hours of a 168-hour week, as one oracle provider put it. A venue that accepts the token as collateral has to decide what it's worth during the remaining 135.5 hours of a typical week.
This is the collateral layer of perpification, and in 2026 it has stopped being theoretical. Tokenized stocks reached about $3.14B in distributed value across roughly 4 million holders as of September 27, according to rwa.xyz, with Ondo, bStocks, xStocks, and Securitize the largest issuers. Several venues now accept them as margin. The way they do it says a lot about what a collateral haircut is actually for.
Why traders want stock collateral at all
The argument is the same one we made for crypto assets in AnyCollateral: Beyond Multi-Collateral Trading: if you have to sell an asset to post margin, you've exited the position you may have wanted to hedge or lever. Tokenized stock collateral lets a holder keep the exposure and still trade.
In traditional finance this is ordinary. A brokerage margin account lets a customer borrow against stocks they already own. What's new is doing it on-chain, around the clock, against an asset whose primary market keeps banker's hours.
The TradFi baseline
Before looking at on-chain haircuts, it helps to know what traditional markets charge.
- Regulation T lets US brokers lend up to 50% of the purchase price of a stock. FINRA Rule 4210 requires maintenance equity of at least 25% of the current market value of margin securities.
- Portfolio margin replaces fixed percentages with stress tests. Under FINRA's rules, a single stock is stressed at ±15%, and a large-cap broad index at +6%/−8%.
- Securities-backed credit lines typically lend 50% to 95% of account value, according to an SEC and FINRA investor alert.
- Clearinghouses go further. Since 2010 the Options Clearing Corporation has run stock collateral through the same simulation as the positions it backs, producing a portfolio-specific haircut. It models the collateral and the exposure together, not separately.
Keep that last point in mind. It comes back later.
What on-chain venues charge
Here's how three venues that publish their numbers treat tokenized stock collateral:
| Venue | Asset examples | Haircut or factor | Other limits |
|---|---|---|---|
| Ondo Perps | Gold 6.5%, S&P 500 / Nasdaq-100 ETFs 10%, Tesla 17%, Nvidia 25% | Haircut off market value | Per-asset caps |
| Kraken (xStocks) | S&P 500 / Nasdaq-100 ETFs 10%, large single stocks 20%, higher-volatility names 30% | Haircut off market value | Per-asset caps; balances above the cap don't count |
| Aave V4 on Base | Seven large-cap stock tokens | Collateral factors of roughly 65–79% | Market-wide collateral and borrow caps |
Two things stand out.
The ranking tracks volatility. Broad index ETFs get the smallest haircuts. Single stocks get more. The most volatile names get the most. That's the same logic as portfolio margin's ±15% single-stock stress versus the narrower index ranges.
A 10% haircut on an index ETF looks far smaller than Reg T's 50%, but the two aren't directly comparable. A collateral haircut discounts what the asset counts for; Reg T sets how much a broker can lend. On-chain venues also rely on other controls alongside the haircut: per-asset caps, position margin requirements, their off-hours pricing policy, and liquidation mechanics. The haircut is one control among several, not the whole defense.
Ondo's documentation is explicit about what the haircut is for. It reflects "price volatility, transaction costs, and weekend gap risk." Traditional margin frameworks deal with overnight and weekend gaps too. What tokenized collateral adds is that the collateral, and the positions it backs, can keep trading while the underlying equity market is closed.
Three ways to price stock collateral on a Saturday
When the primary market is closed, a venue has to pick a pricing policy for stock collateral. Three distinct approaches are live:
1. Freeze the last price. On Aave's Base deployment, the stock price feed publishes on a weekday schedule, running about 120 of 168 hours a week. Risk reviewers noted that from Friday evening to Sunday evening the feed "publishes nothing," and sized the collateral factors to treat each closure "as a window in which the position cannot be repriced." The haircut provides a buffer against moves between Friday's close and the next open, though a large enough move can exceed it.
2. Accept prices only within a band. Kamino, which integrated xStocks as lending collateral in 2025, accepts off-hours oracle prices only within a set deviation from the last market close. Moves inside the band are recognized; outliers are not.
