Self-Custody Bitcoin Cards

Your keys, your bitcoin - the card spends from a wallet you control.

cards on this page: 2 · issuing today (excl. watch list): 1 · verified or reported facts: 50% · direct on-chain BTC top-up: 2

quick picks

card spend from a vault you actually control ether.fi Cash - on-chain funding into a self-custody architecture, 3% cashback on the first $2,000/month, ~1% FX - the only fully documented self-custody card in the live set
euro banking with a card settling against your own wallet Nuri - passkey self-custody wallet with IBAN and Visa card via Wirex BaaS - verify-statused, no public fee schedule, but the keys-and-IBAN combination is unique in the population

Short answer: Two live products genuinely involve your keys: ether.fi Cash (on-chain funding into a self-custody vault, 3% cashback, ~1% FX) and the relaunched Nuri (card debits settle against a passkey wallet). Self-custody moves trust from an operator ledger to a vault contract - it does not remove it, because the merchant still settles in fiat.

#	Card	Bitcoin in	BTC sold	KYC	Monthly	Cashback	Score
1	Nuri Direct BTC top-up (on-chain)Self-custody0% FX	Direct BTC top-up (on-chain)	Not documented	Basic KYC	none	none	3.2	Inspect
2	ether.fi Cash Direct BTC top-up (on-chain)Physical cardSelf-custody	Direct BTC top-up (on-chain)	Not documented	Full KYC	none	3%	6.4	Inspect

"Self-custody card" is usually marketing. The honest version is narrow: the card's balance or collateral sits in an architecture where your keys authorize movement - ether.fi's vault model [2], Nuri's passkey wallet [4], and Blink's documented non-custodial balance mode [6] on the Lightning side. Everything else on the market holds your coin on its ledger and calls the app a wallet. This page covers the two live products that pass that test, what the keys actually control, and the settlement model that no card can escape: the merchant is always paid in fiat by someone who is not you.

2 live products carry a documented self-custody or hybrid architecture. 2 cards match this page's filter; the difference is definitional - one filters the lane, the other filters the marketing.

Three settlement models: custodial IOU, hybrid wallet-following balance, and the self-custody vault.

The two models, side by side

Vault-collateral (ether.fi): funding is on-chain into a self-custody architecture; the card spends through direct-pay (crypto sold per spend) or Borrow mode (a credit line against the vault - the coins themselves are never sold) [2]. The economics are clean: 3% cashback on the first $2,000/month at the free Core tier, 4% at VIP on the first $50,000 [1]; ~1% FX base plus a small tier margin [3]; ATM access at $1,000/transaction, five per rolling day [3]. At the EUR ladder the same tiers anchor at $800/$3,000/$16,000 of monthly spend [1]. The keys never leave your side - the operator's job is settlement, not custody.

Wallet-settled (Nuri): a self-custodial wallet for EUR/USD/stables/BTC with IBAN rails, and a Visa card whose debits settle against the wallet - issued and serviced on Wirex BaaS [4][5]. The wallet layer is genuinely yours ("Your Keys, Your Coins" [4]); the card layer is a Wirex product with Wirex limits (£30,000 per transaction/month, £500/day ATM [5]). Verify status because no public fee schedule exists and a network-label conflict remains unresolved [5] - a transparency gap, not a red flag.

Card	BTC in	Custody	FX	ATM	Monthly	KYC	Status
Nuri	Direct BTC top-up (on-chain)	self-custody	-	-	$0/mo	Basic KYC	verify
ether.fi Cash	Direct BTC top-up (on-chain)	self-custody	1%	$NaN	$0/mo	Full KYC	active

Live values from our database - details and caveats in each review. "-" = not published by the operator.

