DeFi Week 34: TVL slips 0.49% as Spark rotation and Harmony risk split the screen (2026)

The week's net TVL number was small, but the internal rotation was not. DeFi TVL fell from $75.683B on the August 10 daily snapshot to $75.309B on August 17, a $373.5M decline (-0.49%). Spark Liquidity Layer added $224.5M, while Sky Lending lost $205.8M; SSV Network and Morpho Blue also gained, and Bitcoin-denominated records were among the largest decliners. The week also brought a Harmony supply exploit and a set of governance votes that changed who controls treasury operations and how quickly Aave proposals can move. 1 2 3

Data cutoff: August 17, 2026, 10:00 a.m. GMT-5 (15:00 UTC). The historical comparison uses the nearest daily UTC snapshots: August 10 and August 17 at 00:00 UTC. It is not an hour-level reading. Protocol, chain, and yield rows use structured feeds retrieved just after the cutoff. Dollar moves before the current protocol snapshot are back-calculated as current TVL / (1 + change_7d / 100). The protocol screen includes records at or above $50M TVL and excludes CEX records.

Week 34 quick scan

Signal	Result	Trading read
Total DeFi TVL	$75.683B → $75.309B; -0.49%	A mild net pullback hid a sharp protocol-level rotation. 1
Largest protocol dollar gain	Spark Liquidity Layer, +$224.5M to $1.776B	The same Sky ecosystem also saw Sky Lending lose $205.8M; the feed does not prove a migration. 2
Largest protocol dollar loss	Sky Lending, -$205.8M to $5.450B	The cause is unresolved; treat the move as a flow-and-accounting question, not a confirmed risk event. 2
Quantified security event	Harmony: about 4B new ONE reported; no dollar loss in the DeFiLlama row	Mitigation was in progress, with exchange freezes, a patch, and a possible rollback reported; recovery was not reported. 3 4
Governance	ENS approved and executed its Foundation restructuring; Aave approved a shorter governance process	Treasury control and governance execution changed even as TVL slipped. 5 6
Yield anomaly	Aerodrome USDC-VELVET on Base printed 39,794.37% APY	39,789.13 percentage points came from rewards; the rate fell 14,895.24 points over seven days. 7

TVL dropped through August 15 before recovering on August 16 and 17. That path is consistent with a contained pullback, not a broad risk reset: the largest positive and negative protocol moves nearly offset each other inside one Sky-related cluster, while the ecosystem total moved only modestly. 1 2

차트를 불러오는 중…

Chain TVL: Ethereum still holds the balance

The chain endpoint summed to $75.497B, about $188M above the historical aggregate. That difference comes from source definition and timing, not an extra weekly inflow. The endpoint has current chain levels but no comparable 7-day change field, so this table is a composition snapshot rather than a chain-flow ranking. 1 8

Chain	TVL	Share of chain-feed sum	What the snapshot says
Ethereum	$41.424B	54.87%	The main base for Lido, Aave, Morpho, and most of this week's large protocol records. 8
BSC	$4.875B	6.46%	The second-largest balance in the endpoint, without a weekly-flow field. 8
Solana	$4.828B	6.39%	A large balance with several high base-APY pools in the yield screen. 8
Tron	$4.804B	6.36%	Material chain TVL, but not a ranked weekly gainer or loser here. 8
Base	$4.646B	6.15%	The home of the week's most extreme reward-heavy APY prints. 8
Bitcoin	$3.456B	4.58%	BTC-denominated balances were among the largest protocol-level decliners. 8
Provenance	$1.717B	2.27%	A sizeable balance outside the four most monitored DeFi bases. 8
Hyperliquid L1	$1.262B	1.67%	Still material for perp and lending exposure, though its bridge record declined. 8
Arbitrum	$1.241B	1.64%	The chain-level snapshot is stable enough to separate from the Harmony incident, which occurred on Harmony. 8
Monad	$892.4M	1.18%	Smaller base with high-yield records that need separate liquidity checks. 8

Ethereum's 54.87% share tells readers where the balance sits. It does not say that Ethereum supplied the week's net change. A flow conclusion needs historical chain series or event-level attribution; the current chain endpoint cannot provide it.

