The law bans paying people to hold. Cards pay people to spend. Everything difficult sits in the gap between those two sentences.
The GENIUS Act says no permitted issuer “shall pay the holder of any payment stablecoin any form of interest or yield (whether in cash, tokens, or other consideration) solely in connection with the holding, use, or retention of such payment stablecoin.” Cashback on a purchase is paid for a transaction, not for holding, so at first read the card business is untouched.
Then look at how card programs are actually built. Reward rates are set by how much a customer keeps on the platform. Balances that earn while they sit in the app that issues the card. Distribution agreements in which an issuer pays a partner based on the tokens its users hold. Each of those touches the prohibition somewhere, and two live proceedings are deciding where the line falls.
What the statute covers, and what it leaves open
The prohibition binds issuers. It does not obviously bind anybody else, and that is the argument the whole market is having.
A Congressional Research Service note in March laid out the yield debate without resolving it. The law does not define “holder,” so it is unclear whether an exchange custodying tokens for its users is the holder or whether the user is. Under the structure in use today, an issuer passes reserve income to a platform, and the platform decides what to pass to its customers. If the platform is not a holder and not an issuer, the statute may not reach the payment at all.
The stakes explain the intensity. Banks want the prohibition read broadly and point to Treasury research identifying $6.6 trillion of transactional deposits as potentially at risk. Crypto firms read it narrowly and call the alternative incumbent protection. Neither side is being subtle.
The holder question, in one case study
The cleanest illustration is the arrangement everyone already knows about.
Writing in December, Duke’s Lee Reiners argued that Circle likely breaches the statute, because Coinbase is the holder of the USDC sitting in its custodial wallets and Circle’s payments to Coinbase rise with the share of USDC held there. He put the 2024 figure at about $907.9 million. On that reading, the payment is interest to a holder, whatever the contract calls it.
The counter-reading is that end users are the holders and the platform is a distributor being paid for distribution. That is the holder question in a sentence, and it decides whether the largest revenue-sharing arrangement in the industry survives in its current form. How that income gets divided between issuers and their distributors is its own subject; what matters here is that the same pipe funds a good share of what cards hand back to customers.
What the OCC proposed
The banking regulator did not wait for the argument to settle.
The OCC’s proposal would bar issuers from paying any interest, yield or rewards “whether in cash, tokens, or other consideration” in connection with holding a stablecoin, and it adds the part that reaches card programs. An issuer would be presumed to violate the prohibition if it has a contract, agreement or other arrangement with an affiliate or a related third party to pay that yield, and the definition of related third party captures exchanges and third-party wallet services that support the token.
It is a rebuttable presumption, so an arrangement can be defended. It also carries two carve-outs that matter commercially: merchant discounts offered independently stay permitted, and profit-sharing inside white-label or co-brand deals may be allowed.
Read together, the proposal does not ban cashback. It bans routing reserve income to holders through a partner and calling it something else.
What the Senate would add
The second proceeding is legislative and lands this week.
The compromise in the Senate Banking Committee’s text bars crypto firms from paying interest or yield on stablecoin balances that is “economically or functionally equivalent” to bank deposit interest, while preserving rewards tied to “bona fide activities” such as payments, transfers and trading. That would be the first time the distinction cards depend on is written into statute instead of inferred from one.
The bill faces a September 15 cloture vote on the motion to proceed, needing 60 votes, and stablecoin rewards is one of the disputes still open going into it. Banking groups say the language leaves a yield-shaped hole. Crypto firms say tighter wording would end competition. Whatever happens on the floor, the drafting tells card programs what the safe shape looks like: a reward attached to an activity, not to an amount sitting still.
Two buckets, and where programs sit in them
Sort every incentive a card program offers into one of two categories, and most of the uncertainty disappears.
The first bucket is paid for spending. Cashback per purchase, points on transactions, merchant-funded offers, a higher rate in a category. These are transaction rewards, they exist across the whole card industry, and nothing in either proceeding threatens them.
The second bucket is paid for holding. Interest on an in-app balance, an advertised annual percentage yield on stored value, a promotional rate on money parked in a wallet. These are the target. The statute reaches them when an issuer pays, the OCC proposal reaches them when an affiliate or related third party pays on the issuer’s behalf, and the Senate language would reach them when they look economically equivalent to deposit interest.
Most card programs sit safely in the first bucket. Coinbase’s card pays up to 4% back in bitcoin on purchases. Gnosis Pay’s cashback pays in its ecosystem token, funded by a dated program, as this site noted when it examined the card enabler layer. A card paying for spending is a card doing what cards have always done.
The gray zone is where product design lives
Then there are the structures that sit in both buckets at once, and this is where legal review earns its money.
