USDC and the SVB weekend, 48 hours hour by hour

The short version, before the detail

Silicon Valley Bank failed on Friday, March 10, 2023. Circle held 3.3 billion of USDC reserves at SVB, out of about 40 billion in total reserves. The market priced in the worst-case overnight: USDC fell to 0.8774 on Coinbase on Saturday morning, and below 0.82 in specific Curve pools. Circle paused minting and redemption Saturday because US banking rails were closed. The Treasury, Federal Reserve and FDIC announced Sunday evening that all SVB depositors — including the uninsured portion — would be made whole. USDC returned to peg by Monday's banking open.

That is the synopsis every coverage of the event leaves you with. The detail matters because the resolution was discretionary, the timing was fragile, and several smaller decisions that weekend (Coinbase pausing USDC-USD conversion, Curve pool composition shifting, Tether briefly trading above peg) are still teaching examples. The reconstruction below uses Pacific Time where Coinbase ticker history is the source and Eastern Time where the New York close matters.

Background · why SVB held Circle reserves

Silicon Valley Bank was the de-facto bank for the US venture-capital ecosystem. Most VC-backed companies banked at SVB; many had deposits well above the FDIC's 250,000 insurance limit. Circle, as a venture-backed US fintech, banked some operational accounts at SVB. Some of those operational accounts also held a portion of USDC reserves alongside Circle's other custodians (BNY Mellon for the bulk of cash holdings, plus T-bill positions held directly).

Reserve composition disclosure in March 2023 was monthly but not at CUSIP level. Holders knew Circle held cash deposits at "a number of US banks" but did not know the exact SVB exposure until Circle disclosed it on the Friday evening. That information gap is part of why the price reaction was as sharp as it was.

The timeline

What happened to USDT during the same weekend

Worth a section of its own. As USDC fell, USDT briefly traded above peg on global venues as a flight asset. The Binance USDT-USDC spot pair priced USDC at roughly 0.90 against USDT at the Saturday morning trough — implying a USDT premium against a notional USD of around 1.005 to 1.010. The premium was small in absolute terms but consistent across Binance, OKX, Bybit and several smaller venues.

The mechanism is straightforward: holders on global venues, looking at USDC at 0.92 and wanting to stay in a dollar token, rotated into USDT. Arbitrage between venues plus on-chain routes between Tron USDT, Ethereum USDT and centralized-exchange USDT kept the premium from blowing out further. The episode is one of the better-documented examples of the two tokens behaving asymmetrically under stress, which is the standard argument for holding both.

The arbitrage windows that opened

Several arbitrage windows opened during the weekend. Two examples we tracked at the time:

• Cross-venue USDC. The Coinbase USDC-USD price was the most liquid quote for USDC against actual dollars. The on-chain USDC price on Curve and Uniswap was the most liquid quote for USDC against other stablecoins. The two quotes drifted apart during the depeg because the Coinbase venue paused USD conversion. Sophisticated traders bought USDC on Coinbase at the trough, withdrew on-chain, and sold against USDT on Binance for a small premium. The window was tight (the Coinbase withdrawal queue lengthened during the weekend) but executable.
• DAI contagion. Because Maker's DAI was heavily backed by USDC at the time, DAI traded down to 0.93. Traders who believed Maker would either close the USDC peg arbitrage module or that USDC would recover bought DAI at 0.93 and held to peg recovery. The trade was profitable; some pension-style funds bought sizeable positions over the weekend.

We mention these not as recommendations — neither was risk-free, both required exposure during a real crisis — but as documentation of what was actually possible. Crypto-Twitter that weekend was full of "I bought USDC at 0.88" claims. The plausible ones were people who already had dry powder on Coinbase and who could withdraw US dollars after the weekend.

What the desk learned

Lesson 1 · Reserve disclosure transparency matters in real time

Circle disclosed the 3.3 billion SVB exposure on Friday evening. That disclosure was the catalyst for the deeper Saturday morning trough but also the reason the market could price the worst case. If the disclosure had been delayed (until Monday, say), the depeg might have been shallower in the first 12 hours but more severe over the weekend as uncertainty compounded. The desk's read is that the Friday-evening disclosure, while painful, was the correct call.

Lesson 2 · The resolution depended on a political choice

The BTFP and the depositor backstop were discretionary. The Treasury, Fed and FDIC chose to invoke the systemic-risk exception. That choice was made in roughly 36 hours, under pressure, with limited precedent. A different bank failure, with a different depositor profile and different political context, could produce a different choice. USDC's recovery is not a structural feature of its design; it was an outcome of a specific policy decision.

