DeFi Week 29: Hyperliquid takes the flow

The Week 29 DeFi read is not the $73.234B total TVL print. It is the mismatch underneath it. From July 6, 2026 10:00 UTC-05:00 through July 13, 2026 10:00 UTC-05:00, total DeFi TVL slipped only 0.65% from the Week 28 baseline, but Arbitrum lost $593.5M while Hyperliquid L1 gained $535.9M. 12

For traders and farmers, that makes Week 29 a migration week rather than a drawdown week. Liquidity moved toward perp-DEX concentration, protocol risk moved toward governance and oracle surfaces, and high APY pools split into three very different buckets: emissions trades, depeg-fee trades, and data-validation candidates. 23

A flat aggregate can be deceptive when the biggest single-chain flow is larger than the total weekly move. Arbitrum's $593.5M outflow and Hyperliquid's $535.9M inflow almost net against each other, which points to chain-level migration rather than broad DeFi deleveraging. 2 DeFiLlama's chain ranking put Ethereum at about $39.79B, BSC at $4.90B, Solana at $4.84B, Tron at $4.71B, Base at $4.37B, Hyperliquid L1 at $1.37B, and Arbitrum at $1.23B as of July 13. 4

Keyrock framed the Arbitrum-to-Hyperliquid move as the on-chain expression of perp-DEX consolidation. Its wording was blunt: "Arbitrum bled -$593.5M on the week while Hyperliquid pulled in +$535.9M." 2 That matters for positioning because Hyperliquid flow is not a generic L1 beta trade. It is a liquidity and trader-concentration trade around perps.

Robinhood Chain created the other large distortion. Robinhood Chain launched on July 1, before the stated window, as an Arbitrum Orbit L2 for tokenized stocks and real-world assets. 10 The in-window signal came on July 8, when DEX volume spiked to roughly $570M against $21.68M of liquidity, a volume-to-liquidity ratio near 26:1. 11

The Robinhood Chain TVL number depends on the measurement source and timestamp. DefiLlama showed about $145M chain TVL by July 13, while launch-week articles reported TVL reaching about $234M-$250M during the first week. 41012 The harder signal is composition: tokenized RWA value was only about $12.8M, while CASHCAT and other memecoins drove the visible DEX activity. 10

The protocol leaderboard was just as split. BlackRock BUIDL rose 21.03% to $3.69B, adding about $641M, while Ethena USDe fell 12.42% to $3.88B, losing about $550M. 5 Keyrock said aggregate RWA AUM was nearly flat at -0.02%, but that flat number hid dispersion between Securitize's +15.61% gain and Maple's -11.57% contraction. 2

Aave carried the cleaner ETH spot-yield signal. Aave V3 reached $13.01B TVL, up 2.20% week over week, while WETH deposits rose 62.67% and WETH supply APY fell 13.46%. 25 Keyrock read that as ETH holders parking spot and accepting lower yield, which aligns with Ethereum's $51.5M net chain inflow. 2

Week 29 exploit risk was concentrated in three incidents, but the mechanisms were different enough that one security checklist will not cover them.

BonkDAO was the week's most important security event because nothing had to break at the code level. The attacker spent about $4.4M buying enough BONK to exceed the 1% quorum threshold, submitted BIP #76, won with 99.9% approval from only 7 participating wallets, and transferred 4.426 trillion BONK out of the treasury. 713 The relevant comparison is direct: the cost of acquiring quorum was far below the treasury value.

Bonzo Lend showed a different failure path. Bonzo Finance Labs said the loss did not originate in Bonzo Lend's own contracts and instead came from the upstream third-party price oracle. 14 The attacker submitted a forged SAUCE price update, deposited 250 SAUCE as collateral, and borrowed 6.63M USDC plus 34.52M WHBAR. 14 Bonzo Lend and Bonzo Points were paused, while Vaults, Bridge, and staking were reported unaffected. 14

Summer.fi was the slow-burn vault risk case. The Ark that enabled the exploit had been capped at zero deposits on October 30, 2025, but it was never removed from the active accounting set because removeArk required the Ark assets to be zero. 16 The attacker used about $65.4M in Morpho flash loans during the final transaction, but Summer.fi said evidence pointed to at least three months of preparation before the exploit. 16

The common lesson is narrow and practical. DAO treasuries need quorum-cost analysis; lending markets need oracle-verifier dependency checks; automated vaults need offboarding procedures that actually remove stale adapters from pricing. A farmer exposed to all three surfaces is not holding one kind of DeFi risk.

