Where Stablecoin Yield Actually Comes From

The sharpest thing written about the stablecoin yield ban surprisingly came from the White House, not from some crypto lobbyist.

In April 2026 the Council of Economic Advisers published a paper on what the GENIUS Act's yield prohibition would do to bank lending. A few pages in, it concedes that the prohibition "may not fully bind," and then names the live workaround by product, Coinbase's USDC Rewards, funded in part by Circle's revenue-sharing agreement (White House Council of Economic Advisers, April 2026).

That is the government's own reading of its own law, published months before the law switches on.

What people need to know is that a stablecoin is a dollar claim, and the dollars behind it sit in Treasury bills and overnight repo. Those assets earn interest whether or not anybody is permitted to pass it along.

Against roughly $280Bn to $310Bn of supply and a 3-month bill yielding between 3.71% and 3.87%, the sector's reserves throw off something close to $11Bn a year. Call it $30m a day. That is my arithmetic rather than a citation, and it is gross of distribution costs, cash drag and everything else.

The ban doesn't delete that money, it decides which balance sheet writes the check.

I spent years in banks before moving into crypto risk, long enough to watch that question settle a lot of arguments. Not whether the payment happens. Rather, which legal entity books it, and under whose rulebook.

This article argues that the GENIUS Act's yield prohibition will not remove yield from stablecoins. It will relocate it, toward distributors, exchanges and wrapper tokens. The United States already ran this experiment once, for nearly 78 years, and wrote down the results.

A Rule That Lasted Nearly 78 Years

On 29 August 1933, the Federal Reserve promulgated Regulation Q under section 11 of the Banking Act of 1933. Banks were forbidden from paying interest on demand deposits, and the Board set ceilings on what they could pay on time and savings accounts. The FDIC extended the rule to insured non-member banks, and by early 1936 it applied across the system (Federal Reserve History).

The reasoning was that competition for deposits had compressed spreads and pushed banks toward speculative lending, contributing to the failures of the early 1930s. Cap the price of funding, the thinking went, and you protect the lender.

What actually happened is that the money left.

On 15 November 1971 Bruce Bent and Henry Brown opened the Reserve Fund, the first money market mutual fund. The logic was plain. Regulation Q capped what a savings account could pay while market rates ran higher, so the fund bought the market instruments directly and passed the return to the shareholder. It was not a deposit, and it did not have to be.

Money market funds held more than $3.6Bn by 1975. By 1980 the figure was above $61Bn, and by 1982 it was above $230Bn, growth the Federal Reserve's own history records as coming mainly at the expense of banks and thrifts. That is roughly 64 times in seven years.

The demand deposit ban ran into a different workaround. Ronald Haselton, who ran Consumer Savings Bank in Worcester, Massachusetts, worked out that a negotiable order of withdrawal functions exactly like a check while being, legally, not a demand deposit. Congress authorized NOW accounts in Massachusetts and New Hampshire in January 1974, across all of New England in March 1976, and nationwide on 31 December 1980.

Sweep arrangements, Eurodollar deposits and commercial paper did similar work from other directions. None of them required anyone to break the rule.

DIDMCA started a six-year phase-out in 1980, and the rate ceilings were gone by 31 March 1986. The demand deposit ban itself survived until Dodd-Frank section 627 repealed it, effective 21 July 2011. From 29 August 1933 to 21 July 2011 is 77 years and 10 months.

R. Alton Gilbert's 1986 assessment for the St. Louis Fed, published under the title "Requiem for Regulation Q," concluded that the policy never achieved what it was designed to achieve, and that after 1966 it became actively disruptive, with institutions losing deposits every time market rates climbed above the ceilings.

Nearly 78 years. The payment never stopped. Only the payer changed. Unintended consequences of well-meaning regulation.

Follow the Dollar

Today's version of the same question has four answers, and they are all running simultaneously. The reserve income exists, it is large, and at every point in the chain somebody is deciding whether to keep it, share it, or hand it to the holder under a name the statute does not cover.

The Issuer Keeps It

Tether is the purest case. Its Q2 2026 attestation reports net operating profit of $1.5Bn, described as coming "mainly from interest on US Treasury holdings and repurchase agreements," against roughly $115Bn of Treasuries and USDT supply of about $184.6Bn, more than 60% of the market (Tether, August 2026).

