Weekly DeFi Roundup — June 26, 2026

一、市场结构：从垄断性集中到模块化分散 截至2026年6月，DeFi借贷协议在超过380个活跃平台上持有约540亿美元存款，使借贷成为仅次于流动性质押的第二大DeFi类别- 3 - 2 。DeFi总锁仓量（Total Value Locked, TVL）从2025年末的近1780亿美元高点回落至约725亿美元，但借贷类别相对收缩幅度小于整体市场- 。借贷协议TVL约142.95亿美元，已超过DEX协议类别（114.5亿美元），在协议类别排名中上升至第二位- 。 头部协议的市场集中度极高。据DefiLlama数据，2026年5月前五大借贷协议依次为：Aave V3（194亿美元）、Spark（68亿美元）、Morpho Blue（49亿美元）、Compound V3（27亿美元）和JustLend（24亿美元），合计占市场超过360亿美元- - 2 。Aave的累计贷款发放额已超过1万亿美元，占据去中心化借贷市场62.8%的份额- 3 。这一集中度标志着自2020年以来首次有单一协议的市场使用率突破50%- 5 。 然而，市场结构正在经历深刻转型。“简单、 monolithic（单一整体）的流动性池时代正在让位于模块化架构、隔离市场和借贷-DEX混合设计”- 2 。这一转型反映了借贷市场从追求规模效应向追求风险定制化的演变——不同协议正在服务于根本不同的风险偏好。 二、核心协议机制分析 2.1 Aave V3：共享流动性模型的集大成者 Aave V3采用共享流动性模型——所有支持的资产部署在每条链的单一合约中- 2 。据DefiLlama数据，Aave V3当前TVL约为126.6亿美元，其中以太坊主网贡献约103.7亿美元，Plasma约5.79亿美元，Arbitrum约4.26亿美元，Base约3.90亿美元- 8 。Aave V3的年化费用（Annualized Fees）约为9.396亿美元- 8 。 共享流动性模型的核心优势在于深度流动性和紧密的利差，这解释了Aave在TVL排名中的主导地位- 2 。其风险缓释机制包括： 隔离模式（Isolation Mode） ：对高风险资产的借贷进行上限约束- 2 ； 高效模式（eMode） ：在高度相关资产类别中实现更高的抵押效率； Umbrella安全模型：2025年6月引入，提供针对坏账的自动化保护- 7 。 Aave的利率机制采用利用率曲线（utilization curve），在约80%利用率处设置“kink点”，之后利率陡峭上升以保障存款人的流动性- 2 。USDC供应收益率在3%至6% APY之间，Arbitrum和Base因借款人与供应商比率较高，收益率通常略高于以太坊主网- 3 。 2026年4月，Aave因Kelp DAO攻击事件遭遇重大压力测试。攻击者利用跨链桥漏洞盗取约2.93亿美元的rsETH代币，将其作为抵押品存入Aave V3并借出wETH，产生了约1.95亿美元的坏账- 7 。Aave的TVL从约264亿美元骤降至186亿美元- 7 。Aave随即冻结了rsETH市场并暂停了多个链上的WETH储备- 7 。 需特别指出的是：Aave V3的USDT和USDC借贷池在事件后达到100%利用率，超过51亿美元的稳定币被锁定，直至新流动性注入或未偿借款被偿还- 7 。这一事件暴露了共享流动性模型的系统性脆弱性——单一资产的风险可以通过池间传染扩散至整个协议。 2.2 Morpho Blue：模块化架构的挑战者 Morpho最初作为Aave和Compound之上的优化层构建，后扩展为完全可定制的借贷金库- 1 。其核心架构创新在于：用户和开发者可以创建隔离的借贷市场，各自拥有独立的抵押资产、风险参数和利率模型- 1 。Morpho当前支持超过30条链，TVL约74亿美元- 1 。其MetaMorpho金库架构通过Gauntlet Prime等精选策略，将USDC供应APY推高至4.1%至6.8%- 3 。 Morpho的模块化设计在风险隔离方面具有显著优势——一个市场的失败不会波及其他市场。但代价是：其无需许可的市场结构可能将用户暴露于比Aave等精选平台风险更高的池子- 1 。 2.3 Spark与MakerDAO：稳定币基础设施导向 MakerDAO（现重组为Sky生态系统）TVL约63亿美元- 13 。其旗下的Spark Protocol TVL突破5亿美元- 。Spark的架构围绕稳定币流动性和资本效率进行优化，深度集成USDS稳定币系统和以太坊抵押借贷市场- 1 。其优势在于深厚的流动性、严格的抵押标准和多年的运营历史- 1 。 2.4 Compound V3：稳健但落后 Compound V3 TVL约27亿美元- 2 。USDC供应收益率在3%至5%之间，通常低于Aave和顶级Morpho金库- 3 。Compound在每一次重大DeFi压力事件中保持了安全记录- 3 ，但其市场份额已被大幅挤压。 三、市场趋势：杠杆周期与收益率压缩 3.1 借贷规模的扩张与收缩 截至2026年6月，未偿DeFi贷款在2026年上半年增长了超过37%- 。总活跃贷款从2023年初的低个位数十亿美元增长至2026年第一季度的约350亿至400亿美元- 。这一增长主要由以下因素驱动：稳定币借款占未偿DeFi债务的84%，ETH抵押品占39%，流动性质押代币占28%，BTC封装代币占14%- 3 。 然而，2026年6月初的市场抛售引发了一轮剧烈的去杠杆化。Aave V3的7天滚动费用较前期下降60%至672万美元；Morpho Blue费用下降60%至327万美元；Maple Finance下降59%至125万美元- 22 - 。去中心化交易所同样承压：Uniswap V3费用下降57%至374万美元，Curve下降65%至89.1万美元- 22 。 3.2 杠杆周期的结构性解读 去杠杆的根源在于套利交易（carry trade） 的压缩。如Symbiotic联合创始人Misha Putiatin所述：“借款人支付利息是因为他们预期将这些资本部署到能产生更高回报的策略中——这就是套利交易。现在仍有大量可供借贷的资本，但可部署借入资金的有吸引力的、可扩展的机会变少了。”- 22 当可贷资本供给超过生产性需求时，借贷成本下降，贷款收益率压缩- 22 。这一机制解释了为何架构不同、借款人类别各异的协议同步下跌——当套利机会在整个资产类别中枯竭时，每个场所都以大致相同的速率吸收同样的压缩- 22 。 3.3 机构化与产品创新 未来趋势指向机构采用、证券代币化、固定与浮动利率产品，以及合规层与高风险借贷层之间的分化- 。借贷已从最初的杠杆扩张工具演变为涵盖稳定币和真实世界资产（RWA）的多元化资本市场- 。 Aave V4计划于2026年推出，采用中心辐射式架构（hub-and-spoke architecture） - 3 。Apollo Global Management与Morpho已达成协议，在48个月内覆盖高达9000万枚代币- 3 。Coinbase于2025年底通过Base网络上的Morpho协议推出借贷服务，用户可使用比特币或封装比特币作为抵押品借入高达100万美元的USDC- 3 。 四、风险分析 4.1 协议级风险 智能合约风险是所有DeFi借贷协议的基础性威胁。虽然主要协议经过多次审计并经历了多轮市场周期- 1 ，但代码漏洞始终存在。 预言机风险在共享流动性模型中尤为突出。单一资产的价格预言机失败可能危及整个池子- 2 。 坏账风险在极端市场波动中可能累积。Aave的Umbrella模型在Kelp DAO事件中接受了首次重大测试- 7 。值得注意的是，Aave于2026年4月与其长期风险服务提供商Chaos Labs分离，原因是关于Aave v4方向和预算约束的分歧- 7 。 4.2 系统性风险 传染风险是当前市场结构中最值得关注的系统性威胁。Kelp DAO事件生动展示了风险如何通过DeFi的互联基础设施迅速传播- 7 。