Stream Finance's Recursive Stablecoin Loop Unravels Into $285 Million DeFi Contagion

A month after an earlier investigation described the circular relationship between Stream Finance and Elixir as an expensive piece of financial theater, the arrangement finally gave way. What collapsed was not a single hack but a self-referential lending structure: Stream issued its synthetic dollar, xUSD, while Elixir issued the collateral token, deUSD, that partly backed it — the same underlying capital allegedly cycling through both systems until roughly $2 million in real assets supported something closer to $14 million in synthetic tokens, wrapped in language borrowed from institutional finance ("market-neutral strategies," "delta-neutral positions," "institutional-grade returns").

The first public crack appeared on October 28th, when researcher Schlagonia laid out the recursive minting loop in detail. The same day, analyst CBB estimated the system was running roughly 4.1x leverage, with about $170 million in real backing supporting $530 million in liabilities. Six days later, on November 3rd, Stream confirmed a $93 million loss; xUSD fell 77%, from roughly $1.00 to $0.26, and an estimated $285 million in exposure rippled through DeFi lending markets.

Among those caught in the fallout was Hyperithm, which reported roughly $75 million tied up in Stream-linked mHYPER positions, plus $2.6 million stuck in an xUSD lending vault on Euler Plasma that it said could not be withdrawn because of xUSD's illiquid market. A yield strategy had turned into an exercise in damage control.

01Who built the loop

Stream Finance was not a fly-by-night operation. It was founded by Diogenes Casares under Klyra Labs, based in Buenos Aires. Casares had a track record as an activist DAO investor who had previously pushed underperforming startups to return capital to their backers — an ironic résumé given what followed. Stream marketed itself in the vocabulary of professional trading desks: "market-neutral fund strategies," "stat arbitrage," "hedged top of book market making."

Elixir was the other half of the equation. It pitched deUSD as institutional-grade infrastructure for DeFi, described as backed not by cash but by stETH deployed in delta-neutral strategies alongside T-bill exposure routed through MakerDAO's USDS. Its documentation framed deUSD as a fully collateralized, yield-generating dollar meant to let institutional capital access DeFi yield without giving up its original asset exposure.

Stream advertised 18% APY on USDC vaults and 12% on ETH, attributing the returns to lending arbitrage, hedged market-making, and incentive farming. What later reporting established was that deUSD's backing was not maintained as a strict 1:1 dollar reserve, and some analysts have argued Elixir was not relying on centralized real-world-asset reserves at all. On-chain records also show that Stream's own StreamVault contract retained special permission to mint new tokens as yield rounds settled — meaning Stream itself controlled part of the supply it was supposedly earning yield on. Both projects promised proof-of-reserves reporting that, per Stream's transparency page, stayed permanently "coming soon."

02How the loop actually worked

According to Schlagonia's research, the mechanism reduced to four repeating steps: USDC was converted into deUSD, the deUSD was used as leverage, that leverage minted xUSD, and the new xUSD was used to borrow more USDC. Concretely, Stream took user deposits, converted them to deUSD via Elixir's minter, bridged the deUSD to Avalanche or World Chain, borrowed stablecoins against it, swapped the proceeds back to USDC, and used that capital to mint additional xUSD. Elixir then closed the loop by lending further capital against the very xUSD Stream had just minted.

Schlagonia's analysis identified roughly $70 million in USDC supplied to Morpho markets that were not publicly visible, with about $65 million borrowed against it and Elixir functioning as the sole depositor. The cycle — borrow USDC, mint xUSD, borrow against xUSD, mint more deUSD — effectively moved the same pool of USDC around while marketing it as two distinct, separately dollar-pegged assets. By October 27th, three full rounds of this loop had reportedly been completed, and by Schlagonia's estimate the process had inflated roughly $1.9 million of original capital into about $14.5 million in synthetic tokens.

03The warnings that preceded the collapse

The clearest public alarm came on October 28th from CBB, who urged users to pull funds from any Morpho or Euler vault exposed to mHYPER or xUSD, calling the setup "max opacity finance" with an "insane" level of embedded leverage. Analysts subsequently described both xUSD and deUSD as structurally fragile by design, pointing to stacked leverage, recursive loops dressed up as strategy, and the absence of any proof of reserves.

