The Bank for International Settlements fired a warning shot on August 28 that was immediately misread by the market. Chief Pablo Hernández de Cos suggested that stablecoins could make borrowing more expensive. The market interpreted this as a macro statement about interest rates. It is not. It is a statement about the structural mechanics of bank funding—and the endgame of banks issuing their own digital liabilities.
As a smart contract architect, I read this warning through a different lens: the ABI of a bank's balance sheet is being redefined, and the function being deprecated is the demand deposit.
When a bank issues a stablecoin, it is not just launching a product; it is executing a balance sheet transformation that changes the vector of its funding. We are witnessing the unbundling of the core banking product—the transaction account—into a payment rail and a lending pool. The average market participant sees institutional FOMO. I see a codebase where the new features accidentally break the old main loop. Static analysis revealed what human eyes missed: the bank's most critical invariant is being violated by its own innovation.
The stablecoin market cap sits near $304 billion. Tether holds roughly $183 billion. USDC accounts for $74 billion. These are no longer speculative tools sitting in crypto wallets. Federal Reserve researchers are beginning to classify these as direct competitors to the traditional "transaction account." That categorization is the most consequential shift in the stablecoin narrative since the 2022 collapse of Terra.
Arthur Firstov, Chief Business Officer at Mercuryo, articulated this transition sharply in his analysis: "Stablecoins stopped being a crypto product and became a payments product." He is correct to draw this line. Banks have long held the position that stablecoins were a niche crypto phenomenon—a market of hobbyists operating outside the regulated perimeter. That defense mechanism collapses when a Fortune 500 corporate treasury uses USDC for cross-border settlement.
A September 2025 Federal Reserve survey indicates that roughly half of respondent banks are prioritizing growth in at least one stablecoin or digital-asset area over the next three years. This is not adoption; this is capitulation. Banks are adapting to a protocol they cannot fork.
The core issue is not the existence of stablecoins, but the legal character of the liabilities banks are now creating. The market often conflates a "tokenized deposit" with a "bank-issued stablecoin" as if they were interchangeable abstractions. This is a fundamental misreading of the architecture.
Take J.P. Morgan's JPM Coin. It represents a bank deposit on a blockchain. It is a liability of the bank, insured, and subject to capital requirements. Conversely, Société Générale-FORGE's CoinVertible is a MiCA-regulated stablecoin backed by segregated collateral. One is a promise created by the bank. The other is a receipt for an asset held in escrow.
Nitin Gaur, Head of Institutions at Nethermind, makes this distinction with precision: "The interesting question stopped being whether a bank can issue and became what a bank is issuing. A tokenized deposit and a bank-issued stablecoin are two different liabilities with different legal character, different capital treatment, different insurance status and different settlement properties."
This legal variance is not a semantic nuance. It determines the bank's capacity to lend.
Under the US GENIUS Act, a payment stablecoin requires one-to-one backing with eligible reserves, such as cash or short-dated Treasuries. The Treasury proposed implementation rules on August 17. This regulation transforms the stablecoin into a matched, non-lendable reserve pool. Gaur elaborated on the balance sheet consequence: "A stablecoin issued under a GENIUS pathway is not a deposit. It is a payment instrument backed by segregated reserves the issuer cannot lend against. When a treasurer moves a hundred million from a demand deposit into the bank's own coin, the bank has converted a funding source into a matched, non-lendable reserve pool."
I have spent years auditing smart contracts for institutional clients. The most common vulnerability I find is not reentrancy—it is mis-assignment of ownership. The bank-issued stablecoin does precisely this. It re-assigns the "ownership" of the underlying asset from the bank's lending pool to a segregated wallet, effectively deleting the bank's ability to utilize that capital. In code, we call this a loss of composability. In banking, we call it a liquidity drag.
If a bank treasurer moves $100 million from an interest-earning account into the bank's stablecoin, the bank has cannibalized its own loanable funds. The bank trades a liability it can invest (the demand deposit) for a liability it must match with idle reserves (the stablecoin). The lending capacity is mathematically reduced. The curve bends, but the logic holds firm.
The wider effect depends on where reserves end up. If the stablecoin holder keeps the asset on a centralized exchange, the funds may ultimately be re-deposited back at banks, providing circular funding. However, the velocity and concentration of that funding differ drastically. The investment horizon of a stablecoin holder is measured in milliseconds of execution, not in the long division of a custodian. Adrian Wall, Managing Director of the Digital Sovereignty Alliance, identifies the systemic risk: "If stablecoin adoption ultimately shifts funding away from bank deposits rather than recycling those funds back into the banking system, banks could face higher funding costs and potentially less capacity to extend credit."
This is the transmission mechanism from the BIS warning to your mortgage rate. As banks lose low-cost core deposits, they replace them with more expensive, more volatile wholesale funding. The price of money goes up. The cost of credit follows.
The demarcation between the "deposit" and the "coin" is the dividing line between banking and payments.
From a protocol perspective, payments are just a state update. Banking is a state update with a credit component. When a bank issues a stablecoin, it is stripping its core functionality down to a payment primitive, removing its lending advantage, and exposing itself to competition from entities that have no lending overhead.