多个与rsETH或LayerZero桥接相关的协议——包括Curve Finance、稳定币发行商Ethena和BitGo的Wrapped Bitcoin——均暂停了桥接活动- 7 。 流动性枯竭风险在利用率达到100%时尤为严重- 7 。存款人可能无法提取资金，直到新流动性注入或借款被偿还。 4.3 监管风险 在美国，SEC不断演变的框架和拟议的IRS Form 1099-DA报告要求将增加平台和用户的合规义务- 3 。在大多数司法管辖区，以加密货币为抵押借款不属于应税事件，但接收收益代币（如Aave上的aTokens）可能构成应税交换- 3 。 五、治理观察 Aave DAO于2026年4月批准了“Aave Will Win”（AWW）提案，约75%的投票支持，将Aave所有品牌产品的100%收入直接导向社区金库- 。这一决策标志着协议治理的一次重大转向，但同时也引发了对收入分配效率和长期可持续性的讨论。 MakerDAO（Sky）近期出现了一项紧急治理提案，在无预警的情况下迅速进入投票流程并获得通过，大幅提高了MKR代币的借贷上限并降低了抵押要求，引发了社区对治理透明度的质疑- 。 六、结论 DeFi借贷市场正处于从规模驱动向架构分化转型的关键阶段。Aave V3凭借共享流动性模型维持市场主导地位（TVL约126.6亿美元），但Kelp DAO事件暴露了其系统性脆弱性- 8 - 7 。Morpho的模块化架构提供了风险隔离的替代路径，但以牺牲部分流动性和增加用户尽职调查负担为代价- 1 。 市场趋势表明，借贷活动高度依赖于杠杆周期——2026年6月的去杠杆事件使Aave V3费用下降60%- 22 。未来12至18个月，Aave V4的中心辐射式架构、机构资本的持续流入以及监管框架的明晰化将共同塑造市场格局- 3 。 对市场参与者的核心建议：在共享流动性协议中追求收益时，需密切关注池中高风险资产的敞口；在模块化协议中操作时，需对每个隔离市场进行独立的尽职调查。收益率是风险的定价，而非免费的午餐。

DeFi Trends & Insights

Aave V3: Mechanism Design, Market Dominance, and the Risk Premium of Blue-Chip DeFi Abstract Aave V3, the flagship iteration of the largest decentralized lending protocol, commands a structural lead in on-chain credit markets that has widened rather than narrowed through the 2025–2026 cycle. Yet this dominance masks a complex risk landscape—one where Efficiency Mode (E-mode) leverage approaches systemic density, cross-chain bridge vulnerabilities propagate credit risk across deployments, and governance mechanisms strain to keep pace with real-time utilization shocks. This analysis examines Aave V3's core mechanisms, evaluates its market position against empirical data, and flags the principal risk vectors confronting the protocol as of mid-2026. I. Mechanism Architecture: Capital Efficiency at the Cost of Complexity 1.1 Efficiency Mode (E-mode) Aave V3, launched in March 2022 and progressively deployed across networks through 2023 and 2024, introduced a set of risk-management features that fundamentally altered how collateralized lending operates on-chain- 3 . The most consequential is Efficiency Mode (E-mode) , which allows borrowers to access substantially higher loan-to-value (LTV) ratios when their collateral and borrowed asset exhibit price correlation- . The protocol's documentation frames E-mode as a mathematically grounded optimization: if collateral and liability move together in price, the liquidation risk that normally justifies conservative LTV caps is materially lower- 3 . Under E-mode, a borrower using staked ETH derivatives as collateral to borrow ETH can access LTV ratios of up to 93%, compared to the standard 80% cap on mainnet- 3 . For certain stablecoin categories, E-mode permits LTV ratios approaching 97%- . This mechanism is not without trade-offs. E-mode restricts borrows to correlated asset pairs—such as borrowing WETH against ETH liquid staking tokens (LSTs) and liquid restaking tokens (LRTs)—thereby concentrating leverage within narrow asset classes- . As we shall see, this concentration has become a primary risk vector. 