There is a pattern I keep seeing on the custody side of this industry, and the people repeating it are not unsophisticated. A treasurer at a mid-size fintech, after a long conversation about cold storage architecture, tells me his board is convinced they can replace their bank relationship with bitcoin plus a laddered Treasury position. He wants the multisig walkthrough. I give him the multisig walkthrough. But the framing is wrong, and at a $1.65T bitcoin market cap and an $83,000 spot print, the framing is wrong at a scale that now matters. Holding the asset is not the same operation as creating the credit. That distinction is what this piece is about.
The Balance Sheet Confusion That Started All This
The pattern, stated plainly: people who can read a 10-K assume that if they assemble the right asset mix in cold storage, they have reconstructed the function of a bank. They have not. They have reconstructed the asset side of a bank balance sheet. The liability side is missing, and the liability side is the whole point.
Concession first, because this argument deserves it. Bitcoin at $83,000 with a circulating supply of 19.8 million coins is a real asset. A 1.65 trillion dollar market cap is not a rounding error. Short-duration Treasuries are the cleanest collateral the modern financial system produces. If your question is "do I want exposure to these two things on my balance sheet," the answer can absolutely be yes, and a qualified custodian like Anchorage Digital under its OCC Federal Trust Charter can hold both for you in a setup that is operationally tighter than most regional banks. I will concede every inch of that.
What I will not concede is the leap from "I hold these assets" to "I have replicated digital credit." Bank credit is not a thing you hold. Bank credit is a thing that gets created when a bank originates a loan and writes a corresponding deposit into existence on the same line. The deposit funds the loan. The loan creates the deposit. That circular act of issuance is the entire mechanism. Bitcoin does not do this. Treasuries do not do this. A treasurer can stack both in a Coinbase Custody account under New York DFS Trust supervision, and at the end of the exercise the treasurer has assets. The treasurer does not have the ability to write a credit line into existence against the firm's own books. Those are different operations performed by different entities under different licenses.
The board members I keep meeting have read the bitcoin-as-pristine-collateral argument, they have internalized the Treasury-as-cash-equivalent argument, and they have stapled the two together into a thesis that the combination reproduces the bank. It does not. It reproduces the part of the bank that holds inventory. The part that issues new liability-side money against new asset-side claims — that part is absent, and assembling more inventory does not summon it.
The Settlement Finality Trap Nobody Talks About
Pattern two. The same treasurers who insist they have replaced their bank relationship have not run the settlement-finality math, and when you run it for them they get quiet.
Bitcoin settlement finality is probabilistic, not legal. After six confirmations the probability of a reorg is small enough that exchanges treat the transaction as final — but the finality is a statement about the network, not a statement about a court. If a bitcoin sent from a Ledger device to a Trezor device is later disputed, no clearing house unwinds it. The chain says what it says. Compare that to a Fedwire transfer between two depository institutions, which is final under Regulation J, irrevocable, and enforceable. The bank infrastructure carries legal finality. The bitcoin infrastructure carries cryptographic finality. Those are not the same product, and treating them as substitutes is where the credit thesis starts to crack.
Treasuries are worse on this dimension than people assume. A short-duration Treasury position held in a custody account is legally a beneficial interest in securities held by a custodian who holds an interest at the DTC. There are three layers of indirection between the holder and the obligor. When a fintech treasurer says "I hold Treasuries," what they hold is an entry in an omnibus account. Settlement of the Treasury market runs through Fedwire Securities Service. It is fast, it is robust, it is the system the rest of finance trusts — but it is also a system the treasurer does not directly access. The custodian accesses it on their behalf. So the asset is sound and the settlement layer is sound, but neither of those properties travels with the holder to a context where they want to issue credit against the position.
This is where the substitution argument breaks in practice. A bank uses its asset book to back a credit facility because the bank is the entity authorized to issue the credit. The fintech treasurer holding the same assets through a custodian is not. The treasurer can pledge the assets as collateral for credit issued by someone else — a bank, typically — which lands them right back inside the relationship they were trying to escape. The walk-around did not work. It ended at the same counter, just with extra steps and a custody bill.
The asset side of a bank balance sheet is the easy half to copy and the half that does not matter without the licensed liability side that creates the money.
The Maturity Transformation Gap You Cannot Engineer Around
Banks do something that sounds simple and is structurally impossible to replicate with a stack of bitcoin and a Treasury ladder. They borrow short and lend long. Demand deposits, which the bank owes on a moment's notice, fund commercial loans the bank will not be repaid on for years. The spread between the two is the bank's gross margin, and the regulated capacity to run that mismatch is most of what a banking license actually grants.
