A Framework for the Tokenized Equity Basis Trade
Kairos Research, for the Ethena Risk Committee. Market data as of August 26th 2026, from public venues and onchain sources
Summary and recommendation
This report proposes an approval framework for the tokenized equity basis trade. The trade is long a tokenized stock and short the equity perpetual on the same venue, collecting the funding that levered longs pay, with the spot leg drawn from Binance bStocks or from xStocks, the Backed Assets tokens that now trade on Kraken, Bybit, OKX and other venues. We measured the relevant equity perpetual markets across four venues, roughly $2.9bn of one sided open interest, verified every tokenized spot instrument onchain, and backtested the hedge through two earnings seasons, eleven weekends and the first live dividend events.
The design follows the commodity framework’s general format at the Committee’s request. One simple test decides whether a ticker is approved, the approved list is short and stable so the trading team always knows what is in scope, and position sizing within the list is the trading team’s responsibility, guided by the depth based limits we lay out below.
A ticker is approved on a venue when its equity perpetual holds $25mm or more of one sided open interest on a 14 day average, has at least 30 days of funding history, references a listed underlying, is not a leveraged or inverse product, and a matching tokenized spot instrument trades on the same venue.
Applied today the test admits 17 names on Binance and 3 on OKX, and none yet on Bybit or Kraken. Those counts are an output that moves with the market rather than the substance of what we are proposing, and the weekly republication is what keeps them current. We recommend the Committee approve the framework and the list it currently produces, approve bStocks as the spot instrument subject to the side letter now being negotiated, treat the remaining xStocks structural items as a short follow on review, and leave pilot sizing and scaling to the trading team within the guidance below, returning to the Committee at each material expansion.
The approval test
A ticker is approved on a venue when it meets all five of the following.
-
Perpetual open interest, one sided, 14 day average: $25mm or more
-
Perpetual funding history: 30 days or more
-
Underlying: a listed public security
-
Product type: no leveraged or inverse ETFs
-
Spot instrument: a matching tokenized stock on the same venue
The $25mm line serves the purpose the commodity framework’s minimums served, admitting only contracts that are no longer nascent. It is a deliberately simple threshold chosen against current market structure rather than a measured breakpoint, and it sits comfortably above the level where our measurements start showing unstable books and erratic funding. The 30 day funding requirement filters brand new contracts whose early funding prints are mostly noise. Leveraged and inverse ETFs are excluded because they split and reverse split constantly and the token issuer’s handling of those events is undocumented. The matching instrument rule means the perpetual and the token must reference the same listing, which matters in practice, as one currently qualifying contract shows below.
The averaging window needs stating precisely, because a name near the line can pass or fail on the choice. The test uses the fourteen complete UTC days ending on the day before publication. Combined with the two publication rules below, a name sitting on the threshold cannot enter on a single favourable window.
To keep the list stable and easy to track, the list can be republished on a fixed weekly date, and a name changes state only after two consecutive weekly publications agree: added when its 14 day average open interest sits above $25mm on two successive publication dates, removed when it sits below on two successive dates. Between publications the list does not change. Each weekly register from the first publication onward records the prior and current 14 day averages alongside the resulting state, so any addition or removal can be checked against the two observations that produced it. The list in this report is the initial seed, drawn from a single measurement, so it carries no prior observation to compare against. The threshold itself is reviewed quarterly against the size of the market, with changes going to the Committee.
What the test produces across the four venues
The list that follows is an output, not the proposal. It will change as open interest moves, and the weekly republication is what keeps it current. What matters for the Committee’s decision is that the test is simple enough to run without judgment and that it produces a workable book at each venue.
Binance qualifies 17 of its 67 matched pairs, the venue holding $2.14bn of one sided open interest. The names as of August 26th are SanDisk, SpaceX, SK Hynix, Micron, Alphabet, Circle, QQQ, the Roundhill Memory ETF, Intel, NVIDIA, Strategy, Tesla, Marvell, SPY, EWY, Meta and Apple.
OKX qualifies 3, currently SanDisk at $146mm of open interest, SpaceX at $62mm and Micron at $34mm, each with a matching xStock trading on OKX spot. OKX listed its xStocks as spot pairs in mid July, and 93 of them pair with a perpetual on the same venue that references the same listing, and its equity perp complex already holds $585mm across 155 contracts. This makes OKX a genuine second venue with the same clean single venue construction as Binance, not a cross venue patchwork.
One instructive near miss. SK Hynix’s Korean ordinary share perpetual on OKX carries $57mm of open interest and the richest funding in OKX’s equity perpetual complex, roughly 74% annualized over the trailing 90 days. It is not approved, because the xStock tracks the company’s Nasdaq listing while the perpetual tracks the Korean line, and the two listings have historically traded far apart. A hedge across them is not a hedge. This is exactly the mismatch the matching instrument rule exists to catch, and the name joins the list the moment a matching spot instrument exists.