Lesson 3 · Operational pauses read as solvency signals

Coinbase's USDC-USD pause was procedural. Circle's redemption pause was procedural. Both were technically necessary because US banking rails were closed. Both were read by the market as solvency signals and contributed to the deeper Saturday trough. The lesson for issuers: communications around operational pauses need to be more aggressive about the technical reason. The lesson for holders: a procedural pause and a solvency pause look identical from the outside; you cannot tell them apart in real time.

Lesson 4 · Stablecoin diversification is not a slogan

Holders who held only USDC took a paper loss of 12% intraday and a realised loss if they sold. Holders who held only USDT had a small paper gain. Holders who held both rebalanced and ended the weekend close to flat. The diversification argument is not theoretical; it played out in real positions over 48 hours. The desk's two-token policy dates from this weekend.

What changed afterwards

Circle's reserve composition shifted. The cash-deposit share dropped substantially in favour of the BlackRock-managed government money-market fund. Circle moved primary banking to BNY Mellon as custodian for the cash portion. The monthly attestation began publishing CUSIP-level holdings of the BlackRock fund.

The IPO in June 2024 added quarterly SEC reporting on top of the monthly attestations. Circle's customer cohorts, deposit concentrations and counterparty exposures are now public in a way that few stablecoin issuers can match.

The BTFP itself ran for one year as an emergency facility and was retired in March 2024. The precedent it set — that the Treasury / Fed / FDIC will extend uninsured-deposit guarantees during a systemic stress — was not codified into law and remains discretionary for any future event.

Could a USDC depeg this size happen again?

The specific cause is harder to repeat. The BlackRock fund holds Treasury bills and overnight repurchase agreements with the Federal Reserve — instruments that do not "fail" in any conventional sense. The cash-deposit residue is split across more institutions. The disclosure is faster and more granular.

Different causes are not zero. A failure of a custodian bank holding the cash residue, a disruption to the overnight repo market, a US regulator action against Circle directly, or a software vulnerability in the contract layer would each produce a different depeg vector. The desk's working assumption is that the next USDC stress event will look nothing like the SVB weekend, and that holders who pattern-match too literally on the 2023 reconstruction may miss the actual signal when it comes.

Three reflections that aged well

The "stablecoin = T-bill ETF" framing is too clean

In normal weather, USDC behaves like a tokenised T-bill fund: stable, liquid, low-yield-equivalent. In bad weather, it behaves like a deposit account at the bank that custodies it. Both framings are correct in their own context. Holders who treat USDC purely as a T-bill ETF underweight the bank-deposit risk; holders who treat it purely as a bank deposit underweight the actual reserve composition.

The "fully reserved" disclosure language matters less than the reserve mix

Circle's repeated statements that USDC was fully reserved were factually correct throughout the weekend. The problem was not whether reserves existed; the problem was that 8% of reserves were trapped in a failed bank with uncertain timing on recovery. "Fully reserved" became a meme during the depeg because the market intuited that the reserve mix mattered more than the headline number. The 2026 monthly attestation reads this lesson into the disclosure structure.

The peg recovery looked V-shaped because the policy response was fast

The 48-hour recovery is often cited as evidence that USDC's design is resilient. The desk reads it as evidence that the US policy response was fast. The two are not the same. A slower policy response would have produced a longer, shallower, more damaging recovery curve. The peg's resilience is partly structural and partly path-dependent.

If you want to act on this

The most direct action on the SVB lesson is the diversification one: do not concentrate stablecoin holdings in a single token. The second action is venue diversification: do not concentrate working balances on a single exchange. For global users, the desk's working venue is Binance. The referral link opens a Binance registration page pre-filled with code BN16188; registering through that link does not change your fees. Full disclosure on the disclaimer page.

Where we got the SVB hourly data

• Circle, "An Update on USDC" (March 10, 2023 evening); subsequent updates March 11-13 on circle.com/blog.
• Coinbase historical ticker data, USDC-USD, 2023-03-09 through 2023-03-14.
• Curve.fi 3pool subgraph data for the same period (publicly queryable on TheGraph).
• FDIC press releases, March 10-12 2023, "Joint Statement by Treasury, Federal Reserve, and FDIC".
• Maker Protocol forum posts and emergency governance polls, March 11-13 2023.
• Circle Internet Group S-1 (June 2024), filings on SEC EDGAR — useful background on post-SVB custody arrangements.

Anything you can verify yourself, you should. The reconstruction above uses our notes from the weekend plus the public sources listed. If you spot an error, the corrections log is on the corrections page — write to [email protected] with the line.