What the keys do and do not control

The keys control the collateral: nothing moves without them, which is the solvency insulation custodial balances lack. They do not control the settlement leg - the issuer's banking relationships decide whether the fiat reaches the merchant, and a terminal decline is possible in every model. And they do not change the tax physics: the conversion to fiat at settlement is a disposal in the US [7] regardless of where the keys live. Self-custody here means you hold the asset, an operator holds the rail - the honest boundary line for the category.

ether.fi Cash, live capture, accessed 2026-10-05 - the vault-collateral architecture as documented: LTV table, modes and supported networks public.

The honest boundary of the category

It is worth being precise about what self-custody cannot deliver on a card, because the marketing keeps blurring it. A card payment has three legs: authorization (the network), conversion (crypto to fiat) and settlement (fiat to the merchant). Self-custody can own none of those legs - it owns what stands behind them. The vault means the operator cannot lend, stake, freeze or lose your coins, and that is the entire value: an operator failure costs you the rail, not the stack. Products that fold the conversion into the vault contract (ether.fi's borrow mode [2]) tighten the boundary further; products that keep the conversion on an exchange ledger are the spend lane with better branding. The two-model test - whose keys move the collateral, whose ledger holds the claim - is the whole due diligence, and it takes ten minutes.

Field notes

Fund small, test the whole loop. On-chain funding at 1 sat/vB [8] is cheap; the expensive mistake is a large vault load before you have seen a deposit, a spend and a refund each clear once.
Keep the recovery regime yours. Passkeys, seed phrases, vault contracts - the failure mode of self-custody is self-inflicted. Whatever backup regime you trust for a hardware wallet applies double here, because the card depends on it.
Watch the mode you are in. ether.fi's direct-pay versus Borrow modes [2] are different risk products sharing one card - direct-pay sells per spend, Borrow carries liquidation geometry. Know which one the switch is set to.
Check the issuer's rail before the keys. Nuri's card layer is Wirex BaaS [5]; the issuer's banking health is the swipe's dependency, not your vault. The custody guide maps the failure modes.
Treat "hybrid" as a feature to verify, not a label. Blink's agreement documents both modes explicitly [6] - that is the standard a hybrid claim has to meet: the mode in writing, not in an FAQ.

Warnings and hard limits

Self-custody does not transfer to the claim. The moment the card owes the merchant fiat, that leg is custodial by construction. Anyone selling "fully self-custodial spending" is selling a settlement model that does not exist.
Nuri's status is verify, not active, for a reason. No public fee schedule, a documented network-label conflict, and a relaunched operator (Nurisoft Development OÜ - no relation to the insolvent 2022 Berlin Nuri, but also no long track record) [5]. Size the balance as if the card layer could stop tomorrow.
Vault risk is your risk. A self-custody vault means lost keys are lost coins, and a buggy contract is your problem. The operator insolvency risk of the spend lane becomes technology risk here - pick your poison knowingly.
No Lightning yet. 0 of this page's population takes Lightning funding - effectively none. Self-custody cards are on-chain products for now; the Lightning hub covers the fast-rail alternative.

The borrow hub prices the custodial version of collateral; the custody guide is the standing reference for what happens when operators fail. Your keys move the trust - they never remove it.

Sources

ether.fi membership levels - cashback ladders by tier, Core free - accessed 2026-10-05
ether.fi Cash - Borrow mode vs direct pay mode - accessed 2026-10-05
ether.fi transaction limits and fees - ATM model, FX margin - accessed 2026-10-05
Nuri homepage - self-custodial wallet, free Visa card, IBAN in 130 countries - accessed 2026-10-05
Nuri card support - card issued on Wirex BaaS, limits, declines, refunds - accessed 2026-10-05
Blink cardholder agreement v2.3 - Custodial and Non-Custodial balance modes - accessed 2026-10-05
IRS FAQ on virtual currency - the disposal that a conversion still creates - accessed 2026-10-05
mempool.space fee API - 1 sat/vB baseline for on-chain vault funding - accessed 2026-10-05
Freedomia FAQ - the custodial no-KYC contrast case - accessed 2026-10-05

FAQ

Can a payment card really be self-custodial?