Holder yield on USDT is $0. It has always been $0. The company reported more than $10Bn of profit for 2025 and $13Bn for 2024, and none of it flows to the person holding the token.

Two honest caveats. Tether's H1 2026 comprehensive result was negative by about $3.17Bn once unrealized losses on its gold and Bitcoin positions are included, so the operating engine and the company's own risk-taking are very different things. And these are attestations, not audits. Tether announced it had engaged KPMG for a first third-party audit alongside the Q2 numbers.

The Distributor Takes It

Circle is the opposite structure. In Q2 2026 it reported $701m of total revenue and reserve income, of which reserve income was $668m, or 95.3% of the total, on USDC in circulation of $73.3Bn (Circle, August 2026).

Then look at the cost line. Distribution, transaction and other costs came to $412m, roughly 59% of revenue, and net income from continuing operations was $48m. For the full year 2025, Circle paid Coinbase $1.4Bn against $2.7Bn of revenue, about 51%.

The terms explain why. Coinbase receives 100% of the reserve income on USDC held on Coinbase's platform and 50% from other channels. Circle's own S-1 put it plainly, that "the greater the proportion of USDC in circulation held on Coinbase's platform, the greater the proportion of reserve income payable to Coinbase."

Circle now has more than 150 distribution agreements. The issuer is not the capture point in this design. The distribution layer is.

The Platform Pays the Holder

From there it is a short step to the holder. Coinbase advertises around 4.10% on USDC, rising to 4.5% for Coinbase One subscribers, paid by Coinbase rather than by Circle. In February 2026 it restructured, setting 3.5% for Coinbase One and ending rewards for free accounts.

PayPal pays 3.7% on PYUSD. Kraken offers up to 1.75%, or up to 3.75% on its subscription tier. Robinhood advertises an estimated 7% on USDG, structured as lending through Robinhood Earn, and Coinbase launched a USDC lending product in July 2026 at around 7% for Coinbase One.

Two things are worth saying plainly. The highest advertised rates are lending products, which carry counterparty and liquidity risk that holding a token does not, so they are genuinely different products and not merely relabeled ones. And Coinbase's base rate sits above the 3-month bill, which I cannot explain from public disclosure and will not guess at.

None of this is a scandal, and none of it is currently prohibited. Every one of these payments is made by somebody other than the issuer.

The Wrapper Does the Same Job in Code

The on-chain version needs no exchange at all. Sky's sUSDS pays a savings rate set by governance, 3.75% through Q2 2026, funded from real-world asset returns, the Spark borrow rate and stability fees, on USDS supply of $6.65Bn.

Ethena's sUSDe is the outlier, because it is not a Treasury claim. Its return comes from a delta-neutral basis trade, long spot and short perpetual, plus staking. That rate fell from about 9.4% in April 2026 to about 7.1% in June and roughly 4.1% to 4.5% by early August, which tells you exactly what it is made of.

Tokenized Treasuries do it most directly of all. BlackRock's BUIDL sits at about $2.59Bn, Circle's USYC at $2.92Bn, Franklin's BENJI at around $2.3Bn, and Ondo's USDY pays roughly 4.8% (DefiLlama, August 2026). Readers who want the underlying mechanics of how these wrappers accrue will find them in the DeFi primer.

A wrapper is a separate token with its own issuer. The stablecoin underneath it still pays nobody.

The Banks Have a Real Case

The bank trades have made this argument harder to dismiss than crypto commentary usually allows. In January 2026 the American Bankers Association told senators that $6.6Tn of bank deposits were at risk. That figure traces back to a Treasury Borrowing Advisory Committee presentation from April 2025 which identified around $6.6Tn of US transactional deposits as the addressable base, a market-sizing estimate rather than a migration forecast.

The restatement is doing work the original did not. But the mechanism underneath it is real, and I would have made the same case from the other side of the table.

Retail deposits are the cheapest and stickiest funding a bank has. Replace them with wholesale funding and the liability side becomes more concentrated, more rate-sensitive and more expensive, and that shows up first in lending to small businesses and rural borrowers. Andrew Rodrigo Nigrinis made precisely that argument for the Consumer Bankers Association in April 2026, on the basis that the CEA had calibrated its model to a still-small market.