1.2 Cross-Chain Deployment Architecture Aave V3's multi-chain architecture spans Ethereum mainnet, Arbitrum, Base, Avalanche, Polygon, Optimism, BSC, Mantle, Gnosis Chain, and numerous additional Layer-2 deployments- 2 - 3 . As of the latest DefiLlama data, Ethereum mainnet accounts for $10.365 billion of the protocol's $12.659 billion total value locked (TVL), with Plasma ($579M), Arbitrum ($426M), and Base ($390M) representing the largest secondary deployments- 2 . This cross-chain architecture delivers substantial benefits—liquidity providers face minimal switching costs when optimizing for yield, and the protocol's brand recognition provides institutional participants with a credible starting point- 3 . However, it also introduces fragmentation: governance decisions—cap adjustments, collateral rating changes, freezes—must propagate across instances, creating latency under stress- 20 . Risk profiles vary meaningfully by chain: Ethereum mainnet offers deeper liquidation liquidity, while L2 deployments face thinner order books, bridge dependencies, and dispersed oracle coverage- 20 . 1.3 Umbrella Security Model Aave's "Umbrella" security model, introduced in June 2025, provides automated protection against bad debt while allowing users to earn rewards by staking aTokens (such as aUSDC or aUSDT) to support protocol security- 1 - . Umbrella replaces the legacy Safety Module with a modular onchain risk management system designed to automate bad debt coverage for Aave V3 pools- . As of June 2026, aggregate borrowed amounts across ETH, USDT, USDC, and GHO have fallen from approximately $11.8 billion to $6.8 billion—a reduction of roughly 43%, according to the TokenLogic Aave Analytics Dashboard (June 4, 2026)- . This contraction reflects both market deleveraging and the protocol's post-exploit risk recalibration. II. Market Position: Dominance by an Order of Magnitude Aave's market position is not merely dominant—it is structurally unassailable by current competitors. As of mid-April 2026, Aave commanded approximately $17 billion in TVL, dwarfing its nearest competitor by a factor of roughly three: Compound held under $3 billion and Spark Protocol approximately $4 billion, according to DefiLlama's lending category tracker- 3 . More recent data from May 2026 shows Aave V3 at $19.4 billion, with Spark at $6.8 billion, Morpho Blue at $4.9 billion, and Compound V3 at $2.7 billion- . This concentration is not accidental. Aave's TVL lead over the next largest lending protocol exceeds $13 billion—a gap that has widened rather than narrowed through the 2025–2026 market recovery- 3 . The durability of this dominance compounds: deeper liquidity pools mean tighter spreads between supply and borrow rates, which in turn attract more users, which deepens liquidity further- 3 . A 2024 Electric Capital developer report found that Aave maintained one of the highest full-time developer counts among DeFi protocols, signaling sustained infrastructure investment- 3 . However, this dominance also means that systemic risks within Aave V3 are, by definition, systemic risks for the broader DeFi lending sector. The protocol's market share exceeds 50% of the lending category- —a concentration that amplifies the consequences of any material failure. III. Risk Vectors: Where the Risk Premium Resides 3.1 E-Mode Leverage Density The most pressing risk vector is the density of leverage concentrated within E-mode positions. As of May 2026, Galaxy Research data shows that Aave V3 carried $10.7 billion in loans against $17.37 billion in collateral across all markets- 17 - . E-mode alone accounted for $6.3 billion in debt against $7.05 billion in collateral—a debt-to-collateral ratio of approximately 89.4%, with a debt-weighted health factor near 1.05- 17 - 20 . To contextualize: a health factor of 1.0 is the liquidation threshold. A debt-weighted