Bybit qualifies none, but one name is close enough to watch. Bybit runs the broadest equity perp menu, on our classification 209 equity linked contracts of roughly 800 live perpetuals, against 213 under the venue’s own product tags. On the fourteen days ending 25 August, SpaceX averaged $24.29mm of one sided open interest, which is 97% of the threshold, and SanDisk averaged $20.61mm. Both are well clear of the eleven xStocks that trade on Bybit spot, so the construction is ready if either crosses.
SpaceX is the case the framework was built to handle. It fails on the stated window but would have passed on a window ending two days earlier, at $25.29mm, because its open interest has been falling from a peak above $36mm earlier in the month. A single reading would have admitted or rejected it almost at random. The two publication rule is what prevents that, and on the current trend it will not qualify at the next publication either.
Kraken qualifies none. The xStocks home venue’s own perpetuals remain tiny at $9.3mm across fourteen contracts. Nothing approaches the threshold.
What the data shows
Carry has compressed as the market has grown. On Binance the approved names averaged roughly 18% annualized in late July and average roughly 7% today, with two names currently printing negative. The OKX approved names run 8% to 16%.
The hedge held through every event we measured. Across 37 matched earnings events the underlying gapped 9.9% on average while the hedged position moved 20.3 bps on average, with the worst single outcome a loss of 81.7 bps. Across 400 weekend and holiday windows, divergence averaged 14.9 bps. In the four most recently measured weekends, where we have the seasoned split, every token past its first two weeks stayed under 53 bps. The sample includes real stress, SpaceX’s 13.6% earnings drop and 15.8% rebound, and Bitcoin’s 21% three day rally in late August.
Young tokens are the risky ones. Basis dispersion runs 41 to 48 bps in a token’s first two days and settles at 13.2 bps after two weeks, and the worst dislocations we measured all came from tokens in their first days of trading. Note that the 30 day requirement seasons the perpetual, not the token, so a newly listed token paired with an established perpetual can clear the test while its own book is days old. Apple is the live example, approved with a token 27 days old against a perpetual with 143 days of funding. The sizing guidance closes that gap by giving a newly listed token no funded size in its first 14 days.
How the perpetuals price when the stock market is closed is the deepest structural risk. A perpetual is margined and liquidated against the venue’s index price, and the stock market is closed for roughly 70% of the hours in a week. During those hours Binance builds its index from the perpetual’s own order book and Kraken’s follows the token price, so the price that decides whether the short gets liquidated comes from the same thin crypto markets the trade sits in. A hard push in a thin book while the stock market is closed could liquidate the short at a real loss while the matching gain on the token cannot be sold at size. We reverse engineered Binance’s and Kraken’s index and funding mechanics in full; OKX and Bybit publish similar index frameworks but we have not yet measured their closed market behavior, and that study belongs in the trading team’s onboarding work for OKX. The controls are continuous comparison of each venue index against the last real stock price, extra weekend margin, or simply closing positions before the Friday close.
The spot instruments
bStocks, approved subject to the side letter. The structural documentation is public and we have read it, the ADGM approved prospectuses, the weekly certificates outstanding filings, the proof of collateral portal, and the token contracts, whose shared implementation is source verified onchain. Ethena has confirmed that holders have no proprietary interest in the backing shares and that the issuer retains broad flexibility over the underlying and over corporate actions, and is negotiating a side letter covering exactly the points our diligence surfaced. Our recommendation is that bStocks approval be conditional on that side letter covering a confirmed custodian with lending restrictions, an inventory reconciliation by someone outside the Binance group, the audit reports behind the issuer’s published formal verification claim together with a key management plan for the shared upgrade key, guaranteed creation and redemption access at committed size, definitive dividend treatment in writing, and a binding corporate action methodology.
On dividends the record needs the side letter to settle it. Ethena’s understanding was that no dividends are paid out, while the issuer has since filed four corporate action disclosures with the ADGM listing authority showing certificates outstanding increased by the dividend net of 30% US withholding, with the arithmetic reconciling exactly. Until the treatment is stated in writing, the trading team should assume zero dividends are shared.
xStocks, structurally sound, approval per venue as open interest crosses the line. The product has the stronger disclosure stack, a public prospectus, named custodians barred from lending the shares, live onchain proof of reserve and direct redemption at net asset value, and it has grown into a genuinely multi venue instrument, roughly 714 products with about $799mm of float as of August 26th, spanning ten chains plus a venue native deployment, trading on OKX, Bybit, Kraken and Gate. We found no product level reason to exclude xStocks beyond the five part test and the venue onboarding work each venue still needs. The remaining items are real but small, and we recommend they run as a short follow on review rather than holding up approval: reconciling supply across ten chains against the proof of reserve, the Kraken group’s stacked roles as issuer owner, venue, market maker and custodian, the mechanics of the dividend reinvestment rebase during its adjustment window, and the settlement model behind OKX’s shared order book listing. One live example shows why the corporate action item matters, after Netflix’s ten for one split the token traded on one venue at ten times its price on the others for a period. Venue conventions around corporate actions must be verified per venue before size.