Protocol ranking: rotation inside a flat week

The current ranking below uses the same non-CEX, $50M-plus screen as the mover tables. Rank delta compares each protocol's current position with the position implied by its back-calculated TVL before the seven-day change. A positive number means the protocol moved up. TVL is a dollar-valued balance, not a clean measure of net deposits: token prices, bridge accounting, and methodology changes can move it without a matching wallet flow. 2

Current top 10 by TVL

Screen rank	Protocol	TVL	7-day change
1	Lido	$18.113B	+0.17%
2	Aave V3	$14.274B	+0.26%
3	SSV Network	$9.605B	+2.07%
4	Morpho Blue	$8.050B	+1.50%
5	WBTC	$7.258B	-1.98%
6	Binance staked ETH	$7.076B	-0.38%
7	LayerZero V2	$6.712B	+0.87%
8	Coinbase Bridge	$6.233B	-0.78%
9	Hyperliquid Bridge	$5.872B	-0.53%
10	Sky Lending	$5.450B	-3.64%

The top-ten order did not change on this screen. The meaningful changes happened below the top line: Spark Liquidity Layer moved up three places, while several smaller BTC and DEX records fell sharply.

Largest dollar gainers

Protocol	TVL before	Current TVL	7-day change	Dollar move	Primary reason in available evidence
Spark Liquidity Layer	$1.551B	$1.776B	+14.47%	+$224.5M	Unresolved. The same Sky family shows a $205.8M Sky Lending decline, but no source here proves an internal migration. 2
SSV Network	$9.411B	$9.605B	+2.07%	+$194.7M	Unresolved. The feed records the move but supplies no incentive, launch, or migration explanation. 2
SparkLend	$3.724B	$3.860B	+3.66%	+$136.3M	Unresolved. A positive move inside the Sky group is observed, not attributed. 2
Morpho Blue	$7.931B	$8.050B	+1.50%	+$119.3M	Unresolved. The feed does not identify a new product or incentive change. 2
Figure Markets Exchange	$1.615B	$1.717B	+6.28%	+$101.5M	Unresolved. This is a TVL observation, not a verified deposit event. 2

The Spark numbers deserve attention because the direction is split within one ecosystem: Spark Liquidity Layer rose $224.5M, SparkLend rose $136.3M, and Sky Lending fell $205.8M. That pattern can be consistent with product reclassification or capital movement, but the feed does not identify which. The correct label is observed concentration, not confirmed migration.

Lido was nearly flat on the weekly protocol screen, despite an attributed institutional-staking development. The Defiant reported on August 13 that SharpLink planned to stake $200M of ETH through Lido into wstETH, with Anchorage Digital as custodian. The report describes a plan, not a completed deposit, and the DeFiLlama row shows only a $31.1M seven-day increase. It is context for monitoring Lido, not proof of the cause of that move. 2 9

Largest dollar losers

Protocol	TVL before	Current TVL	7-day change	Dollar move	Primary reason in available evidence
Sky Lending	$5.656B	$5.450B	-3.64%	-$205.8M	Unresolved. The mirror-image Spark moves are a lead, not a confirmed explanation. 2
WBTC	$7.404B	$7.258B	-1.98%	-$146.4M	Unresolved. BTC-denominated valuation is an alternative to redemptions; the feed cannot separate them. 2
Binance Bitcoin	$4.425B	$4.338B	-1.97%	-$87.4M	Unresolved. The same BTC-denominated caveat applies. 2
Falcon Finance	$1.254B	$1.172B	-6.52%	-$81.8M	Unresolved. No exploit or incentive cause is identified by the TVL feed. 2
Uniswap V3	$1.414B	$1.336B	-5.49%	-$77.6M	Unresolved. The row does not distinguish LP withdrawals, price effects, or accounting. 2
Portal	$1.425B	$1.351B	-5.24%	-$74.6M	Unresolved. No causal announcement was verified in the available evidence. 2
Uniswap V4	$822.4M	$756.6M	-8.01%	-$65.8M	Unresolved. The size is notable, but the cause is not in the feed. 2

The two BTC-denominated losses should not be read as two independent withdrawal signals. A price component can move dollar TVL even when the token balance is unchanged. The available protocol snapshot cannot quantify that component, so the safer comparison is the size and direction of the record, not a claim about user exits.