health factor of 1.05 means that the margin for error across E-mode positions is approximately 5%—a 5% adverse price movement in correlated assets could trigger cascading liquidations. As Galaxy Research analyst Charles Yu stated: "The leverage density in E-mode creates a fragile equilibrium; even a 5% adverse price movement could trigger a cascade of liquidations"- 20 . The concentration is not merely numerical—it is structural. Concentrated leveraged positions are clustered on highly correlated assets: ETH, liquid staking derivatives, and liquid restaking derivatives, whose price movements affect both sides of leveraged positions simultaneously- 20 . 3.2 Cross-Chain Bridge Vulnerability: The KelpDAO Incident The April 2026 KelpDAO exploit exposed a hidden systemic risk: not buggy lending code, but a compromised cross-chain bridge- . Attackers exploited KelpDAO's LayerZero-powered bridge, minting approximately 116,500 rsETH tokens valued at roughly $293 million without any corresponding burn- 1 - . The stolen rsETH was then deposited as collateral on Aave V3 to borrow WETH, generating approximately $195 million in bad debt- 1 . Crucially, Aave's Chainlink exchange-rate feed continued pricing rsETH at its pre-exploit ETH redemption rate throughout the attack. The oracle functioned exactly within its design scope—but had no mechanism to detect that the asset had become instantaneously under-backed at the bridge layer- . The consequences were severe: Aave's TVL fell from roughly $26.4 billion to $18.6 billion, stripping the protocol of its position as the largest DeFi platform by deposited assets- 1 . Aave V3's USDT and USDC lending pools reached 100% utilization, leaving more than $5.1 billion in stablecoins locked until new liquidity arrived or outstanding borrows were repaid- 1 . The AAVE token fell nearly 20% in roughly 25 hours, dropping from $112 to $89.5- 1 . Recovery was partial: attacker positions on Aave V3 (Ethereum Core and Arbitrum) were liquidated on May 6, with 106,993 rsETH recovered out of approximately 112,103 unbacked rsETH on affected L2s—a recovery rate of 95.4%- 20 - 17 . However, residual bad debt remained for governance to address- 20 . 3.3 Stablecoin Cap Pressure and Governance Latency The USDe supply dynamics on Aave V3 MegaETH illustrate a second-order risk: the speed at which supply caps can fill and the latency of governance responses. USDe reached 99.5% of its $400 million supply cap, with a single supplier providing over $200 million in a single position- 20 . The Risk Steward process doubled the cap to $800 million after reserves refilled within three days—but the episode demonstrated how rapidly filling caps can alter utilization rates and borrowing costs, impacting borrowers reliant on stablecoin liquidity for basis trades or leveraged strategies- 20 - 17 . This governance latency is not merely theoretical. On April 23, 2026, Circle's Chief Economist Gordon Liao proposed adjustments to Aave V3's USDC interest rate curve, citing near-100% utilization and a flat 14% variable rate, accompanied by a $60 million drop in supply within 24 hours- . The USDC pool had maintained a 99.87% utilization rate for four consecutive days following the KelpDAO vulnerability event- . 