The market data suggests adoption of these new rails is not hype-driven. In July, Citi reported a dollar payment from London to Thailand over a US holiday weekend, using its tokenized-deposit service. This is a concrete demonstration that these rails operate when traditional systems are offline—a functionality that resonates with corporate treasurers who fear settlement risk. Western Union launched USDPT in May, with Anchorage Digital Bank issuing the stablecoin on Solana.
The scale of these projects differs. J.P. Morgan reports around $7 billion in daily activity across Kinexys products. CoinVertible, by contrast, reported €156.6 million of euro tokens and $12.55 million of dollar tokens outstanding on August 31. This is a classic apples-to-oranges comparison: daily transaction volume versus circulating supply. These figures cannot establish which model is winning, but they indicate that even the "losing" model is attracting nominal volume.
This proliferation of bank tokens creates a new problem: fragmentation. If every bank issues its own coin, we get dozens of thin, illiquid pools instead of one deep, liquid instrument. Users would be relegated to exchanging one bank's token for another at unknown conversion rates during market stress. Connecting the technology does not guarantee parity at face value during a liquidity crisis.
Europe's Qivalis consortium is trying to solve this fragmentation problem without a token standard. It has assembled 37 banks across 15 countries around a planned euro stablecoin, targeting a launch in the second half of 2026. Ernesto Olmedo Pereira, Head of Strategy & DeFi at Qivalis, states: "If every bank launches its own token, you get dozens of thin, incompatible pools instead of one deep, liquid euro instrument."
This is an acknowledgment of the standardization problem that web3 developers solved at the application layer with ERC-20. But while the token standard is universal, the banking model is not. Qivalis is essentially building a private interoperability layer for European deposits. The question is whether they can replicate the utility of a 37-bank consortium against the convenience of a stablecoin that is simply... issued by one entity and used everywhere.
The contrarian blind spot in the bank-issued stablecoin narrative is the assumption that regulated money is inherently more trustworthy than collateralized money. This is a false equivalence rooted in the "too big to fail" doctrine.
A bank deposit is a claim on the bank. If the bank fails, the deposit requires a taxpayer bailout (or insurance fund) to be made whole. A GENIUS Act-compliant stablecoin is a claim on a segregated reserve pool. If the issuing bank fails, the stablecoin should be redeemable from the reserves. The bank-issued stablecoin is actually a more sovereign asset than a bank deposit—it does not require the bank to be solvent; it only requires the reserve pool to be accurate.
This is where my audit background kicks in. I have audited custody solutions where the "segregated" reserve was held by a sub-custodian of the issuer's parent company. The segregation existed in the documentation, not in the chain of custody. In the digital asset space, we call this "multisig on a single hardware module"—a false sense of security based on a logical separation that doesn't exist in physical reality.
If a bank places "segregated reserves" in a special purpose vehicle that is consolidated for tax purposes, the legal separation is a fiction during a bankruptcy. The court will look at the economic reality. The stablecoin holder would be a creditor of the bank, not a secured party. This is an unaddressed legal risk that no amount of "miCA compliance" can fully mitigate.
During my work auditing institutional custody smart contracts, I identified a critical flaw in role-based access control that could allow a compromised administrator to drain funds. The fix involved a simple check for a block number in the authorization logic. The point is that invariants must be enforced in code, not in risk policies. Similarly, the stability of a stablecoin relies not on the bank's promise, but on the verifiability of the reserve on an immutable ledger. If the ledger is just a database controlled by the bank, we are back to square one: trusting the institution.
The BIS warning is not a prediction about the future; it is a description of the present. Stablecoins are forcing banks to accept a lower return on a substantial portion of their liabilities. This is a structural tax on bank profitability. We build on silence, we debug in noise.
The misconception is that banks will pass on these costs to stablecoin users. They will instead pass them on to legacy loan customers. The credit-intermediation premium will rise as banks seek to maintain return on equity while holding non-lendable reserves. Borrowing costs will increase, not because of monetary policy, but because of the changing composition of bank funding.
As a final thought on the Qivalis model, it is the only structurally sound approach to private money issuance. A common rail reduces fragmentation. But the governance of a 37-bank consortium is a nightmare of competing incentives. In my experience with smart contract governance forums, we call this "the finality problem": everyone agrees the state should be updated, but no one agrees on when. Qivalis requires true off-chain coordination that must be codified into on-chain settlement.
Banks are entering a world where the codebase is global, but the jurisdiction is local. The GENIUS Act and MiCA are not just regulatory frameworks; they are fork conditions in the consensus of global finance. The block confirms the state, not the intent; the bank holds the balance, but not the trust.
Ultimately, the stablecoin race will bifurcate the traditional banking model into two distinct categories: banks that issue payment instruments (and become low-margin utilities) and banks that provide credit (and become high-margin risk takers). The middle ground—the all-purpose bank that issues a stablecoin, accepts deposits, and lends—will face a constant audit of their own balance sheet, with the auditor being the market itself.
The future of banking is not a question of code experimentation. It is a question of whether banks can survive the removal of the demand deposit from their core function. The invariant that gave them value—the ability to lend against fractional reserves—is being challenged by an architecture that prefers full reserves for digital assets. The curve bends, but the logic holds firm.