The letters have not stopped. BPI and the trade bodies filed on the OCC's 1 May deadline, BPI and the CBA filed on the FDIC proposal in June, and on 13 July 2026 the ABA, the ICBA and 76 state associations wrote to the Senate targeting the ambiguity in CLARITY section 404.

What the Ban Actually Says

Section 4(a)(11) of the GENIUS Act is one sentence, and the sentence is narrower than the headlines suggest. It provides that "No permitted payment stablecoin issuer or foreign payment stablecoin 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."

Read the subject of that sentence. It binds issuers. It does not bind exchanges, affiliates, distribution partners, DeFi protocols, or the issuers of a different token that happens to hold the first one.

The CEA said as much. The prohibition, it wrote, "does not explicitly bar intermediaries between issuers and holders from offering yield-like rewards funded by revenue-sharing arrangements with issuers. As a result, stablecoin holders continue to earn yield." It went further, noting that as of February 2026 the yield paid by stablecoin providers looked similar to that paid by high-yield savings accounts, "since both are passing through returns on Treasuries."

Regulators noticed. The OCC's proposed rule of 2 March 2026 attaches a rebuttable presumption of violation where an issuer has a contract with an affiliate or third party to pay yield and that party has a separate arrangement to pay holders (OCC, March 2026). The FDIC published a parallel proposal on 10 April 2026. Forbes described the gap in May 2026 as a Coinbase-shaped hole, which is memorable and, on the mechanics, fair.

Note what has not happened. Not one of these rules is final. The agencies missed their 18 July 2026 rulemaking deadline, so the statutory backstop now governs and the Act takes effect on 18 January 2027. Everything described above is happening in anticipation of a law that has not switched on.

Who Captures the Yield, Instrument by Instrument

The table below tracks the same dollar of reserve income through each structure. Sizes come from DefiLlama on 22 August 2026, rates from company disclosure, and the reference rates from the Federal Reserve's H.15 release for 20 August 2026.

Set the last two rows side by side and the scale argument mostly answers itself. Stablecoins are around 1.7% of US commercial bank deposits (Federal Reserve H.8, August 2026). No study I have found has identified measurable deposit outflows to stablecoins in the H.8 data, which is not the same as saying there have been none.

The CEA put numbers on it. Eliminating stablecoin yield raises bank lending by $2.1Bn, or 0.02%, at a net welfare cost of $800m, a cost-benefit ratio of 6.6. Stack every adverse assumption and the worst case reaches $531Bn of extra lending, but that scenario requires stablecoins at roughly six times their current deposit share and all reserves held in unlendable cash.

Its conclusion was blunt for a government paper, that "a yield prohibition would do very little to protect bank lending, while forgoing the consumer benefits of competitive returns on stablecoin holdings."

The Quarter That Complicates the Story

Anyone quoting the growth of yield-bearing stablecoins is usually quoting Q1 2026, and Q1 was spectacular. Yield-bearing supply grew more than 22%, adding about $4.3Bn of an $8Bn total quarterly increase, so roughly 54% of all net new stablecoin supply that quarter went into something that pays (CEX.IO, April 2026). Total supply crossed $315Bn.

Then Q2 went the other way. Yield-bearing supply fell by more than $3.5Bn, around 15%, ending a run that had lasted nearly three years, and the overall market posted its first quarterly contraction since Q3 2023, down to $312Bn.

The composition of that fall is the interesting part. Ethena's sUSDe dropped 52%, about $2Bn, as perpetual funding compressed. Sky's sUSDS fell 16%. Yet the Treasury-backed products grew through the same quarter, with USDY up more than 66%, USYC up around 16% and BUIDL up 2%.

That is not a retreat from yield. It is a rotation out of yield manufactured by a basis trade and into yield that is simply a T-bill claim wearing a different wrapper.

In my reading, that is the Regulation Q pattern arriving on schedule. Money did not leave the banking system in the 1970s because savers wanted exotic risk. It left because a plain government instrument paid more than the regulated account and somebody built a legal container to hold it.

What to Watch Between Now and January

Four dates carry most of the information.