Earlier signals had already surfaced. In May 2025, a Chainlink oracle on Avalanche briefly misreported deUSD's price at $1.03, triggering more than $500,000 in liquidations on Euler Finance for accounts holding deUSD debt. In August 2025, Inverse Finance's governance voted to wind down its deUSD markets, citing risk and operational concerns — a protocol quietly declining to accept the collateral is its own kind of warning.

Stream's transparency page had for some time carried the line: "Coming soon! We are currently integrating with third parties to offer proof of reserves to improve our transparency" — a promise that analysts noted stayed unfulfilled indefinitely even as the protocol advertised large backing and high yields. DefiLlama's tracked TVL numbers also diverged from Stream's own figures. On October 30th, Stream stated that total assets deployed across its strategies were roughly $520 million, separate from about $160 million in direct user deposits — a figure that became the reference point for its later "proof of assets" claims. Community estimates at the time put real backing closer to $170 million against roughly $530 million in liabilities, implying about 4x leverage. By October 31st, the accumulating red flags prompted the original piece asking whether Stream's structure amounted to a house of cards.

04The break

On November 3rd, Stream confirmed the $93 million loss. By then, the trajectory had been visible for days.

Prior to that, the interdependence of the ecosystem had already been documented: YieldFi's yUSD borrowed from Morpho and served as collateral backing mHYPER, which in turn lent into xUSD — a chain where each protocol's solvency depended on the next. When the chain first strained, roughly 24% of yUSD's market cap disappeared within 24 hours.

Hyperithm had moved early. On October 28th it publicly confirmed exiting all yUSD and xUSD exposure, retaining $10 million unleveraged as a general-partner commitment and publishing wallet addresses so the move could be independently verified — a repositioning based on data rather than a panicked retreat. Because Morpho is permissionless, much of this looped exposure had been sitting in plain sight the whole time; retail depositors several steps removed from the core loop had no visibility into the fact that their "safe" yield was ultimately backed by collateral worth a fraction of face value. With only about $170 million of real backing against $530 million of borrowing — a leverage ratio above 4x — the buffer was thin to begin with; BlockEden's later estimate put the effective collateral even lower, under ten cents on the dollar.

05How Stream and its defenders explained the mechanism

On October 30th, Stream published its own account of the recursive looping strategy: "Recursive looping is when a protocol loops its own asset to capture a spread in interest rates. This increased yield then goes back to users either in the form of yield or our insurance fund. This allows users to earn more while taking on more risk." What that explanation omitted, as the earlier investigation had noted, was that the "asset" being looped was one Stream itself controlled the supply of — the protocol was simultaneously minting and borrowing its own token, meaning the leverage was not applied to independent stable assets but to already-leveraged synthetic ones.

The insurance fund referenced in that statement was opaque and not segregated from other capital, functioning in practice as retained surplus profit; observers estimated the team held onto roughly 60% of deposits in this way, a detail that only became public after outside scrutiny forced an explanation.

DCF God, who had disclosed a personal stake in Klyra, Stream's parent company, defended the underlying logic of looping — arguing that if a token pays a given APR and it's possible to loop at half that rate, doing so is rational. He also added a caveat that he had never advised anyone to deposit in xUSD, noting his own looping had stopped around 5x leverage because going further felt like too much hassle. Of Stream running the same approach at scale with client funds, he said it was "their business" and "the most degen farm," summarizing his view as: "If you're gonna be degen, might as well be full degen."

Stream's founder, using the handle 0xlawlol, posted on October 29th — in a message later deleted — that the protocol held an insurance fund exceeding $10 million and that positions outside the main wallet were "fully liquid immediately," promising a forthcoming transparency report. No such report materialized before the founder stopped posting on the platform altogether.

06The collapse timeline

On November 3rd, Stream disclosed the $93 million loss, attributing it to an unnamed "external fund manager," suspending all deposits and withdrawals, and announcing that law firm Perkins Coie had been engaged to investigate. Comments on the announcement post were disabled. xUSD's price collapsed within hours — from roughly $1 to $0.50 according to PeckShield, and further down to $0.26 (a 77% decline), with some data showing intraday lows near $0.10. Panic selling into USDC across decentralized exchanges drained remaining liquidity.

The same day, Balancer suffered a separate $128 million exploit, and some analysis has suggested the resulting market-wide alarm compounded the run on Stream, as users pulled funds simultaneously from both situations. Whether coincidental or reinforcing, both events landed on November 3rd.

