Which balance actually earns

93 of 139 obtainable crypto cards advertise yield. 38 pay it on the balance the card spends from.

“Earn while you spend” is the line every yield card uses, and for most of them the two halves are describing different pots of money. The card spends from one balance. The interest accrues on another. Nothing in the marketing distinguishes them, because the distinction is the only thing that would make the offer sound smaller than it is.

So we went through every obtainable card in the database and asked one question of each: when this product pays yield, is it paying on the money the card debits, or on money held somewhere the card cannot reach? The answer is a field in our public dataset, not an opinion, and it turns out to split the market roughly in half — with the larger half on the side readers assume they are not on.

Key findings

1. 38 of 139 obtainable cards pay yield on the balance the card spends from. That is the group where the phrase “earn while you spend” describes the mechanics rather than the mood.
2. 50 more advertise yield and pay it somewhere the card cannot debit. A separate Earn account, a savings product, a staking programme. Real yield, real rate, and you have to move the money before you can buy anything with it.
3. 5 claim yield and never say which balance earns. Not a rate that is hard to find — a mechanism that is never stated anywhere the customer can read it.
4. A reader trusting the marketing would over-count by 55 cards. That is the size of the gap between “this card has yield” and “this card grows the money I am about to spend”, and it is bigger than the group it is mistaken for.
5. Only 12 of the 38 pay by default. The other 26 need something switched on, subscribed to, or moved first. Deposit and walk away and they earn nothing.
6. 7 of the 38 publish no rate at all. They advertise that the balance earns without saying how much, which makes the product uncomparable by design.

Where the yield actually sits

Every obtainable card falls into exactly one of four states. The chart counts cards, not products, and the two middle bars are the ones that get conflated.

The interesting comparison is not yield against no-yield. It is the 38 against the 50: two groups that market themselves in the same words and behave differently every time you use the card. On one, the balance keeps accruing until the transaction clears. On the other, spending means first moving money out of the thing that was earning, which is a step the advertising never mentions and the rate never accounts for.

What the 38 actually require of you

Knowing a card is in the right group is half an answer. The other half is what it makes you do before anything accrues, and here the products separate into three shapes.

The cards that debit a vault at the till

10 of these products hold the money in something branded as an Earn account, a strategy or a vault, and then let the card spend it in place. Nexo Card, Ether.fi Cash Card, Exa Card, Bybit Card, KuCard, Tuyo Card, Bit2Me Card, Plasma One Card, Xapo Bank Card and Brighty Card all work this way. When you pay, the issuer debits the earning product directly rather than asking you to withdraw first, so the balance accrues until the transaction settles.

This is the class most often filed wrongly, and wrongly in both directions. Read the product name and you conclude the money is parked away from the card, which is the mistake that pushes a genuine on-balance card into the wrong column. Read the marketing and you conclude nothing is required of you, which is the mistake in the other direction — every one of these needs the money in the vault first, and Bit2Me in particular funds the card from a non-earning wallet unless you go and change it.

The distinction that settles it is the deduction order, not the branding: if the issuer documents the card pulling from the earning product at purchase, the money is spendable in place. If the issuer documents a transfer step, it is not.

The 5 that will not say

These issuers advertise yield and never state which balance earns it. This is not us failing to find a rate. In each case the claim is public and the mechanism is absent from the site, the help centre and the terms.

We class these as unstated rather than guessing, because the flattering guess and the unflattering one are equally unsupported and only one of them costs the reader money. A field with no value for “unknown” will invent one, and it will invent the answer that makes the product look better.

The honest counter-arguments

Separate Earn accounts are not a trick. A card that pays 8% inside a savings product and nothing on the spending balance may still beat a card paying 2% on the spendable balance, if you keep most of your money in the savings side and top up deliberately. The distinction on this page is about what the advertising implies, not about which product is better. It only becomes a defect when “earn while you spend” is printed over a mechanism that does not.

Every rate here is advertised, none is measured. 7 of these cards publish no figure, and the ones that do are quoting ceilings: the top tier, the promotional headline, the rate before commission. Uphold shows its yield net of a 20-25% cut, which most issuers do not do. Treat the rate column as what the issuer says, and see advertised versus measured for the handful we have put a number on ourselves.

The line between the groups is a judgement in a small number of cases. An auto-sweep into a same-day earning balance counts as on-balance here; a manual transfer does not. Products near that line — where the sweep is automatic but the destination is separately branded — could reasonably be filed either way, and we have named the gate on every row so you can disagree with a specific card rather than the whole count.

Rates move and tiers change. Everything on this page was read on the dates in the dataset. Variable rates are variable, and a ceiling advertised today is not a commitment for next quarter.

Questions this research answers

Methodology

The population is every card in the Sweepbase database that a consumer can currently obtain — 139 of them. Cards closed to new applicants and cards announced but never issued are excluded, because a yield rate you cannot sign up for is not an option.

Each card carries a free-prose field describing its yield product, and each of those fields opens with one of four fixed phrases: the balance the card spends from earns, the yield exists but not on that balance, the issuer claims yield without saying which balance earns, or there is no yield. That contract is enforced by a test, so a row that stops matching fails the build instead of quietly landing in the wrong bucket. The four counts on this page are read from those prefixes at page-render time and are never typed by hand.

The test for the first group is whether the money you can pay with grows without being moved into a different product. A vault or auto-sweep attached to the same spendable balance passes. A separate account you have to transfer out of does not, however good the rate.

The three mechanisms — pays by default, needs a toggle, debits a vault — are a reading of issuer documentation rather than a field in the dataset, so they are curated separately and cross-checked against it: every card classified here must still bucket as on-balance in the data, and every on-balance card must have a classification. A card added to the database without one fails the test rather than silently shrinking the table.

Rates are transcribed from each card’s own row, and every figure in the rate column is checked against that row automatically — a number copied from the neighbouring entry fails the build. This is a real failure mode rather than a hypothetical one: it is how a rate belonging to one card ended up printed against another during an earlier audit.

What this page does not do is verify that any advertised rate was paid. Advertised yield is a claim by the issuer, and we treat it as one throughout.

Corrections

This page was first published on August 11, 2026. Corrections will be logged here with dates, in line with our editorial policy. If a card is filed in the wrong group, the fastest fix is to tell us which issuer sentence contradicts us.

Sources: issuer help centres, terms and fee schedules for every card listed. Those documents were last read on July 20, 2026, which is the date recorded against every row in this dataset; this page was assembled from those readings on August 11, 2026 and did not re-open each source. Card-level facts come from the Sweepbase dataset, published under CC-BY 4.0 so the counts here can be reproduced independently. Ratings and the formula behind them are at our rating methodology. Spot an error? Report a correction.

About the author: Mihail B. is the founder and editor of Sweepbase, which tracks the crypto card market card by card. Sweepbase earns affiliate commission on some cards; the rating formula does not read the affiliate field, and no placement on this page was paid for — see the affiliate disclosure and how we test.