How Spark Is Rebuilding DeFi Lending | Sam MacPherson

Spark protocol operates as a prime agent sub-DAO within the Sky ecosystem, formerly MakerDAO, with Phoenix Labs as its main contributor. The rebrand to Sky was partly motivated by wanting USDS as a ticker because it immediately signals stablecoin to new users. The governance restructuring replaced token plurality voting, which Sam MacPherson says produced politicking and lobbying rather than sound lending operations, with a sub-DAO model where Sky acts as an unopinionated wholesale credit and stablecoin issuer and sub-DAOs pursue growth independently within loosely defined risk rules. Sub-DAOs must post junior first-loss capital, so a 10 percent requirement means a sub-DAO needs 10 million of its own capital to access 100 million in exposure.

The Spark Liquidity Layer allocates approximately one billion dollars in stablecoins, with Spark Lend as its primary destination restricted to blue-chip collateral including highly liquid bitcoin variants, ETH, Lido staked ETH, and EtherFi ETH. Emode, enabling 92 to 93 percent LTV and leverage of roughly 10 to 15 times, is restricted exclusively to Lido stETH, a decision rooted in recognizing growing ETH staking yield demand as early as 2021 to 2022. MacPherson frames Spark Lend's ETH product as the safest possible construction for ETH yield, targeting users who want liquid, low-risk yield generation. Spark also allocates through the Liquidity Layer to Morpho and has at times allocated to the Aave core market, though Aave core was generally avoided because its risk-adjusted yield was not viewed as attractive enough. Spark's yield evaluation takes absolute yield and discounts it by expected losses over time, and it uses Morpho's isolated markets infrastructure heavily because positions can be unwound easily when yield drops below that threshold.

Kelp's rsETH was onboarded into Spark Lend but never in emode and was off-boarded in January 2025 as part of a general policy of reducing risk surface. MacPherson says that even if rsETH had not been off-boarded before the Kelp incident, damage to the protocol would have been minimal or zero due to supply caps and the absence of emode for that asset. Spark Lend ETH deposits doubled in the month following the Kelp incident, which MacPherson attributes to users seeking lower-risk ETH yield. Before the incident MacPherson viewed Aave as Spark's primary competitor on high-liquidity USDC and USDT products, but following it, liquidity on Aave dropped to approximately 400 million total, and MacPherson now identifies fintechs with large balance sheets as Spark's primary competitive layer.

Spark's Ethena exposure peaked at approximately 1.2 billion dollars, which MacPherson says was below 20 percent of the balance sheet, reached only briefly around the time of the ENA token announcement when yields were double-digit percent even in senior Morpho positions. Ethena was never onboarded into Spark Lend directly and was kept exclusively in Morpho isolated markets because pooled lending markets cannot differentiate rates by borrower risk profile. MacPherson explains that a borrower using ETH as collateral to borrow stablecoins is materially lower risk than a borrower using sUSDe, but a pooled market charges both the same rate, which is why higher-risk looping products belong in isolated markets. Spark wound down Ethena exposure when risk-adjusted return fell into negative alpha territory.

Spark's risk framework is built on four pillars: the collateral asset, the custodian or admin authority, the entity facilitating the lending, and the borrower. MacPherson identifies a fundamental difference between DeFi and TradFi credit in that DeFi borrowers can be anonymous and can walk away from insolvent positions with no legal obligation to repay. Spark recently went live with Credora risk ratings across six Spark Savings vaults, and S&P issued a credit rating for USDS approximately one year before the recording, receiving a decent but not top-tier rating and identifying specific concerns that give Sky a direction for improvement. MacPherson says competitive pressure in DeFi lending creates a structural tendency for risk curators to keep increasing risk while avoiding noticeable losses, with retail LPs typically relying on brand trust rather than performing their own due diligence, and he says he was worried this dynamic could create a structural risk capable of killing the DeFi industry as a whole.

Spark Savings USDT launched at the end of 2024 and peaked at approximately 1.5 billion dollars about one week before the recording, sitting at 1.2 billion at time of recording with 750 million in liquidity representing approximately a 70 percent liquidity ratio. MacPherson describes this as a significant correction to a strategic fault, noting that Sky had been oriented almost exclusively around USDC prior to the USDT initiative. Spark Liquidity Intents allow users to submit asynchronous withdrawal requests rather than requiring synchronous atomic liquidity, with the automated allocation system fulfilling requests typically within the next block. The previous synchronous liquidity cap was 10 million dollars, which MacPherson describes as a poor user experience for depositors with positions such as 600 million dollars.

Spark Prime is described as a crypto prime brokerage enabling collateralization across exchanges, qualified custodians, and DeFi protocols simultaneously using overcollateralized loans rather than unsecured credit. MacPherson estimates the total perpetual futures borrowing market at roughly 22 billion dollars, derived from open interest peaking near 220 billion dollars in late 2024 and applying approximately 10 percent of open interest as the relevant borrowing demand. As of the recording the prime brokerage was in a scaling beta phase.