Start with tiering by balance. Coinbase’s bitcoin-backed rates run at 2%, 2.5%, 3%, and 4% depending on the customer’s assets on the platform at the moment a purchase starts. The reward is paid on a transaction, which is bucket one. The rate is set by holdings, which resembles bucket two. Nothing in the current text says which way that resolves, and a program relying on the answer should not assume it is the comfortable one.
Next, the bundle. The European market shows the pattern already: a card marketed with tiered cashback alongside a separate savings product paying an annual yield. Reporting on September 1 described one exchange card launched at the end of August with cashback tiers of 4%, 6% and 10% against monthly caps, sold next to a flexible savings product advertising up to 7% annualized. The cashback is defensible. The savings product is exactly what the ban addresses, and the marketing that presents them as one offer is the risk.
The funding source is the third case. A reward paid out of interchange is a cost of acquisition. A reward paid out of reserve income routed through a partner is the thing the OCC presumption is aimed at. Two programs can advertise the same percentage and sit on opposite sides of the line, and only one of them can show a supervisor where the money came from.
What interchange can actually fund
Strip out reserve income and a reward budget has one main source, which puts a ceiling on the headline rate.
In the United States the ceiling is comparatively high. Credit interchange averaged 2.36% of transaction value in 2025 by the Merchants Payments Coalition’s count, which is why a 2% to 4% cashback offer can exist at all, though the top tiers usually carry monthly caps that bring the blended rate down to something interchange supports.
In Europe, the ceiling is written into law. Regulation 2015/751 caps consumer interchange at 0.2% of the transaction for debit cards and 0.3% for credit cards. A European program paying 4% out of interchange would be paying roughly twenty times what it collects, which is why the generous European offers are funded by an ecosystem token or a venture budget instead, and why those offers change when the budget does.
That contrast is the practical forecast. As the reserve-income route narrows, advertised rates converge on what the local interchange regime funds, plus whatever a token treasury is willing to spend on distribution.
Read more: Rain’s $250 million round prices stablecoin card plumbing • PayPal rebuilds around crypto as PYUSD’s float shrinks
Europe already ran this experiment
The European Union settled the holding question three years ago and produced a market that still has cashback.
MiCA’s Article 50 says issuers of e-money tokens “shall not grant interest in relation to e-money tokens,” extends the same bar to crypto-asset service providers, and then closes the obvious workaround: any remuneration or benefit related to the length of time a holder holds the token counts as interest, including compensation, discounts or pricing benefits provided directly or through third parties.
That is a broader rule than the American one, and European card programs did not disappear. They pay for spending. The lesson for United States programs is not that the ban is survivable in the abstract. It is that the ban is survivable once rewards are tied to activity and the paperwork can prove it.
What a program should do before January
Four pieces of work separate a program that is ready from one that will be rewriting terms in a hurry.
Document the funding source of every reward, per program, in a form a bank partner can review. Interchange, ecosystem token, merchant-funded and reserve-derived are four different answers, and only the last one carries regulatory weight.
Separate any yield product from the card product, in the contract and in the marketing. A savings balance and a spending card can coexist. A single page that advertises them as one offer invites the reading nobody wants.
Re-examine balance-based tiering. If reward rates depend on holdings, decide now whether the program can defend that as a loyalty structure or would prefer to tier on spending volume, which raises no question at all.
And read the distribution agreements upstream. If the program’s economics depend on a share of reserve income that ultimately reaches customers, the OCC’s presumption is pointed at that arrangement, and the fix is a contract change rather than a marketing one. PayPal’s experience with promotional rewards on PYUSD balances shows how quickly a program can be withdrawn when the regulatory reading shifts.
What rebutting the presumption looks like
The OCC’s presumption is rebuttable, which means the burden lands on documentation rather than on argument.
A program in that position needs to show three things. That the payment to the partner is not calculated on balances held, since a formula keyed to holdings is the fact pattern the rule describes. That the reward reaching the customer is computed per transaction, with the ledger to prove it. And that the contract does not oblige anyone, anywhere in the chain, to pass reserve income to holders.
Programs whose agreements were written before this proposal existed will often fail the first test without intending to, because tying a distributor’s payment to assets under custody was the normal way to write these deals. Reopening that clause is slow work, and January is not far away.
What happens next
Three dates frame the rest of this.
The Senate vote on September 15 determines whether the activity-based carve-out gets written into law this year or stays a proposal. The OCC has said it wants final GENIUS Act rules by November, which is when the presumption and its carve-outs either become binding or get softened. And the law takes effect on January 18, at which point a program either has a defensible answer for each reward it pays or does not.
None of this threatens the thing customers actually respond to. Cashback on spending has survived every version of this rule in every jurisdiction that has written one. What is ending is the quieter arrangement underneath, where a balance earned something because the reserve behind it did, and nobody had to say so on the marketing page.