The Senate takes its first procedural vote on the CLARITY Act on 15 September 2026, after Senator Thune filed cloture on the motion to proceed on 8 August. Section 404 is the contested provision, and a January draft would have applied the prohibition to digital asset service providers by name. Galaxy cut its odds of 2026 passage from 50% to 30%, and prediction markets have been pricing lower still.

Watch whether the OCC's rebuttable presumption survives into a final rule. That single drafting choice matters more than the headline ban, because it is the only proposed mechanism that reaches a payment made by somebody other than the issuer.

Then 18 January 2027, the statutory effective date. Because the agencies missed the July deadline, any final rule landing after roughly 20 September 2026 pushes its own 120-day clock past that backstop, so the backstop governs. Service providers get three years after that to stop handling non-compliant stablecoins.

The fourth date is the quietest and possibly the most consequential. FASB's proposed accounting standards update, issued 18 August 2026, would let qualifying stablecoins be classified as cash equivalents alongside Treasuries, commercial paper and money market funds, subject to on-demand redemption, direct redemption with the issuer and segregated reserves (FASB, August 2026). Comments close on 19 November 2026.

Sit those two frameworks next to each other. The statute insists a stablecoin is not a deposit and must not pay interest. The accounting standard would treat it as cash on a corporate balance sheet. A zero-yield cash equivalent backed by assets yielding close to 3.8% is an arbitrage waiting for a counterparty, and corporate treasurers are not known for leaving that on the table.

For anyone building a monitoring dashboard, three metrics beat the commentary. Circle's distribution cost ratio, which was 59% of revenue in Q2 2026. The split between Treasury-backed and synthetic yield wrappers, quarter by quarter. And the H.8 deposit series, which so far shows nothing.

The Workaround Has a Price

There is a coda to the Regulation Q story that the crypto version of this argument tends to skip.

On 16 September 2008 the Reserve Primary Fund broke the buck, its net asset value falling from $1.00 to $0.97. It was then the third-largest money market fund in the world at $62.5Bn, and it held $785m of Lehman Brothers debt. Lehman had filed the day before.

More than $40Bn was redeemed in two days. The wider run across money market funds reached roughly $439Bn. The fund had been founded by Bruce Bent, the same man who opened the first money market fund in 1971 to route around Regulation Q.

The SEC adopted reforms on 23 July 2014, with core provisions effective 14 October 2016, moving institutional prime funds to a floating net asset value with liquidity fees and redemption gates (SEC, July 2014).

Trace the whole arc. A rule intended to stop banks paying for deposits produced, across four decades, an uninsured shadow deposit system of more than $230Bn that broke in a crisis and had to be re-regulated twice.

My view is not that workarounds should be blocked. It is that a reroute moves risk to wherever the original rulebook is not looking, and the supervisory apparatus arrives afterwards, usually at speed and usually after somebody has lost money. That is the actual cost of writing a prohibition that the market can price its way around.

Where I Land

The most candid document in this whole debate is still the one the White House published in April, and it says the ban may not fully bind. I have not seen a serious rebuttal to that sentence, only arguments about how much the leakage matters.

My view is that the yield does not go away, because it cannot. It is the coupon on a Treasury bill, and somebody receives it every day whether the statute names them or not. The only open questions are who books it and how visible that payment is to a regulator.

The Regulation Q comparison is not a rhetorical flourish. It is the closest thing this debate has to a controlled experiment, and it ran for nearly 78 years across two full rate cycles and one financial crisis. The ban held on paper the entire time. The economics never obeyed it for a single quarter once market rates moved.

So here is a falsifiable flag. If, twelve months after 18 January 2027, the combined supply of yield-bearing stablecoins and tokenized Treasury wrappers is lower than it is today, I was wrong, and the prohibition did more than relocate the payment.

I do not expect that. I expect the number to be higher, the yield to be paid by entities with less name recognition than Coinbase, and a fresh round of letters from the bank trades asking for a broader definition. That is what the last reroute looked like from the inside, and it took Congress nearly 78 years to admit it.

Further Reading

• Open USD Has 140 Partners. Libra Had 28., a consortium stablecoin built to route reserve yield to its distribution partners rather than a single issuer, which is the capture-point question in this article compressed into one product design
• A 25-Year-Old Just Got a US Bank Charter. The OCC Is Telling You Something., the charter fight underneath all of this, because whoever holds the float decides who is allowed to earn on it