Aave DAO had the highest throughput. Five proposals passed during the window: Aave Risk Framework, sGHO Cross-Chain, Aave V4 on Avalanche, a dedicated Aave V4 whitelabel instance for EtherFi on Optimism, and the Technical Asset Listing Framework. 8 The vote counts were tightly clustered, with 75-82 voters per proposal and only the EtherFi whitelabel proposal showing meaningful opposition at about 80 votes against. 8

ENS governance moved from crisis response toward redistribution of authority. ENS DAO closed a ranked-choice Security Council election on July 12 with 50 voters and 1,756,770 total scores; the top scorers included Nick Johnson, Alex Van de Sande, Kevin Gaspar, Alex Netto, Colton Liberacki, Griff Green, and Hudson Jameson. 18 Alex Van de Sande also proposed delegating 5M ENS from the treasury across five stakeholder categories, with voting rights delegated but token ownership retained by the DAO. 19

Uniswap Labs proposed activating protocol fees on a subset of Uniswap v4 pools on July 7, extending the UNIfication fee switch that already covers v2 and v3 pools across 11 chains. 20 The proposal targets static-fee pools without hooks, CCA-launched pools, and aggregator-hook pools, and it introduces a V4FeeController built from a V4FeePolicy contract plus a V4FeeAdapter contract. 20 Guillaume Lambert of Panoptic opposed the move, arguing that "Turning on the v4 fee switch risks killing the protocol." 20

Arbitrum also added a revenue angle to the Robinhood Chain story. Offchain Labs co-founder Steven Goldfeder announced on July 9 that Arbitrum will collect 10% of fees from Robinhood Chain and other Orbit-based L2s, with 8% going to the ARB tokenholder-controlled treasury and 2% to development. 21 That fee claim does not offset Arbitrum's Week 29 TVL outflow by itself, but it changes how ARB holders should read external Orbit chain volume. 221

Lido's governance catalysts remained mostly pending rather than completed. Two Lido Snapshot proposals were created during the window in pending state, and voting was expected to begin around July 18 for the new permissionless CSM module and the CMv2 Penalty Framework. 22 Separate reporting said the LDO Staking Module, Liquid Buybacks via NEST, and LIP-35 Staking Router v3 remained in discussion or pre-Snapshot stages. 23 For LDO holders, the Week 29 move was expectation pricing, not completed revenue pass-through.

The most serious stablecoin signal was apxUSD. The peg worsened from $0.898 at the Week 28 close to $0.864 on July 13, and the seven-day range included a low near $0.825. 9 The Curve APXUSD-USDC pool on Ethereum had about $4.49M TVL, and its APY collapsed from 3,882% to 79% over the week. 3 That is a bad combination for stablecoin LPs: deeper depeg, lower arbitrage fee compensation, and thinner evidence that normal peg-recovery flow is still active.

Aerodrome's USDC-VELVET pool was the opposite kind of anomaly. The Base pool showed about $3.89M TVL and 159,321% APY on July 13, up roughly 6,937.5 percentage points over seven days. 3 That print did not confirm durable organic yield; it confirmed that the pool still needs incentive and liquidity-source analysis before sizing. The prior Week 28 expectation was cooling, so a renewed spike should be treated as a risk flag as much as an opportunity. 3

The new anomaly list had several smaller but tradable pools. Concrete CTFRXUSD+ on Ethereum showed about $3.95M TVL and 414% APY, Pharaoh V3 BTC.B-WAVAX on Avalanche showed about $2.50M TVL and 306.8% APY, and Hyperliquid WHYPE-USDC showed about $1.97M TVL and 136.5% APY. 3 Across the filter set, DeFiLlama data identified 55 anomalous pools with TVL above $1M and either APY above 100% or a seven-day APY move above 20 percentage points. 3

Robinhood Chain's stablecoin yield is cleaner than its memecoin volume, but the two should not be merged. Ethena deposited $50M USDe into a Steakhouse Financial-curated USDG vault on Morpho on July 8, which reporting linked to a 160% one-day Robinhood Chain TVL surge. 24 Maple also launched syrupUSDG on Robinhood Chain on July 1, bringing institutional credit strategies to USDG and supporting Robinhood Earn's approximately 7% APY product. 25 Those stablecoin flows can build TVL even if memecoins generate most of the attention and transaction velocity.

For chain exposure, Hyperliquid now deserves a separate bucket from generic L1 beta. Its $535.9M Week 29 inflow nearly matched Arbitrum's $593.5M outflow, and the underlying explanation is perp-DEX concentration rather than a broad rotation into every new chain. 2

For Arbitrum exposure, split the TVL and revenue questions. Arbitrum lost material chain TVL during the week, but Orbit fee-sharing gives ARB a claim on external Orbit activity such as Robinhood Chain. 221 Those are different time horizons and should not be collapsed into one bullish or bearish view.

For stablecoin LPs, apxUSD remains the highest-severity screen. A pool paying 79% after previously printing 3,882% does not compensate the same way when the asset trades near $0.864 and recently touched $0.825. 39

For vault users, Summer.fi should move offboarding status into the pre-deposit checklist. A capped adapter that still affects NAV can be economically live even when it looks operationally inactive. 16

For DAO-token holders, BonkDAO and ENS show the two sides of governance power. BonkDAO showed what happens when quorum is cheap relative to treasury value; ENS is trying to dilute concentrated voting power without transferring treasury ownership. 719