The treasurer with bitcoin in cold storage at Fidelity Digital Assets and Treasuries laddered out to twelve months has done the opposite of maturity transformation. They have matched the duration of their assets to a defined holding period and they hold those assets to themselves. There is no liability funded by short-term obligations that gets deployed into long-term claims on borrowers. There is no spread. There is just inventory, and inventory at $83,000 a coin is volatile inventory that is now marked-to-market against a Treasury position which is not. Aggregating two assets with uncorrelated returns is portfolio construction. It is not credit creation.
People sometimes counter this with the DeFi lending argument — that a bitcoin position can be deployed into a protocol where it earns yield against borrowed stablecoins, and that this approximates the bank's spread business. It does not. It approximates a margined inventory loan against the bitcoin, which is the bank-doing-the-lending side from the opposite direction. The borrower of the stablecoins is the treasurer's counterparty, not the treasurer themselves. Whichever way you slice it, the entity creating the credit is a different entity from the entity holding the collateral. If the treasurer's goal is to be the credit-creator, this architecture leaves them in the credit-borrower seat, with a custody bill on top.
And the bigger problem with the DeFi-as-credit-substitute thesis is that DeFi credit is overcollateralized. A bank issues a loan against future cash flows of a borrower that often exceed the collateral or replace collateral entirely with covenants and personal guarantees. DeFi cannot do this. The chain has no enforceable claim on a future cash flow that does not exist on-chain. So DeFi credit is structurally a different product — it is collateralized inventory finance, not credit. Calling the two interchangeable is a category error, and category errors get expensive when a fintech CFO acts on one.
The Custody-Is-Not-Credit Conflation That Keeps Getting Repeated
Last pattern, and the one that triggered this piece. Custody and credit are different industries that share none of the same regulatory permissions, and watching people conflate them on Crypto Twitter has stopped being funny.
Custody is a fiduciary holding function. Coinbase Custody operates as a New York DFS Trust Company. Fidelity Digital Assets is a New York DFS Trust. Anchorage Digital holds an OCC Federal Trust Charter — the first crypto-native institution to get one, and the closest thing to a bank charter the industry has so far produced — and even that charter is a trust charter, not a depository charter. A trust charter authorizes the custodian to hold assets on behalf of a client. It does not authorize the custodian to create money against those assets. The OCC was careful to draw exactly that line when it granted the Anchorage charter, and the line is the entire reason the charter was grantable at all.
Hardware wallet vendors are even further from credit issuance, and I say this with affection for the category. Ledger ships devices from Paris. Trezor ships devices from the Czech Republic under SatoshiLabs. GridPlus Lattice1 ships hardware that uses a co-signer abstraction which is genuinely clever for institutional setups. None of these companies issue credit. None of them clear payments. They sell a secure element that signs transactions. That is a critical piece of plumbing for self-custody and it is not the plumbing of a bank.
The conflation happens because the surface vocabulary overlaps. A custodian "holds your assets" the way a bank "holds your deposits." But the bank's deposit is the bank's liability, redeemable on demand, fungible with every other dollar in the banking system, and able to fund the bank's loan book. The custodian's holding is not the custodian's liability — it is the client's asset, segregated, not fungible with the custodian's own balance sheet, and contractually unable to fund anything else. Those are opposite legal constructions wearing similar surface language. Treating them as interchangeable is how the treasurer ends up explaining to a board, six months later, why the bank relationship is back and why it was always going to come back. The substitution was never on the table. The vocabulary made it look like it was.
The version of this argument I find most worth running for boards: digital credit, as a product, exists in well-defined places. It exists at depository institutions that hold demand-deposit charters. It exists, in a degraded but functional form, at money market funds that hold redemption claims against short-dated instruments. It is starting to exist, in early and not-yet-resolved form, at stablecoin issuers whose regulatory status is the live legal question of the decade. It does not exist in a cold-storage account holding bitcoin and Treasuries. Not at the volumes the question implies, not under the supervision regime the question implies, and not at the speed the question implies.
So What Do You Actually Do
If your firm needs credit, you maintain a bank relationship. You do not have to like it. You do not have to enjoy paying for it. But you do not, in 2026, replace it with assets — you replace it, eventually and partially, with a stablecoin and tokenized-deposit architecture that nobody has built yet, and you accept that the architecture is not built yet. Pretending it is built because bitcoin trades at $83,000 and Treasuries are easy to ladder is a category error that boards make when they have read the asset-side argument and have not run the liability-side argument. Run the liability-side argument before you sign off on the substitution.
If your firm wants exposure to bitcoin as an asset on its balance sheet, that is a separate question with a clean answer. A qualified custodian — Anchorage if you want the OCC charter, Coinbase Custody or Fidelity Digital Assets if you want the New York DFS trust frame — solves the holding problem at an operational standard most fintech treasurers cannot match in-house. Pair the custody account with a self-custody secondary using Ledger or Trezor or a GridPlus Lattice1 in a multisig configuration as a redundancy layer against custodian counterparty risk. That is the architecture for holding the asset. It is not the architecture for replacing the bank.
And if you are advising a board on this question, the single most useful intervention is to separate the two decisions out loud, on slide one, before the meeting drifts. Decision one is "what assets do we hold." Decision two is "where does our credit come from." Conflating them is how you end up with the wrong custody choice, the wrong banking relationship, and a six-month lag before the consequence shows up in a treasury report. Keep them apart, decide each on its own merits, and the substitution thesis tends to retire itself.
This piece does not address the regulatory live wire around stablecoin issuance under the proposed federal frameworks — that question is moving too fast for an analytical post to track usefully. It does not address tokenized-deposit pilots being run inside specific bank holding companies, which are private and which I do not have receipts on. And it does not address the question of whether the bank-credit model itself deserves to survive the decade unchanged, which is a different argument from the one I just made and worth its own piece.
FAQ
Can a fintech really run treasury operations with only bitcoin and Treasuries in custody?
For inventory and balance-sheet exposure, yes. For credit issuance, working-capital lines, payroll rails, vendor payments at scale, and any function that requires a counterparty to extend credit against your operations, no. The custody-only setup gives you holdings. It does not give you the liability-side capacity of a depository institution, which is what most operational credit needs upstream. A bank relationship is still required somewhere in the stack.
What does a qualified custodian like Anchorage Digital actually do for a corporate treasurer?
Anchorage operates under an OCC Federal Trust Charter, the first granted to a crypto-native firm, which authorizes it to hold digital assets in a fiduciary capacity. For a treasurer, that means segregated custody with bank-grade regulatory supervision, institutional withdrawal controls, and reporting integration. What it does not provide is deposit-taking or credit issuance — the trust charter is intentionally scoped to custody. Comparable trust charters exist at Coinbase Custody and Fidelity Digital Assets under New York DFS.
Why doesn't a Treasury ladder give the same liquidity as a demand deposit?
A demand deposit is the bank's own liability, redeemable on demand at par, and is fungible across the payment system. A Treasury ladder held in a custody account is a beneficial interest in securities held through DTC, settled via Fedwire Securities Service, with conversion to spendable cash requiring sale or repo. The economic distance is small. The operational distance — and the legal-finality distance — is meaningful for anything that needs same-second settlement.
Is DeFi lending a viable credit substitute for an institutional treasury?
No, for two reasons. DeFi lending is structurally overcollateralized, so it functions as inventory finance rather than credit against future cash flows — a category of credit that banks issue routinely and that on-chain protocols cannot enforce. Second, the institutional treasurer using DeFi lending is the borrower, not the lender of record. They are not creating credit; they are consuming credit from a different counterparty.
Does holding bitcoin through Ledger or Trezor change the credit-replication argument at all?
No. Hardware wallets from Ledger, Trezor, or GridPlus solve the custody-security problem at the individual or small-team level. They do not change the regulatory or balance-sheet status of the asset being held. Self-custody is an answer to "who holds the keys." It is not an answer to "who issues credit against the position," which is the question this article is about.
What is the bitcoin market cap and price context the original treasurer conversation referenced?
At the time of writing, bitcoin trades at $83,000 with a circulating supply near 19.8 million coins and a market cap of roughly $1.65 trillion. The all-time high of $109,000 was set on January 20, 2025. The numbers matter because they are large enough that institutional balance-sheet allocations are no longer thought experiments, which is exactly why the credit-substitution thesis is now being argued in board rooms rather than on podcasts.
When does self-custody actually compete with a qualified custodian for institutional use?
Almost never as a sole solution. The institutional standard is qualified custodian primary plus self-custody multisig secondary as a redundancy and counterparty-risk hedge. The trust-company setup gives you regulated reporting and institutional controls. The multisig setup gives you sovereignty over a portion of the holdings if the custodian relationship fails. Treating self-custody as a full replacement for a qualified custodian in an institutional context underestimates the operational burden of doing it correctly.