What actually happens when a stablecoin depegs

A stablecoin trading below a dollar looks like one event and is usually one of three: the reserves are impaired, the reserves are fine but unreachable, or the reserves are irrelevant because the price you are looking at came from the wrong place.

Two well-documented breaks show the second and third, and neither involved a stablecoin that was actually unbacked.

$3.3bn of Circle's reserves were held at Silicon Valley Bank when it failed. Over a weekend, with no clarity on whether those deposits would be recovered, USDC traded down to about $0.87.

The reserves were never missing. They were in a bank that had stopped functioning, and the market priced the possibility that a portion would not come back. The peg recovered when the US government guaranteed SVB deposits — a political decision, not a market one.

The lesson is uncomfortable and worth stating plainly: full backing did not prevent the break, because backing held somewhere you cannot reach on a Saturday is not the same as liquidity.

On 10 October 2025, USDe printed $0.65 on Binance while trading at $0.99 everywhere else. The protocol was 110% collateralised throughout.

The cause was mechanical: Binance's unified-account oracle read its own $8m order book rather than the $400m of on-chain liquidity next door. A price on a thin venue became the price used to liquidate positions, those liquidations sold into the same thin book, and the feedback loop did the rest. About $1bn of forced liquidations followed, and roughly $8bn of supply left over the following months.

The identical failure hit wBETH the same day: a $7bn fully-backed wrapped staking token printed about $430 against ETH trading above $3,800 — an 88% discount — because the collateral was priced off a thin internal pair. Binance paid $283m in compensation and changed how the token is priced. The staked ETH behind it was never impaired for a second.

Neither was a backing failure, and both cost people real money. The risk was not in the reserves; it was in the plumbing between the reserves and the price on your screen.

This is why comparing stablecoins on reserve quality alone misses the failures that actually occur. Backing is not liquidity, and liquidity is not the venue you happen to be looking at.

• Never hold leveraged positions collateralised by a stablecoin on a venue that prices it off its own thin book. That is the specific mechanism that liquidated people in both October 2025 events.
• Split across issuers. Every failure in this category has been issuer-specific, and both events above were confined to one token at a time.
• Do not sell into a break unless you have to. In both cases the price recovered; the permanent losses were forced liquidations, not voluntary sales.
• Check where the price you are acting on comes from. A quote on one exchange is not the market — see why the same coin costs different amounts.

The failures that do not recover are the ones where the backing was never there or was reflexive — collateralised by the issuer's own token, so that a falling price destroyed the collateral that was supposed to support it. Those are structurally different from either event above, and they end at zero rather than at $0.87.

Our stablecoin rating records what kind of reserve each token holds and what type of document evidences it, because that is the distinction between a scare and a total loss.

Backing is not liquidity, and liquidity is not the venue you happen to be looking at.

Most depegs of well-backed stablecoins are liquidity or price-feed events, not solvency events, and they recover. The money that is permanently lost is lost by leveraged holders liquidated at a price that existed on one venue. Split across issuers, avoid leverage collateralised by stablecoins on thin venues, and distinguish a reserve problem from a plumbing problem before acting.