DeFi Week 28: TVL rebounds to $73.7B

DeFi TVL recovered to about $73.7B during the Week 28 window from June 29 at 10:27 a.m. to July 6 at 10:00 a.m. GMT-5, reversing last week's SSV-led drawdown rather than starting a clean risk-on cycle. All six tracked major chains gained week over week, but the rebound was concentrated in staking and restaking while exploit risk clustered around vault accounting, weak collateral wrappers, and deposit-validation logic. 12

The actionable read for yield farmers is narrower than the headline TVL bounce. ETH-linked staking collateral recovered, apxUSD improved but remained off peg, Base incentives still produced extreme headline APYs, and three new exploits totaled about $7.22M in confirmed losses. 234

The broad TVL move was real. Ethereum added about $2.222B to reach $39.652B, which put it at roughly 53.8% of the tracked chain total. Tron was the strongest percentage mover at +7.97%, Base added $231M, and Arbitrum rose only $16M even though Robinhood Chain launched as an Arbitrum-based Layer 2 during the window. 18

The protocol leaderboard shows where the recovery came from. Lido gained $1.670B, SSV Network gained $853.6M, Aave V3 gained $801.5M, Binance staked ETH gained $642.9M, and EigenCloud gained $430.0M. 2 Those five names are mostly staking, restaking, or lending infrastructure, so this was not a uniform rotation into every DeFi category.

SSV Network's move needs a narrower reading than a normal inflow. SSV recovered to $8.592B, up 11.03% for the week, but the protocol was still below the level before Week 27's $5.26B decline. 2 The DAO's June 22 decision to end the Incentivized Mainnet Program on June 30 likely changed how DeFiLlama classified staked ETH rather than proving that the entire amount exited the staking system. 9

Babylon Protocol was the main exception. Babylon fell 13.24% to $2.675B, a $408.1M decline, while Ethereum-linked staking names rose. 2 The clearest takeaway is category divergence: Bitcoin restaking did not participate in the ETH staking rebound.

Spark Liquidity Layer and Uniswap V4 also landed on the loser list. Spark fell 12.96% to $1.753B, while Uniswap V4 fell 16.99% to $791.0M. 2 The measured point is narrower: both protocols lost TVL during a week when the largest staking names recovered.

Maple Finance was another mixed signal. Maple TVL declined 2.80% to $2.359B, even as Maple launched syrupUSDG on Robinhood Chain on July 1 and said the product brings its institutional credit engine to USDG. 210 For now, the Robinhood distribution story is strategic, while the TVL data says it has not yet produced net protocol inflow.

Three new in-window exploits accounted for about $7.22M in losses: Summer.fi at roughly $6M, Hinkal Protocol at roughly $820,000 USDC, and Edel Finance at roughly $403,000. 41112 The common thread is not chain selection. The common thread is accounting assumptions inside contracts and wrappers.

Summer.fi was the largest incident. The Lazy Summer LVUSDC vault was exploited on July 6 through a donation-based ERC-4626 inflation attack, and the attacker used a $65.4M Morpho flash loan to manipulate vault accounting before extracting about $6M. 4 ForkLog cited Summer.fi's statement that protocol guardians were pausing all Lazy Summer vaults while the team investigated the root cause. 13

The APY print was part of the warning sign. PeckShieldAlert reported that LVUSDC's displayed APY briefly spiked to about 2.08M%, and CertiK's analysis said the attacker manipulated FleetCommander's totalAssets() accounting before redeeming $70.9M after a $64.8M deposit. 13 For farmers, any ERC-4626 vault with unusual share-price behavior now deserves a check on donation protection, share inflation guards, and oracle/accounting dependencies.

Hinkal Protocol was smaller but direct. Hinkal lost about $820,000 USDC on Ethereum on July 2 after an attacker abused the prooflessDeposit() function and then made repeated transact() calls to extract funds. 11 AMBCrypto reported that stolen funds were traced to Tornado Cash and Thorchain. 11

Edel Finance showed the same class of risk through tokenized-equity collateral. An attacker used a roughly 180,000 USDC Morpho flash loan to manipulate the wGOOGLx wrapper exchange rate, pushing the assets-per-share ratio from about 6 to about 79 and creating roughly 78x overvaluation. 12 DEXTools reported that wGOOGLx had 0 holders, a DEXT Score of 0, and almost no on-chain market behind it. 12 Edel paused V1 contracts and said the team would absorb the bad debt, but the larger lesson is independent of Edel: thin wrappers should not be treated as deep collateral just because the underlying reference asset is familiar. 12

Taiko Bridge moved in the other direction: from unresolved exploit to controlled reopening. Taiko completed a four-stage recovery by July 2, restored bridge reserves to full 1:1 backing by June 30, completed an independent security review, and reopened the bridge under controlled quotas. 14 BlockSec's prior analysis traced the original June 21 incident to an SGX enclave signing key exposed on GitHub and an attestation contract that did not reject DEBUG-mode enclaves. 15

The half-year baseline keeps the week in perspective. CertiK's Hack3D H1 2026 report recorded more than $1.31B in Web3 security losses across 344 incidents, with net losses of about $1.2B after freezes and recoveries. 16 CertiK also reported that code vulnerabilities accounted for 204 incidents and that attackers are increasingly revisiting contracts deployed more than one year earlier. 16

Arbitrum DAO narrowed Arbitrum Gaming Ventures after a vote that ended June 30. The proposal limits AGV to managing the existing portfolio and stops new gaming investments; AGV originally received a 225M ARB allocation roughly two years ago. 17 The useful read is scope: Arbitrum governance is pulling authority back from a specialized investment vehicle rather than expanding it.

