Depeg risk

The peg is a promise about conversion, not about price

No mechanism sets the secondary market price of a token. What a peg actually provides is a conversion right: somebody, under stated conditions, can hand over the token and receive the referenced asset, or the reverse. Traders enforce the price because that right is profitable to exercise whenever the market drifts away from par.

So the question that predicts a depeg is never only whether the reserve exists. It is who can convert, at what size, on which days, at what cost, and how fast. Narrow the answer to any of those and the price band widens, whether or not a single dollar of backing has gone missing.

Four channels, four different fixes

The reserve channel is the one people model: assets lose value or become unrecoverable, and backing falls below liabilities. The fix is composition and haircut analysis. The access channel is different: the reserve is fine, but the redemption desk is closed, the minimum ticket is $100,000, or the banking rails do not settle on a Saturday. The fix there is operational, not financial.

The third channel is arbitrage cost. If closing a two cent gap requires posting collateral, waiting a day for settlement and paying a fee, the band cannot be tighter than those frictions, and in stress they all widen at once. The fourth is reflexivity, where the mechanism that restores the peg is priced in the same asset under attack. That one has no fix in the design once it starts, only a decision not to build it.

March 2023: solvent at 87 cents

Circle disclosed on 11 March 2023 that $3.3 billion of the USDC reserve was held at Silicon Valley Bank, about 8% of the total reserve, and that the reserve at that point was 77% short dated US Treasury bills worth $32.4 billion with the remaining 23%, $9.7 billion, held in cash.¹ CNBC reported the same weekend, citing CoinDesk market data, that USDC fell below 87 cents.²

Read those two figures together, because they do not fit. An 8% exposure of uncertain recovery produced a discount of more than 13%, and it lasted until US banks reopened and Circle confirmed the deposits were fully available.¹ Almost none of that was a solvency judgement. It was the redemption door being shut on a bank holiday weekend while a real but bounded loss was unquantified. Anyone modelling this event as a reserve failure will size their liquid tranche and miss the actual mechanism.

May 2022: when the restoring force is the asset under pressure

The SEC's complaint against Terraform Labs records the sequence and it is worth reading in order. In May 2021 UST dropped below $1.00, and the defendants secretly arranged for a third party to purchase large amounts of UST to restore the peg. One year later, in May 2022, under selling pressure from large UST holders, UST depegged again, and this time without intervention the price of UST and LUNA fell to nearly zero, wiping out over $40 billion of total market value.³

The 2021 event is the part that gets skipped, and it is the diagnostic one. A mechanism that required an undisclosed buyer to hold par once was already broken; the second event only revealed it publicly. The demand that made the position so large came from a protocol advertising 19 to 20% interest on deposits.³ In our view a yield promise at the token layer is not a marketing choice, it is a liability, and any design where the peg depends on that yield continuing should be treated as a leveraged position rather than a stablecoin.

What both cases change about the design work

Separate solvency from liquidity and test them on different clocks. Solvency asks whether reserves cover liabilities at some future settlement date. Liquidity asks what can be converted at par this afternoon. USDC was solvent throughout and still traded at a discount, and that gap is the design surface.

Then publish enough for the market to price the difference. Circle maintains a public transparency page carrying reserve composition and third party reporting.⁴ The useful question for any issuer is not whether disclosure exists but whether its cadence matches the redemption cadence: monthly reporting against daily redemption is a thirty day window in which holders are pricing a rumour rather than a number.

Finally, write down what happens when the primary channel is unavailable. Which venue absorbs redemption pressure over a weekend, at what size, and who is authorised to say so publicly. Structures that improvise this under stress usually do it badly.

How this differs from a redemption run

A depeg is a price event. A redemption run is a volume event at the primary channel, and the two travel together often enough that they get treated as one thing. They are not. A run can happen with the peg intact, and a peg can break with almost nobody redeeming, which is exactly what March 2023 was.

The practical difference is in what you monitor. Depeg risk is watched on secondary market spreads, depth on both sides, and the cost of the arbitrage round trip. Run risk is watched on redemption queue size against liquid reserves. Mechanism analysis of either is not a comment on whether any particular token is safe to hold, which is a question for the holder and their own advisers.