Not the way a wallet is. The merchant receives fiat, the card networks settle in fiat, and the conversion is always operated by someone. Self-custody on a card means the *funding and collateral layer* is under your keys: ether.fi's card debits a vault you control [1][2], and Nuri's card settles against a passkey wallet that belongs to you [4]. The claim creation at settlement is still a third-party act - what changes is who can move the backing.

What does self-custody actually buy me here?

Two things, both real. Solvency insulation: a frozen or insolvent operator cannot take coins you hold in your own vault - they can cut the card, not the keys. Compulsion resistance: no operator freeze, no withdrawal gate on the balance itself. What it does not buy: the fiat leg still depends on the issuer's banking relationships, and a declined transaction at the terminal is possible in every model. Self-custody protects the stack, not the swipe.

How does the ether.fi model work exactly?

Two documented modes [2]: direct-pay (spend crypto as you go) and Borrow mode (the card draws a credit line against your vault collateral, so the crypto itself is never sold). Funding is on-chain - at a 1 sat/vB baseline [8] - into an architecture where the assets remain under your control [1]. Cashback runs 3% on the first $2,000/month at the free Core tier, 4% at VIP [1], with ATM limits of $1,000 per transaction and $5,000 per rolling 24 hours [3].

What is Nuri's model, and why is it verify-statused?

Nuri is a self-custodial wallet (Euros, Dollars, Stablecoins, Bitcoin) with a free Visa card whose cardholder support runs on Wirex BaaS [4][5]. The wallet keys are yours; the card layer is a Wirex BaaS product with its own limits (£500/day ATM, £30,000 transaction ceilings [5]). Verify status because the operator publishes no fee schedule and a network conflict (Visa branding versus a Mastercard-labeled agreement in the docs) remains unresolved [5] - the architecture is real, the economics are not publicly checkable.

What happens when a self-custody card is declined?

The decline is in the fiat leg - issuer policy, merchant category, or the settlement rail - and no key can fix it. Keep a fiat or stablecoin tranche for the payment the vault cannot settle. The custody guide documents decline-and-hold patterns across operators; they are not unique to self-custody products, but they are more visible here because there is no customer-service float to mask them.

Do self-custody cards avoid tax events?

No. Moving coins between your own wallets is not a disposal, but the moment the card converts value to fiat for the merchant, a taxable event exists (IRS [7]) - in direct-pay mode at each spend, in borrow mode only when the loan cycle realizes value. Self-custody changes custody, not the tax character of a conversion. The tax guide covers the lots.

Is the vault custodial during a borrow draw?

The collateral stays contract-bound to you; the drawn credit is a liability against it. That is the whole design difference versus a custodial spend card - the operator's claim is against a vault you control, not a ledger entry it owes you. The risk profile inverts: instead of trusting an operator's balance sheet, you manage liquidation risk - the borrow hub covers that math for the custodial variant.

Why do so few cards do this?

Because self-custody removes exactly the float that funds card economics - the operator can neither lend nor stake a balance it never held, and the issuing bank still needs regulated fiat rails. Two live products (ether.fi, Nuri) plus Blink's documented non-custodial balance mode [6] is the entire honest market at check time. Everything else marketed as "self-custody card" is redefining the word - ask whose ledger holds the coin between top-up and spend.

How do fees compare to custodial cards?

Favorably at the funding layer and honestly at the FX layer: ether.fi documents ~1% FX base plus a tier margin (0-0.5%) [3] against a market median of ~1%; on-chain funding at current 1 sat/vB base fees [8] costs cents. No monthly fee at Core [1]. The fee surface is cleaner than most exchange cards - the trade is that you operate the security yourself.

What happens to my vault if the operator disappears?

The keys - and the coins - remain yours; that is the design. The practical loss is the payment rail: a dead operator means no settlement leg, and pending rewards or drawn credit lines die with the agreement. Exit risk is therefore an *operability* risk, not a balance risk. Keep the recovery phrase in your own custody regime and the stack survives any operator wind-down.