ENS governance had the week's sharpest legitimacy problem. Brantly Millegan, an ENS Constitution author and former ENS Labs Director of Operations, said on July 3 that he was leaving ENS, removing .eth from his identity, and winding down EthID, Ethereum Follow Protocol, Grails Market, and ENS Market Bot. 5 The exit followed a Security Council renewal vote in which 71.3% of voting power abstained and only 26.9% voted For. 18

The ENS working-group vote also changed operating structure. The Term 7 Meta-Governance Working Group election closed June 30 with the "Metagov only" option at 601,311.95 votes, or 40.0% of voting power. 19 That result leaves Public Goods and Ecosystem working-group functions outside the new Term 7 structure, which matters for builders funded by ENS grants.

Solana launched a more formal governance rail on July 1-2. Solana Governance Proposals use stake-weighted, Merkle-verified voting; a validator needs at least 100,000 SOL delegated to open a proposal, a proposal needs 15% cluster stake support to enter voting, and passage requires a two-thirds supermajority of voting stake. 6 The Defiant reported that delegators can override validator votes, which keeps stake owners from being locked into a validator's default governance position. 20

The first queued Solana governance fights are economic. Solana Compass reported that SIMD-550 would double the disinflation rate from 15% to 30%, compressing the terminal-inflation timeline from about 5.7 years to 2.8 years. 21 SIMD-553 would introduce resource-based fees at 0.5 lamports per compute unit and burn 100% of those fees, lifting estimated daily SOL burn from about 648 SOL to 7,500-9,000 SOL. 21 SIMD-123 would route priority fees to delegators through a separate block-reward commission mechanism, after a March 2025 community vote passed with 74.91% support. 21

Aave's governance thread was less dramatic but directly relevant to risk parameters. The Aave Risk Framework ARFC moved to Snapshot pending vote on July 6, and the sGHO Cross-Chain ARFC was also pending. 22 Stani Kulechov also denied reports that Kraken was buying a 15% Aave stake at a $385M valuation and said Aave Labs is a DAO service provider while protocol revenue belongs to AAVE holders. 23

The yield regime flipped back toward incentives. DeFiLlama tracked about 15,800 pools, and the Week 28 filter found 75 anomalous pools with either APY above 100% or a seven-day APY move above 20 percentage points, provided TVL exceeded $1M. 3 The count was up 17% from Week 27's 64 pools, and the anomalies were concentrated on Solana (25 pools), Base (22), Ethereum (12), and Arbitrum (5). 3

Base still had the loudest incentive print. Aerodrome's USDC-VELVET pool showed 152,383.86% APY with $3.57M TVL, but the seven-day APY change was -500,629.54 percentage points from last week's 653,013% level. 3 Its reward APY was 152,379.88% while base APY was only 3.97%, so the farm remains an emissions trade rather than a fee-yield position. 3

Curve's APXUSD-USDC pool was the opposite kind of anomaly. The pool showed 3,882.34% APY with $4.52M TVL, and the source data attributed the full APY to base fees rather than token rewards. 3 The fee spike lines up with apxUSD's continuing depeg: apxUSD recovered from $0.794 to $0.898, but it remained about 10.2% below the $1 target. 7 This is fee yield created by unstable-asset arbitrage, not a low-risk stablecoin pool.

Hyperliquid's growihf USDC pool needs caution because the classification is not fully explained by the available yield fields. DeFiLlama showed 4,243.95% APY, a +4,243.95 percentage point seven-day move, and $9.17M TVL, while both base APY and reward APY fields were listed as 0%. 3 That mismatch makes the pool a data-validation candidate before sizing capital around the headline APY.

Solana had the largest count of anomalous pools among major chains. Orca USDT-SUSD showed 6,023% APY with $2.66M TVL, while Orca SOL-USDC cooled from 130.76% to 69.24% even though its TVL remained around $25.32M. 3 GMTrade's XAU-USDC pool also crossed 100% APY, which points to volatility in synthetic gold trading rather than a broad SOL staking-yield repricing. 3

Maple's token moved even though Maple's TVL did not. SYRUP rose 19.4% during the week to about $0.1484, helped by syrupUSDG and syrupUSDC launching on Robinhood Chain and Robinhood Earn offering 7% APY on USDG. 24 Maple's protocol TVL still declined to $2.359B, so traders should separate token momentum from deployed-credit growth. 2

For staking and restaking exposure, Week 28 argues for recalibrating buffers rather than assuming the Week 27 drawdown fully healed. Lido, SSV Network, Binance staked ETH, ether.fi, EigenCloud, and Rocket Pool all gained around 10% or more, but SSV remained below its pre-collapse level and Babylon moved against the group. 2

For vault users, Summer.fi and Edel put accounting paths at the center of the risk checklist. ERC-4626 share inflation, wrapper exchange rates, donation handling, and thin collateral markets all deserve review before a vault APY or collateral factor is treated as usable. 412

For APY hunters, the first filter is source of yield. Aerodrome USDC-VELVET was almost entirely reward-driven, Curve APXUSD-USDC was fee-driven by depeg arbitrage, and Hyperliquid growihf USDC had a field mismatch that needs validation. 3 Those three pools should not share the same sizing model just because their APYs all look extreme.

For governance-sensitive positions, Solana and ENS deserve the most attention. Solana is moving tokenomics questions into binding stake-weighted votes, while ENS is still working through a legitimacy fight around Security Council authority, Foundation scope, and working-group structure. 56