Crypto cards with a trading desk built into the app

One pot or two

Several of these apps advertise trading and spending from one account, and on all 12 the issuer’s own documents say otherwise. The tell is in the docs: a minimum deposit to open the trading account. MetaMask asks $10 into a Hyperliquid account. Coin98 wants 5 USDC in a dedicated futures account. Tria asks $10. Based doesn’t publish a figure but says the two wallets cannot be converted at all. Ledger, xPortal and Jupiter show it from the card side: the card spends a balance held by a card partner, and you move money into it before you can spend. BloFin Wallet and Bitget Wallet work the same way, and Bitget’s help centre adds: “Your Perps account is separate from your wallet account.” Revolut keeps CFDs in its separate brokerage, Revolut Securities Europe UAB, while the card spends the main account. Bitpanda moves the collateral into a margin wallet whose assets “cannot be withdrawn, swapped, or used elsewhere”. Money that has to be moved was never in one place.

Which pot the money sits in decides what happens on your worst day. A liquidation costs you what you deliberately moved into margin, and the card keeps working on the rest. That is the fence, and it is only as good as your own restraint about topping the margin back up: Hyperbeat puts a Sweep to Card button on one side and a one-tap conversion of the spending balance into collateral on the other. Look at the funding screen before you fund anything.

The card that read like one pot until it didn’t

Hyperbeat ran in this slot from August until 28 September 2026 as the one card whose own terms put spending and trading on the same wallet. Its terms still do, word for word. What changed is that Hyperbeat published a funding guide, and the guide describes the transfer the terms deny: “Before you can trade, you need funds in your trading account.” The spending account is listed as one of six deposit routes, and the money comes back by Transfer or by a Sweep to Card button. The one-balance language that survives covers spot and perps, which is the trading side talking to itself. So the fence exists here too. It is just a low one, and the app hands you the gate.

Hyperbeat sells this as “the same account you save and spend from”, and the order book it calls the world’s fastest is Hyperliquid’s, not its own: the same one 9 of these 12 cards route to, so the superlative is borrowed. Card credit is collateralised into lending protocols while perp margin sits on the venue, and no published sentence says a liquidation on one side reaches the other. Take the method from the episode: marketing that sounds like one balance is not evidence of one balance, and neither is a terms page. The funding screen settles it.

How all 12 keep them apart

The ring-fenced shape is “one app, two accounts”, and ring-fenced is not safe: the margin you moved can still go to zero, and on an on-chain desk a wick can liquidate you with nobody to appeal to. What the fence buys is that the loss stops at the money you moved. On 12 cards the issuer states the split: Hyperbeat Card, Based Card, Tria Card, MetaMask Card, Ledger Card, xPortal Card, Coin98 Fusion Card, Jupiter Card, Revolut Card, BloFin Wallet Card, Bitget Card, Bitpanda Card. Separate pots also do not mean self-custodied pots. Coin98, Ledger, xPortal, Jupiter, BloFin Wallet and Bitget Wallet pair a self-custody wallet with a card balance held by the card’s issuer or programme partner, and Ledger’s own FAQ says funds moved to the card account “are no longer under self-custody”.

Ledger publishes no leverage ceiling anywhere we could find, which is unusual in a group where everyone else leads with a multiple. Its card is run by Baanx, and the card balance is custodial while the wallet stays self-custodial. Revolut does not run a crypto desk. It trades CFDs on stocks, indices and commodities through a brokerage authorised by the Bank of Lithuania, at 1:5 on single stocks. It is also the only entry where a licence condition stops the account going negative, which is stronger than a promise the firm wrote itself. Bitpanda runs its own book too, but as spot margin rather than derivatives: up to 10x on crypto, long only, and up to 20x on stocks, ETFs and ETCs. Its terms stop crypto losses at the collateral; on stocks its help centre says you may still owe money if the closing proceeds fall short.

Who is behind the order book

9 of these 12 cards are interfaces onto Hyperliquid, so two of them chosen at random are more likely than not to be the same market twice. That is the finding from our research on trading-desk concentration, and it is the same shape as the BIN-sponsor concentration we mapped in May: many consumer brands over one entity, and a failure mode nobody prices until it fires. This one has fired. A validator vote force-settled every open position in one market in March 2025, and the first cross-margin auto-deleveraging closed profitable positions that October.

If you hold two of these cards because you like having a backup, the trading half of that backup is not independent. As payment cards they are two issuers and two networks, but the positions sit in one place. An outage or a forced settlement at the venue reaches both at the same moment, and neither app can do anything about it.

Where the desk is blocked

A card’s country list and its desk’s country list are different documents, and comparison pages routinely quote the first as if it covered both. Four cards in this group show the gap:

• Ledger’s desk names five restricted jurisdictions (the UK, the US, Ontario, France and Belgium), and the card is sold in all five. Two of those are EEA states, so a card row that says “EEA” hides half the problem. The card also sells in Switzerland, Brazil and the rest of the EEA, where the desk works, so the split is partial.
• Revolut’s account reaches eight markets beyond the EEA, from the US to Singapore, and the CFD desk reaches none of them: a Swiss or Australian cardholder gets the card and no CFD account. For the UK, Revolut never says CFDs are unavailable. The UK help pages 404 and the UK product list does not include them, and silence proves nothing either way.
• MetaMask is the plainest of the four, on the help page for the feature itself: perps are unavailable in the USA, the UK, Ontario and Belgium, plus the US sanctions list.
• xPortal names the US, the UK and Canada, “including but not limited to”, and only in the terms. The help-centre article that explains the feature carries no geography at all, and the same clause says xPortal “does not monitor or validate your jurisdictional eligibility”.

Six more publish nothing about the desk that I could find: Hyperbeat, Based, Coin98, Jupiter, BloFin Wallet and Bitget Wallet. Jupiter’s app-wide terms, as I read them in August, exclude US and Singapore wallets from everything, the desk included. That matters more now: Jupiter’s card is offered in 31 US states and in Singapore, so a cardholder there may find the desk closed. Bitget Wallet’s only country list is also app-wide (sanctioned regions plus Singapore), and BloFin Wallet sells its card in 48 listed countries while its app shows a generic restricted-region message with no list. Bitpanda publishes stock margin as open in the EEA and Switzerland and closed to UK users, with no list for crypto margin; its card is sold in the eurozone, inside the stock-margin area. Tria bars sanctioned jurisdictions without naming one, so a reader cannot check their own. Revolut publishes no country list for the desk either; what it publishes is the licence the desk runs under, which is what makes it an EEA product. The per-card wording, and what was checked to establish each “not published”, is in the research.

Cards people expect to find here

Three groups get proposed for this list and none of them belongs. A fourth set sits closer to the line, and three of those made it.

Exchange cards. Bybit, OKX, Binance, Kraken, Gate, KuCard, MEXC and the rest are issued by exchanges that already sell derivatives on the same account. Of course the desk is there. The card is an accessory bolted to a trading account, which is the opposite of what this page tracks. I do not publish a count for this group: nothing in our database records issuer type, so any figure would be a brand-name guess. That also means the 12 on this page are 12 outside the exchange cards, and the whole catalogue holds more. The exclusion is a judgement, not a property of the data, and it follows the app rather than the brand: a card issued from the exchange account stays out, while a wallet app from the same group that sends its perps elsewhere can come in.

Borrowing against crypto. Six cards let you draw stablecoins against a crypto position to fund spending: Ether.fi Cash, Exa, Xplace, RedotPay, Busha and Wirex. The word “leverage” appears near all of them. A loan opens no position and changes no exposure. It is a useful feature and a different product.

Tokenised shares. Nine cards put equities, indices, ETFs or metals inside the app. None offers leverage on them: Fasset states outright that margin is unavailable, and Uphold publishes leverage fixed at 1:1. A tokenised share is owned outright, with no derivative involved. One of the nine, Fizen, has since paused card creation, so it is off the table either way.

The close calls. Bitget issues its card from a self-custody wallet app that carries a perp desk, and was cut until October because the parent is a centralised exchange. It is in now: the wallet sends its perps to Hyperliquid and PopDEX, not to the Bitget exchange. BloFin Wallet came in on the same ground, a Rain-issued Visa inside BloFin’s self-custody wallet whose perps go to Hyperliquid. The older BloFin card, the exchange’s own, stays out. Phantom was on this list until I compared the two country lists and found they share nothing, which is set out in full in the rejection table. Jupiter stays, though not for the reason you would expect from its store listing: that listing names the card and the cashback and never once says perp, leverage or futures, and the desk turns up in the documentation instead. A store listing lags the app. The full rejection list is in the research.