3. Keep marking, with an internal price. Ondo Perps marks collateral continuously, with "no price freeze." Over the weekend, an internal order-book oracle takes over from Friday afternoon until Sunday evening. If collateral has to be converted to cover losses while the market is closed, a 2.5% closed-market settlement fee applies.
None of these is free. Freezing the price means the collateral value can be stale by the time Monday opens. Banding means large genuine moves are recognized late. Continuous internal marking depends on off-hours price discovery, and converting collateral while the market is closed carries an explicit cost.
How far do tokens actually drift? One data point: CryptoSlate tracked four stock tokens that stayed within about 0.6% of their Friday closing prices over a weekend in August. That's one quiet weekend, not a stress test. The haircut exists for the weekend that isn't quiet.
Who buys the collateral on Saturday?
Pricing is half the problem. The other half is liquidation.
Liquidation always depends on how much secondary-market depth is available. Tokenized stocks add three complications:
- Redemption is limited. xStocks, for example, can be minted and redeemed only on US business days, with KYC and a minimum size. Some stock tokens give secondary holders no direct redemption right at all; the Aave risk review noted that a liquidator of those tokens would have to sell on the secondary market or hedge rather than redeem.
- The primary market may be closed. A liquidator selling on Saturday is limited to on-chain liquidity, with no open primary market to arbitrage against.
- The token is not the share. Tokenized stocks are typically tracker instruments backed by the underlying, without shareholder rights. Corporate actions like dividends and splits are handled by issuer rules, and venues need their own rules for adjusting balances and reference prices. Ondo, for instance, holds the collateral reference price fixed during a split.
That's why the caps in the table above matter as much as the haircuts. Depending on how it's scoped, a cap limits how much of one asset counts as collateral or how much aggregate exposure it can back. Kraken, for example, excludes balances above the cap from both collateral value and liquidation.
The risk that counts twice
The subtler problem is correlation.
If you post a tokenized stock as margin for a different position, you carry two distinct risk factors: the position's PnL and the collateral's price. We walked through that basic case for a crypto token in the CHOG AnyCollateral announcement: a drop in collateral value can trigger liquidation if it pushes the position past its margin requirement, even when the trade itself hasn't lost money.
Now suppose the collateral and the position move together. A trader posts tokenized Nvidia as margin and goes long a Nasdaq-100 perp. If tech sells off, the position loses and the margin shrinks, at the same moment. In banking, the closely related concept is wrong-way risk, which the Basel framework defines as exposure that rises as the collateral or counterparty weakens. The point carries over: assessing collateral and position separately can miss their combined effect.
This is where the clearinghouse approach earns its complexity. Simulating collateral and positions together can capture that dependence, depending on the model and scenarios. Fixed per-asset haircuts don't capture it on their own, so they need other controls alongside them, and traders should account for it too. The practical rule is simple: stock collateral backing a position in the same direction as the stock is closer to extra leverage than to a hedge.
What a stock-collateral design has to answer
Put the pieces together and any venue accepting tokenized stocks as margin has to answer five questions:
- What is the collateral worth when its market is closed? Freeze, band, or keep marking, and on whose price?
- How big is the haircut, and what else backs it up? Caps, conversion rules, and liquidation mechanics all carry part of the load.
- Who can liquidate it, and into what market? Redemption windows and on-chain depth set the real exit.
- How are corporate actions handled? Dividends and splits need explicit rules for adjusting balances and reference prices.
- Is correlation between collateral and position accounted for? And if so, by the venue's model, its limits, or the trader?
Several of these questions, especially pricing, liquidity, liquidation, and correlation, apply to any non-stablecoin collateral. LeverUp's AnyCollateral currently supports approved Monad ecosystem tokens beyond USDC and MON; it does not support tokenized stocks today. Candidate assets are reviewed for liquidity, oracle availability, and risk profile, each accepted asset carries a collateral ratio set by the protocol, and the supported list is shown in the trading interface. For the underlying argument about why a flexible collateral layer matters as more asset classes move on-chain, see RWA Meets Perpetuals.
Tokenized stocks will keep moving on-chain. Supported bStocks can already be bridged to Monad, for example, though being bridgeable doesn't make an asset eligible collateral. Whether tokenized stocks work well as margin depends less on the token than on the rules around it: token rights, off-hours pricing, liquidity, caps, and how liquidation works while the stock market is closed.
Further reading: