Let me concede something upfront. The big two stablecoin issuers publish more reserve disclosure today than any non-bank issuer of dollar-denominated liabilities ever has in financial history. That is true. Monthly attestations, breakdowns by instrument, named auditors. The transparency floor has moved.
Now let me tell you what I found when I tried to follow the chain past the disclosure. I am not an ex-Circle treasury analyst. I am not on the trading desk at Cantor. I am a developer who got curious enough to spend a weekend reading attestation PDFs, NY DFS guidance, and the OCC charter for Anchorage Digital — and what surfaced is that almost everything Crypto Twitter says about stablecoin reserves collapses against the public record once you stop quoting other people quoting other people. The six beliefs below are the ones that broke first.
Myth: "Stablecoins are backed 1:1 by dollars in a bank account."
This is the load-bearing myth. The one that gets repeated in podcast intros, exchange marketing copy, and the explainer paragraph of every "what is a stablecoin" Medium piece. It is wrong in a specific and important way.
People believe it because the price is one dollar and the mental shortcut is "one dollar in, one dollar parked, one dollar out." The shortcut is comforting. It is also incompatible with how a regulated reserve actually gets composed.
The reality is that a stablecoin's reserve is a treasury portfolio. The dollar peg is the output. The input is a mix of short-duration T-bills, overnight reverse repo positions, cash held at commercial banks, and money market fund shares. The proportion shifts month to month based on yield curve and outflow forecasts. USDC has a circulating supply of 58 billion units pegged at one dollar — every one of those units is a claim against a reserve pool that is overwhelmingly not idle dollars. Idle dollars do not pay yield. The whole business model of a stablecoin issuer is "we hold yield-bearing instruments and pay you nothing." If the reserve were actually a bank account, there would be no Circle.
Practical implication: when you hold a stablecoin you are holding a tokenized claim on a treasury fund operated by a non-bank counterparty. Pricing that risk at zero is a category error.
Myth: "If Circle or Tether holds the reserves, the custody risk is theirs."
The intuition here is corporate-veil thinking. Circle holds the reserves, so if something blows up, it blows up at Circle. You, the holder, only carry the price risk on the token.
This is wrong because the issuer does not custody its own T-bills. Nobody custodies their own T-bills. T-bills settle through the Federal Reserve's book-entry system and are held by qualified custodians — for stablecoin reserves that typically means a primary dealer, a custodial bank, or a regulated trust. The question "where are the reserves?" has at least three layers: the issuer, the asset manager running the reserve fund, and the custodian holding the underlying instruments.
I tried to map the full chain for the public record. The reserve fund manager and named bank counterparties show up in attestation reports and SEC filings of the underlying money market funds. The actual custodian of book-entry T-bills sits one layer deeper and is not always disclosed in the monthly attestation — it is in the fund's separate filings.
Practical implication: every additional layer is a counterparty. The token holder sits at the bottom of a stack with at least three credit exposures. The 2023 USDC depeg, where the token briefly traded near 87 cents, was a bank counterparty story — Silicon Valley Bank — not an issuer story. The risk was not where the marketing said it was.
Myth: "T-bills are safer than bank deposits, so reserve composition does not matter."
This one shows up from people who have read just enough macro. T-bills are the closest thing to a risk-free asset in dollar markets. So a reserve held in T-bills is "safer than cash." Therefore composition is a distraction.
The reasoning collapses on the word "safer." Safer against what risk?
T-bills carry essentially zero credit risk against the U.S. Treasury for the holder of the bill. They do carry duration risk, liquidity risk during a fire sale, and — crucially — operational risk in the redemption pipeline. If a stablecoin issuer faces a 4 billion dollar redemption wave on a Friday afternoon and its T-bills mature Monday, the dollars do not exist on the redemption side until they exist on the custody side. That gap is where pegs break.
The 2023 depeg episode demonstrated this. The instruments were sound. The pipeline was not. Money market funds have gating powers — a stablecoin issuer's promise of instant 1:1 redemption is structurally tighter than the redemption terms of the funds and bills that constitute its reserve.
Practical implication: composition matters less than redemption mechanics. Read the attestation for the share of overnight reverse repo and cash at named banks — that is the liquidity floor. T-bill share is the yield floor, not the safety floor.
Myth: "Proof of reserves means proof of solvency."
Proof of reserves is the most-cited transparency mechanism in crypto. Exchanges publish it. Stablecoin issuers publish it. The phrase suggests that if reserves are proven, the issuer is solvent. The phrase is doing work it cannot do.
A reserve attestation is a point-in-time snapshot of assets. Solvency is assets minus liabilities. If the liabilities side is not attested with the same rigor, the proof is asymmetric. An exchange can publish a Merkle tree of customer balances and a wallet address showing reserves at that block — and still owe more than it holds if some liabilities are off-tree. Every major exchange in the grounding here advertises CER-verified reserve status, with attestation dates clustered between February and March 2025 — Binance on 2025-03-01, Bybit on 2025-03-12, OKX on 2025-03-01, Bitget on 2025-02-20. Useful. Not the same thing as solvency.
For stablecoin issuers the asymmetry is sharper. The liabilities are radically transparent — every token in circulation is a public, countable on-chain liability. That is the structural advantage stablecoins have over exchanges in this argument. The reserve side is the harder one to prove, and that is where the attestation focus has rightly been.
Practical implication: when you read a stablecoin attestation, you are reading the strong side of the balance sheet. The weak side — the legal mechanics of who has a claim on the reserve in a bankruptcy waterfall — is in the issuer's terms of service and the relevant trust law, not in the attestation.
Myth: "Self-custody removes you from stablecoin issuer risk."
This is the self-custody maximalist's version of the same confusion. The argument is that holding USDC on a Ledger device or in a multisig contract removes the issuer from the equation. You hold the keys, you hold the asset.
You hold the token. The token is a database entry — an ERC-20 balance, or its equivalent on a non-Ethereum chain — backed by a contractual promise from the issuer to redeem at one dollar. Self-custody changes who controls the private key that authorizes a transfer of that database entry. It does not change the identity of the entity that has to honor the redemption.
Ledger devices, Trezor units, GridPlus Lattice1 hardware — every one of them stores keys, not value. The value lives on the chain and is, for stablecoins, an IOU from a centralized issuer that is one regulatory enforcement action or one bank counterparty failure away from being worth less than face. The 2023 episode hit self-custodied USDC just as hard as exchange-held USDC. The peg break was issuer-side. The hardware did not help.
Practical implication: self-custody hedges against custodian failure and exchange insolvency. It does not hedge against issuer failure. For stablecoin exposure, those are different risks and they need different mitigations.
Myth: "Qualified custodians make stablecoins safer for institutions."
The institutional version of the self-custody myth. The argument: an asset held at Coinbase Custody — a New York DFS Trust Company — or at Fidelity Digital Assets — also NY DFS Trust — or at Anchorage Digital — the first OCC-chartered crypto bank — is structurally safer than the same asset held at a hot wallet at an exchange.
That is partially true. Custody is segregated, audited, and bankruptcy-remote in the documented sense that a Coinbase Custody trust account is supposed to be remote from Coinbase the operating company. Anchorage Digital's federal trust charter gives it a regulatory posture closer to a bank than to a crypto firm. That layer matters. It is the reason institutions use these custodians.
But for stablecoin holdings specifically, the custodian protects you from custodian failure, not from issuer failure. If Coinbase Custody holds your USDC in a segregated trust and Circle has a reserve crisis, your bankruptcy-remote custody account contains a tokenized claim that has lost value. The custodian did its job. The asset is still impaired.
Practical implication: qualified custody and stablecoin issuer risk are stacked, not overlapping. An institution holding ten million in USDC at Anchorage carries Anchorage custody risk plus Circle issuer risk plus the bank counterparty risk in Circle's reserves. The custodian decision and the issuer decision are independent and both need diligence.
What to Actually Believe
A stablecoin is a tokenized claim on a treasury portfolio operated by a non-bank counterparty. That sentence does more work than any reserve-composition table will. Hold it in your head and the rest of the analysis follows: the relevant risks are issuer credit, bank counterparty credit inside the reserve, redemption pipeline mechanics, and the legal status of your claim in a bankruptcy waterfall. Custodial choice — self-custody on Ledger or Trezor, qualified custody at Coinbase or Anchorage — sits orthogonal to all four.
The practical move for size positions is to read the monthly attestation as a balance sheet of counterparties, not as a guarantee. Note the share at named commercial banks. Note the share in overnight repo. Note the share in money market funds and identify whether those funds are themselves gated. The names are in the disclosures. The relationships between them are the actual risk graph.
And accept that the disclosures are the strong side of the balance sheet. The weak side is the terms of service. Read those too. The redemption mechanics, the right to suspend, the priority of token holders against operating company creditors — those clauses do not make the front page of CoinDesk and they govern what happens on a bad Friday.
FAQ
Where exactly are the T-bills behind USDC custodied?
I could not pull a single named custodian for the book-entry T-bills inside USDC's reserve to the level of "this CUSIP at this institution." The monthly attestation discloses the asset manager running the reserve fund and the named bank counterparties holding cash deposits. The custodian of the underlying Treasury instruments sits one filing layer deeper — in the SEC documents of the money market funds that hold the bills, not in the stablecoin attestation itself.
Is USDC safer than USDT because Circle is U.S.-regulated?
The two issuers operate under different regulatory regimes and that matters, but framing it as "safer" collapses several distinct risks. U.S. regulation gives Circle a defined enforcement counterparty, named bank relationships, and audit standards. It also concentrates bank counterparty risk in U.S. banks — which is exactly what hit USDC in March 2023 with the SVB exposure. Different risk shapes. Not strictly ordered.
What does NY DFS Trust Company status actually give a custodian?
NY DFS Trust Company status — held by Coinbase Custody and Fidelity Digital Assets — imposes capital requirements, segregation rules, custody standards, and ongoing supervision by the New York Department of Financial Services. Functionally, customer assets are held in trust and treated as bankruptcy-remote from the operating company. It does not regulate the underlying crypto assets themselves, only how the trust holds them on behalf of clients.
How is Anchorage Digital different from Coinbase Custody?
Anchorage Digital holds a national trust bank charter from the OCC — it is supervised at the federal banking level rather than at the New York state level. That puts it in a different and arguably stronger regulatory category than NY DFS Trust Companies for institutional clients who care about federal banking supervision. The custody mechanics are similar. The supervising regulator and the legal framework around it are not.
Does holding USDC on a Ledger or Trezor protect me if Circle fails?
No. The hardware wallet secures the private key that authorizes transfers of your USDC balance. The balance itself is a claim on Circle. If Circle's reserve is impaired or the issuer becomes insolvent, the value of the token drops regardless of where the key is stored. Self-custody hedges exchange and custodian failure. It does not hedge issuer failure. Those are independent risks and need independent thinking.
What was the lesson of the March 2023 USDC depeg?
USDC briefly traded near 87 cents in March 2023 when Silicon Valley Bank failed and Circle had reserve cash deposited there. The lesson was that stablecoin pricing is sensitive to specific bank counterparty exposures inside the reserve, not just to T-bill composition. The instruments were fine. The bank holding part of the cash leg was not. Reading attestations means reading the named banks, not just the asset categories.
What is this article not covering?
This piece does not cover the specific reserve composition of USDT — Tether's reserves are disclosed under a different regime than USDC's and a like-for-like trace requires more space than this format gives. It does not cover algorithmic stablecoins, which have a fundamentally different failure mode. And it does not cover the tax treatment of stablecoin holdings in any specific jurisdiction. Each of those is its own article.