On November 4th, researcher Yields and More (YAM) published a creditor breakdown putting total direct debt at approximately $285 million, with major exposures at TelosC ($123.64 million), Elixir ($68 million), MEV Capital ($25.42 million), Varlamore ($19.17 million), Re7 Labs ($14.26 million), Enclabs ($2.56 million), Mithras ($2.3 million), TiD ($0.38 million), and Trevee ($14.7 million, detailed in a post-mortem three days later and largely recovered by November 14th). YAM cautioned that the list was likely incomplete, with more affected vaults yet to surface. Multiple protocols froze markets tied to xUSD, deUSD, and related assets. Euler Finance published an FAQ clarifying that losses fell on individual curator-managed vaults rather than Euler DAO's own markets, which it said had zero exposure — noting that curators, not Euler itself, had chosen both the oracles and risk parameters (xUSD priced via eOracle, USDX and sUSDX via RedStone). Liquidations executed as coded, but once collateral value collapsed while oracles continued reporting stale prices, no one wanted to take on the resulting bad debt. Euler subsequently removed the affected vaults from its interface to prevent new deposits into the exposed positions.

On November 6th, Elixir shut down deUSD entirely, stating that Stream held about 90% of deUSD's supply (roughly $75 million) and had "decided not to repay or close positions." Elixir said it processed redemptions for 80% of holders as deUSD itself fell more than 97% within 24 hours.

On November 7th, Trevee released its own post-mortem, disclosing $14.7 million in exposure spread across several products. Unlike many other affected parties, Trevee moved quickly — partnering with Telos to freeze relevant Euler markets and ring-fence Stream's collateral — and within a week announced full recovery for splUSD holders through assertive creditor action. Notably, Trevee's post-mortem included the observation that "Stream's losses originate either from the 10/10 drawdown or later," language suggesting the exact starting point of Stream's insolvency was not precisely known. If "10/10" refers to October 10th, Stream would have operated for close to a month after first going underwater. That date corresponds to the largest liquidation event in crypto market history — roughly $19 billion wiped out in 24 hours following President Trump's announcement of 100% tariffs on China. While much of the market deleveraged in response, Stream reportedly moved the opposite direction: increasing leverage, borrowing across additional chains, and continuing to accept new deposits.

07The aftermath

StableWatch data indicated that yield-bearing stablecoins experienced their largest collective outflow since Terra's UST collapse, totaling roughly $1 billion, of which about $411 million came from xUSD alone. DeFi curator-managed vault TVL fell from about $10.3 billion to $7.5 billion in the aftermath — the largest capital-loss event in the sector since Terra/UST in May 2022.

Perkins Coie's investigation remains ongoing, and the "external fund manager" Stream blamed for the loss has still not been identified publicly. Analysts estimate eventual recovery in the range of 10 to 30 cents on the dollar; against roughly $285 million in competing creditor claims and an estimated $79 million in remaining assets, the math suggests most depositors will recover only a small fraction of what they put in.

Taken as a whole, the episode did not involve broken code — the smart contracts underlying both Stream and Elixir reportedly executed exactly as written. The failure was structural: an estimated 60% of protocol supply effectively controlled by the operators themselves, a recursive minting process that turned roughly $1.9 million into $14 million of synthetic tokens, and a proof-of-reserves promise that never arrived. Where Terra's 2022 collapse played out publicly with algorithmic promises, Stream's unwound quietly behind institutional-sounding language. Hyperithm, having examined the on-chain loop structure directly, withdrew about $10 million ahead of the collapse; most other depositors had no comparable way of knowing they needed to.

Recovery has been uneven — Trevee's aggressive legal and operational response returned its users to whole, while others remain in limbo as Stream's legal process continues, and further undisclosed exposure may still surface as more protocols review their own books.

The episode reinforces a recurring argument in DeFi risk circles: that the space still lacks independent analysts capable of properly evaluating recursive leverage and circular collateral arrangements, as opposed to curators who are financially incentivized to approve the yields they curate. Two protocols that spoke fluently in the language of institutional finance ultimately relied not on innovation but on the absence of scrutiny — until sophisticated capital had already exited and everyone else was left holding a stablecoin that was neither stable nor, in any meaningful sense, backed.