3.4 Oracle Risk: The CAPO Incident On March 10, 2026, a misconfigured CAPO oracle system in Aave triggered $27 million in forced liquidations, undervaluing wrapped staked Ether by 2.85% against its actual market rate- . According to the post-mortem by Chaos Labs, the oracle error caused cascading liquidations that were fundamentally mispriced- . This incident underscores that oracle risk—often treated as a "solved" problem in DeFi—remains an active vulnerability, particularly in volatile market conditions. IV. Governance Response and Evolving Risk Framework In response to these events, Aave governance has moved toward a more comprehensive risk posture. LlamaRisk submitted two governance proposals covering a four-layer protocol-wide risk standard: Asset Risk, Bridging Risk, Automated Risk Oracles and Monitoring, and Chain Risk- - . The framework applies to every asset on Aave V3, V4, and Aave Horizon, with non-compliant assets slated for removal- - . On May 29, 2026, Aave Labs introduced an ARFC (Aave Request for Comment) proposing a standardized Technical Asset Listing Framework across all three deployment versions, establishing unified technical requirements for ERC20 compatibility, oracle design, access controls, minting and burning mechanisms, upgradeability, cross-chain bridge exposure, audits, and external dependencies- - . The "Aligned Delegates Framework" proposal, introduced in April 2026, established an 85% voting participation threshold and 75% rationale publication baseline for automated governance systems- . This reflects a broader governance maturation—but also highlights the inherent tension between decentralized decision-making and the speed required to respond to real-time risk events. V. Conclusion: The Risk Premium Is Real Aave V3's dominance of on-chain lending is empirical and structural. With $12.659 billion in TVL per DefiLlama, a market share exceeding 50% of the lending category, and a lead of over $13 billion over its nearest competitor, the protocol has achieved a scale that compounds its advantages- 2 - - 3 . However, "blue-chip" status does not eliminate collateral stress- 17 . The data reveals three principal risk vectors that demand ongoing attention: E-mode leverage density: $6.3 billion in debt at an 89.4% debt-to-collateral ratio with a health factor near 1.05 creates a fragile equilibrium where a 5% adverse move could trigger cascading liquidations- 17 - 20 . Cross-chain propagation: The KelpDAO incident demonstrated that a bridge vulnerability in a connected protocol can become Aave's problem—and that oracle infrastructure designed for normal market conditions may fail to detect instantaneous asset devaluation- . Governance latency: The gap between utilization shocks and governance responses—whether for USDC rate curves or USDe supply caps—creates windows of elevated risk that active market participants must price- - 20 . For borrowers and treasuries operating on Aave V3, the implication is clear: price a risk premium, size positions conservatively on newer assets, maintain measurable health-factor headroom, and recognize that "blue-chip" infrastructure does not eliminate path-dependent collateral dynamics- 17 . The protocol's rails work—but they work within constraints that are increasingly well-documented and, for the attentive observer, increasingly well-priced.

What to Watch

• 一、市场结构：Aave V3的主导地位与集中化风险
• 截至2026年6月，Aave V3的总锁仓量（TVL）约为126.6亿美元-
• 。这一数字较4月中旬约170亿美元的峰值有所回落-
• ，主要受6月市场去杠杆化及KelpDAO安全事件引发的连锁反应影响-
• 。即便如此，Aave V3仍以数倍优势领先于其主要竞争对手——Compound（约30亿美元）与Spark Protocol（约40亿美元）-