Security: Harmony's supply event is the week's risk marker

The DeFiLlama hack feed returns one record inside the cutoff window: Harmony Bridge, dated August 11, classified as a token-and-share-accounting incident with an Infinite Mint technique. The record has no dollar amount, no returned-funds figure, and no source URL. 3

A Whale Alert report published August 12 says the apparent exploit created roughly 4B new ONE tokens. It says Harmony was working with exchanges to freeze funds, preparing a software patch, and considering a rollback. That makes the status mitigation in progress, not confirmed recovery or a completed patch. The 4B figure is newly minted token supply, not a dollar-loss estimate. 4

Incident	Chain	Amount lost or at risk	Attack vector	Current status in retrieved evidence
Harmony Bridge	Harmony	About 4B ONE created; USD loss unavailable	Infinite-mint / verification flaw; the DeFiLlama record classifies it as token and share accounting.	Ongoing mitigation: exchanges were being asked to freeze funds, a patch was being prepared, and a rollback was under consideration. No recovered-funds figure was reported. 3 4

No separate flash-loan-labeled record appeared in the returned DeFiLlama window. That is a statement about this feed, not proof that no other incident occurred elsewhere. For any position tied to Harmony or ONE, the first checks are the patch or rollback status, exchange freeze scope, and whether the affected verification path has been replaced rather than merely paused.

Governance: treasury control moved even while TVL stalled

Snapshot's official weekly summary says ENS closed its restructuring vote with 70.27% support on August 8. ENS Labs then announced on August 11 that the executable had been carried out onchain. The change established a staffed ENS Foundation with a full-time Executive Director and five-member board, while the Endowment gained a nine-day timelock and Security Council cancellation protection for transactions outside the Foundation's mandate. The summary does not state quorum, so quorum is unavailable here. 5 6

ENS's official explanation keeps 54.6% of total ENS supply in the DAO, leaves the roughly $16M operational wallet with the DAO, and permits one 1M ENS transfer restricted to employee compensation. Tokenholders retain protocol control and the ability to appoint or remove Foundation directors. The nine-day timelock and Security Council cancellation path change the operational risk surface around the roughly $65M Endowment; they do not turn the vote into a yield signal. 5

Snapshot's summary also reports two Aave votes closed during the week. The governance rulebook vote passed with 392,127 AAVE, 100% For, and removes the mandatory first review stage, shortening the standard process from 19 days to 13. The second vote, on accepting a fixed-yield claim on Agora's AUSD as collateral in Aave's Monad market, passed with 392,114 For, 11 Abstain, and 0 Against. The summary does not report quorum for either vote, and it does not claim that the collateral vote itself executed an onchain parameter change. 6 10

The Aave framework's stated execution path still requires an on-chain proposal after the binding Snapshot stage, followed by the applicable timelock. That distinction matters: a passed signaling vote changes the governance path or authorizes a next step; it is not automatically a deployed market parameter. 10

DAO / proposal	Result	Voter signal	Execution impact
ENS — Next Era of ENS DAO	Passed; executed	70.27% For; quorum not stated in the available official summary	Foundation operational, five-member board, nine-day Endowment timelock, one-time 1M ENS employee-compensation transfer. 5 6
Aave — Governance Framework v2	Passed	392,127 AAVE; 100% For; quorum not stated	Mandatory first review removed; stated standard process shortened from 19 days to 13. 6
Aave — Agora AUSD fixed-yield claim as Monad collateral	Passed	392,114 For; 11 Abstain; 0 Against; quorum not stated	Signals support for the collateral path; the summary does not claim execution of the market change. 6
Arbitrum — Security Council election process improvements	Pending in the August 11 roundup	Vote scheduled to end August 13; final result not in the roundup	Proposed a two-year council cohort, lower nominee threshold, and key-rotation permissions. 11

Governance is part of the risk screen because it changes the time and authority behind future parameter changes. ENS moved from proposal to execution in this window. Aave shortened its route to a vote, but the available evidence still separates community approval from deployed market changes. Arbitrum's roundup shows an active constitutional proposal rather than an outcome, so it should not be counted as passed.

