DeFi Week 30: TVL rebounds, but flows split (2026)

DeFi TVL rose from $73.318B to $75.885B across DeFiLlama's July 13 and July 20 daily snapshots, a $2.567B increase or 3.50%. The rebound was concentrated: Ethereum added about $1.644B, while Monad and Robinhood Chain posted much faster percentage gains. Hyperliquid L1 lost about $127M, and Arbitrum lost about $20M. 1 2

Data cutoff: July 20, 2026, 10:00 a.m. GMT-5. The chain comparison uses DeFiLlama's daily snapshots nearest that cutoff, so it is a daily-point comparison rather than an exact hour-by-hour measure.

Week 30 quick scan

Signal	Result	What it means
Total DeFi TVL	$75.885B, up 3.50% from $73.318B	A broad rebound, but not a uniform one. 1
Largest chain dollar gain	Ethereum: +$1.644B, +4.15%	Most of the aggregate increase came from the largest chain. 2
Fastest large-chain gain	Robinhood Chain: +95.23%, to about $260M	The percentage move is large because the base is still small. 2
Largest protocol dollar gain	Lido: +$890.5M, +5.45%, to $17.244B	Liquid staking moved with the Ethereum rebound. The API does not separate price effects from new deposits. 3
Largest protocol dollar loss	DoubleZero Staked SOL: -$284.4M, -71.09%	The loss happened while Solana TVL rose, so it was not a chain-wide drawdown. 3
Security total	$19.659M across six quantified records, plus one unquantified Across entry	Ostium accounts for about 92% of the quantified amount. 4
Governance	ENS SPP3 closed with For at 84.22%; two Lido votes were still active at cutoff	One completed social/procurement decision, two pending staking decisions, and no usable quorum field in the returned Snapshot data. 5 6 7

TVL: the rebound is real, but the flow is selective

The chain table is more useful than the aggregate headline. Ethereum reached about $41.37B of TVL, Base about $4.58B, Solana $4.90B, BSC $4.89B, Tron $4.77B, Hyperliquid L1 $1.31B, and Arbitrum $1.23B at the current API snapshot. 2

Chain	Latest TVL	7-day move	Dollar move	Read-through
Ethereum	$41.37B	+4.15%	+$1.644B	The main source of the week's dollar growth.
Solana	$4.90B	+1.15%	+$55.9M	Positive at chain level despite a large DoubleZero outflow.
Arbitrum	$1.23B	-1.60%	-$20.1M	A smaller continuation of the prior week's weakness.
Base	$4.58B	+2.81%	+$125.3M	Inflow continued, but at a slower pace than Ethereum.
BSC	$4.89B	-1.14%	-$56.2M	Mild outflow against the broader rebound.
Hyperliquid L1	$1.31B	-8.93%	-$127.2M	Last week's flow leader reversed.
Monad	$701.2M	+30.47%	+$163.8M	A high-growth move from a smaller base.
Robinhood Chain	$260.4M	+95.23%	+$126.9M	Fast percentage growth, but still too small to move aggregate DeFi TVL by itself.

The protocol leaderboard reinforces that concentration. In a DeFiLlama screen excluding CEX records and requiring more than $50M of current TVL, Lido rose from a back-calculated $16.354B to $17.244B, Aave V3 from $13.073B to $13.853B, and Babylon Protocol from $2.540B to $3.287B. Those are the three largest dollar gains in the screen. 3

The loss side was more fragmented. DoubleZero Staked SOL fell from about $400.1M to $115.7M, Ethena USDtb from $727.2M to $454.1M, and Portal from $2.579B to $2.323B. BlackRock BUIDL fell from $3.689B to $3.438B. The raw protocol feed does not attach causes to these moves, so they should be treated as flow signals rather than proof of deposits, withdrawals, redemptions, or user growth. 3

The useful comparison is between chain and protocol direction. Ethereum, Lido, and Aave all moved up together, which is consistent with an ETH-linked rebound, but the data does not tell us how much came from token-price appreciation versus fresh capital. Solana rose while DoubleZero collapsed, which points to protocol-specific repositioning rather than a Solana-wide exit. Hyperliquid's reversal also matters: the chain that absorbed much of Week 29's attention was down nearly 9% on the daily TVL comparison this week.

Exploits: oracle, governance, lending, and account abstraction

DeFiLlama lists seven security records in the window from July 13 through July 20. Six have amounts, totaling $19.659M; the Across record has no amount in the feed. The quantified records are Lumi Finance ($270,000), Chi Protocol ($8,500), BarnBridge ($776,000), Drips Network ($24,900), Ostium ($18M), and DeFiTuna Lending ($580,000). 4

Protocol	Chain	Reported loss	Attack surface	Status at cutoff
Ostium	Arbitrum	Up to $18M	A registered PriceUpKeep forwarder and future-dated authorized oracle reports triggered a vault payout.	All trading paused; investigation ongoing; final loss not confirmed. 8
BarnBridge	Ethereum	$776,600 USDC	The attacker bought BOND voting power, passed a proxy-upgrade proposal, and replaced the SmartYield/controller logic.	Funds were converted to about 415 ETH and remained at another address in the reviewed report. 9
DeFiTuna	Solana	$580,000	The lending pools were exploited, leaving a matching deficit in the USDC pool. The specific vector was not disclosed in the report.	Attack path patched; recovery and investigation ongoing; compensation not disclosed. 10
Lumi Finance	Arbitrum	$270,000	A logic flaw in the Sodium smart account's `validateUserOp` flow let a failed signature check reach an attacker-controlled `isValidSignature`.	No recovery or remediation status was established in the reviewed report. 11

Ostium is the dominant dollar event, but BarnBridge is the more important governance warning. A protocol can have audited or unchanged application logic and still be vulnerable if token voting can cheaply authorize an upgrade. The relevant check for DAO-token holders is therefore not only quorum, but also the cost of buying enough voting power, the delay before execution, and whether an upgrade can move funds immediately.