Position sizing and risk guidance for the trading team
While approval designates a ticker as eligible, determining position size rests with the trading team. We advise monitoring allocations against the thresholds below, with order book depth serving as the governing constraint in practice.
-
No more than 10% of the perpetual’s open interest, less for volatile names.
-
No more than 20% of the token’s circulating supply
-
No more than half the visible order book depth within 1% of mid, taken on the weaker side of the weaker leg and measured at the 25th percentile of the trailing series rather than at a favorable moment. On Binance the books quote roughly $0.56mm per side on the deepest name and far less on most, so depth based caps land well below what open interest alone would allow, and that is the correct outcome.
-
No more than 10% of daily traded volume on either leg.
Around key events such as weekends or earnings days, we would suggest to either carry extra stablecoin margin of at least 10% of short notional over weekends and halve exposure through earnings, or simply close positions before the US market close going into a weekend and before the close on earnings day. Entry on any name requires expected net carry above the trading team’s hurdle rate after fees, financing, dividend loss and slippage, spreads inside 15 bps on both legs, and prices in line with the stock.
Three conditions warrant an automatic halt and same day escalation to the Committee rather than desk discretion. First, any change to the implementation, upgrade authority, administrative roles or pause roles behind a held token, covering bStock contracts, xStock contracts and their bridges, checked daily against chain state. Second, a backing or reconciliation failure at either issuer, meaning the issuer reported backing below 100%, a failed reconciliation, or the backing record for a held name going stale beyond 72 hours. The 72 hours is set against observed behaviour rather than an ideal: the Binance proof of collateral portal refreshes daily in aggregate but individual ticker rows have been seen up to five days old, which does not meet a tighter standard today. Bringing per name reporting inside a defined and monitored window is one of the side letter asks, and until it is agreed the trading team should treat a stale row on a held name as a stop on adding to that name. This control is separate from the monthly independent reconciliation, which addresses whether the shares exist rather than how fresh the reporting is. Third, an index divergence, evaluated per name in two forms because the exposure differs by session. During a live session, the venue index diverging by more than 1% from the live underlying for 15 minutes. Outside of US market hours, which is the majority of the week and the period we identify as carrying the deepest structural risk, the venue index and the tokenized spot diverging from each other by more than 1.5% for 15 minutes, or either diverging by more than 2% from the last regular hours print adjusted for any approved after hours reference. The closed market thresholds are wider because no independent price exists in that window, which is precisely the condition that makes them necessary. Escalation is the same day and positions in the affected name stop growing immediately. Everything else, carry fading, depth thinning, funding parameter changes, is the trading team’s territory with our monitoring to compliment it.
Open risks
Issuer credit on bStocks Holders have no proprietary interest in the backing shares and the issuer can deal broadly with them. Until the side letter restricts that, the spot leg is unsecured credit exposure to a Binance affiliate rather than a claim on the stock. This is the strongest argument for capping bStock exposure as a single concentrated position.
The shared upgrade key Every bStock token sits behind one upgrade key that remains a single externally owned account with no multisig and no timelock. Two attempts to transfer its ownership failed on 24 July and were never retried. One compromised key would touch all 67 tokens at once. The issuer’s website claims formal verification by tier one security firms with no report published anywhere.
Index pricing while the stock market is closed Described above. Monitoring detects it, position size limits the damage, and the OKX and Bybit closed market index study is open work.
Concentration in one corporate group On Binance the issuer, spot market, perpetual market and margin account all sit inside one group. The xStocks venues genuinely diversify this, though the Kraken group itself holds stacked roles around xStocks, and the Kraken group also custodies some assets for Ethena.
Carry can keep fading Approved name carry halved in a month as the market grew, and two names print negative today. Nothing in this framework assumes carry persists. The hurdle rate governs entry, and positions unwind when net carry stays negative.
USDT dependence The Binance index divides by the USDT price, and the Binance, OKX and Bybit contracts in scope are USDT margined, a stablecoin exposure unrelated to equities. Kraken supports several collateral currencies. We recommend posting only approved stablecoins as margin across every venue, which remains subject to Ethena confirming its collateral setup for this strategy.
Limited History of Data The oldest matched pair has 75 days of history. The evidence is strong for its length, 37 earnings events and several weekend windows, but it covers one market regime, which is why sizing discipline and staged growth carry the weight that confidence cannot yet.