Yield anomalies: reward emissions and fast-moving base rates

These are screening signals, not promised returns. The yield feed separates current APY into base and reward components and provides 7-day and 30-day changes. The most useful question is whether the headline rate is paid by durable fees, by a reward token, or by a transient market condition. 7

Pool	Chain	TVL	Current APY	Base / reward	7-day APY change	30-day mean	What stands out
Aerodrome Slipstream USDC-VELVET	Base	$4.586M	39,794.37%	5.24% / 39,789.13%	-14,895.24 pts	54,587.75%	The headline rate is almost entirely rewards and is falling rapidly. 7
Pharaoh V3 WAVAX-USDC	Avalanche	$1.607M	654.69%	0% / 654.69%	+533.17 pts	149.80%	The current print is entirely reward APY and is far above its 30-day mean. 7
Royco V2 JRROYAPYUSD	Ethereum	$1.148M	556.44%	556.44% / no reward field	+462.35 pts	222.21%	The feed reports a base-driven rate, but it does not explain the underlying cash flow. 7
Aerodrome Slipstream WETH-CBBTC	Base	$8.045M	138.82%	62.24% / 76.58%	-159.52 pts	295.35%	A larger pool still has a reward component, but its current rate is well below the 30-day mean. 7
Uniswap V4 ETH-01	Ethereum	$1.712M	126.38%	126.38% / no reward field	-231.63 pts	329.33%	A base-labelled rate also collapsed against its recent mean; the feed does not identify why. 7

The highest number is the least useful number in the table. Aerodrome USDC-VELVET has $4.586M of TVL, but 99.99% of its displayed APY comes from rewards. Pharaoh's pool has the same problem in a different form: its entire 654.69% print is reward APY. Before sizing either position, a farmer needs the reward-token liquidity, emission schedule, pool depth, and exit slippage. The yield feed alone cannot provide those checks.

Position screen for the next check

Concentrated protocol rotation: Spark Liquidity Layer's $224.5M gain, SparkLend's $136.3M gain, and Sky Lending's $205.8M loss are the week's clearest cluster. The current evidence does not establish whether the cluster reflects a migration, a methodology update, or separate balance changes. 2
Liquid staking: Lido's $31.1M gain was modest. SharpLink's reported $200M allocation was still a plan when The Defiant published it, so the next check is execution evidence rather than a larger headline. 2 9
BTC-backed records: WBTC and Binance Bitcoin lost $233.8M combined in the screen. Token-price effects remain an unresolved alternative to withdrawals; compare token balances or vault-level data before treating the decline as an exit. 2
Security: Harmony's roughly 4B newly minted ONE is a supply and verification risk, not a normal TVL drawdown. The next check is whether the patch and rollback are complete and whether exchanges released or retained freezes. 3 4
Governance: ENS execution changes Endowment controls, while Aave's passed rulebook changes the time from proposal to implementation. Neither result is an automatic yield or safety upgrade. 5 6
Yield farming: The largest APYs are reward-led, and several rates fell by more than 100 percentage points in seven days. Compare base versus reward, then test liquidity and exit mechanics before treating the print as an opportunity. 7

Week 34 was a small ecosystem pullback with a large internal split: TVL fell 0.49%, Spark-related records gained while Sky Lending lost, and the security screen moved from ordinary exploit accounting to an active supply-integrity incident on Harmony. The next useful update is not simply whether the ecosystem total rises; it is whether the Spark/Sky divergence receives a causal explanation, whether Harmony completes remediation, and whether the governance changes show up in deployed parameters rather than only in vote results. 1 2