DeFiTuna is a different risk. The reported $580K deficit is a direct balance-sheet problem for the USDC lending pool, even though the attack path has been closed. Until a recovery or recapitalization plan is disclosed, the patch is not the same thing as making depositors whole.

Governance: one closed ENS vote, two Lido decisions still pending

ENS's [7.1] [Social] SPP3: Marketplace RFP vote opened on July 13 and closed on July 18. For received 1,191,365.53 score, Against 156,420.09, and Abstain 66,836.54, for an 84.22% / 11.06% / 4.72% split of the returned voting score. The vote was marked closed, so the decision was approved by the Snapshot result. Its title points to a marketplace procurement decision rather than an immediate lending, collateral, or emissions parameter change. 5

Lido had two votes running at the cutoff: Adopt the CMv2 Penalty Framework and 0x02 CSM: New Permissionless Module Launch. Both were scheduled to close at 11:00 a.m. GMT-5 on July 20, one hour after this edition's data cutoff. The live Snapshot scores showed For at 100% and Against at 0% at the time of retrieval, but neither should be treated as a final result before the deadline. The stated impact is on staking-operator penalties and access to a permissionless Community Staking Module, not on a completed change to current staking rewards. 6

The returned Snapshot records reported quorum: 0 for the ENS and Lido items. That is not a usable quorum threshold, so this edition does not claim that either vote met or failed quorum. The Aave DAO query had no proposal starting or ending inside the current window; its latest listed vote closed before the July 13, 10:00 a.m. GMT-5 window start. 7

Yield anomalies: extreme APY is a data-quality and sizing problem

The current DeFiLlama yield feed contains several prints that are too extreme to treat as baseline income. The right comparison is current APY against the pool's seven-day base APY and 30-day mean, not the headline number alone. 12

Pool	Chain	TVL	Current APY	7-day APY move	30-day mean	Trading read
WSTETH-AAVE on Balancer V2	Ethereum	$12.109M	260,979%	+260,979%	40,695%	Extreme fee or denominator behavior; do not size from the print without inspecting volume and fee history.
USDC-VELVET on Aerodrome Slipstream	Base	$3.659M	224,990%	+65,669%	249,378%	A persistent-looking high print is still not proof of durable organic yield.
APXUSD-USDC on Curve	Ethereum	$6.569M	4.90%	-74.12%	5,455%	A collapsing APY beside a stressed synthetic dollar is a risk signal, not a stablecoin carry trade.
BTC.B-WAVAX on Pharaoh V3	Avalanche	$2.409M	19.61%	-287.18%	55.81%	The rate has fallen sharply from its recent average; size only after checking incentives and liquidity.
WHYPE-USDC on Ramses CL V2	Hyperliquid L1	$1.779M	130.66%	-5.81%	109.85%	High current APY, but the chain-level TVL reversal argues for a smaller risk budget.

The APXUSD pool is the cleanest warning because the anomaly is a collapse rather than an eye-catching spike. The feed shows current APY at 4.90% against a 30-day mean of 5,455%, while the asset pair's seven-day APY change is -74.12%. That combination says the historical fee opportunity has not persisted; it does not say the peg has recovered or that the pool is safe. 12

The two six-figure APY readings should be treated as filters for further investigation. They can reflect a short fee window, a rapidly changing pool denominator, or incentive mechanics. DeFiLlama's payload provides the rate and comparison fields, but it does not establish that either pool's return is repeatable.

Position implications

For ETH-linked exposure: Ethereum's +$1.644B chain move and the gains in Lido, Aave V3, EigenCloud, and SparkLend support the direction of the rebound. The protocol data does not isolate asset-price effects, so do not equate TVL growth with new deposits. 2 3
For Hyperliquid and Arbitrum exposure: Hyperliquid L1's -8.93% daily TVL comparison and Arbitrum's -1.60% move show that last week's rotation did not continue evenly. Recheck perp liquidity and protocol-level flows before treating the Week 29 pattern as a trend. 2
For security-sensitive lending and vault positions: Ostium, BarnBridge, and DeFiTuna require different checks. Oracle authorization, governance upgrade paths, and pool deficit recovery are separate risks; a general audit badge does not answer all three.
For yield farmers: Compare APY with the seven-day base and 30-day mean, then inspect pool TVL, incentives, and exit liquidity. The largest APY prints in this feed are precisely the ones least suited to blind sizing. 12
For governance-token holders: Treat ENS as a completed Snapshot decision, but keep Lido's two staking votes in the pending bucket until they close and any execution step is visible. 5 6

The next checkpoint is whether Ethereum's rebound persists while Hyperliquid and Arbitrum remain soft. That split will tell traders more than the 3.50% aggregate increase by itself.

DeFi Week 30: TVL rebounds, but flows split