DeFi Week 39: TVL gains 6.6% as ether's price does the work
DeFi TVL rose 6.6% to $93.12B, a move that tracked ether's own price, while Nostra lost $3.5M to a manipulated oracle on Starknet and Arbitrum's DAO permanently banned three grant recipients.
DeFi's aggregate total value locked ended the weekly window at $93.120B, the daily snapshot for 20 September, up $5.768B or 6.6037% from $87.351B on 14 September. 1 Ether moved 6.27% over the same seven days, from $2,476.68 to $2,632.01. 2 The two figures track each other closely, and the ether-denominated reading settles what that means for anyone reading the number as a flow: the book went from 35.270M ETH on 14 September to 35.380M ETH on 20 September, a gain of 0.31%.
The scheduled window runs from 14 to 21 September 2026, 10:00 GMT-5, and the TVL series uses the nearest daily historical balances rather than hour-level readings. Two security incidents, three permanent DAO bans on Arbitrum, and a set of tokenized-equity pools carrying four-figure yields all landed inside it.
Quick scan
| Metric | Verified reading | Strategic significance |
|---|---|---|
| Aggregate DeFi TVL | $87.351B to $93.120B; +$5.768B / +6.6037% 1 | Ether's 6.27% weekly gain accounts for nearly the whole move; in ether terms the book was flat. |
| Largest dollar gainer | Lido: +$2.371B; $24.336B to $26.707B 3 | Liquid staking balances repriced with ether across five chains. |
| Largest dollar loser | Grove Finance: -$696.8M; $1.971B to $1.275B 3 | Its Ethereum book stepped down twice, on 16 and 17 September, while Base, Plume and Avalanche held flat. 4 |
| Second-largest loser | Anemoy Capital: -$374.8M; $725.1M to $350.2M 3 | The tokenized Treasury fund lost the same shape of step, on Ethereum between 17 and 18 September. 5 |
| Largest security incident | Nostra Money Market: $3.5M on Starknet 6 | A manipulated NSTR price feed let one account borrow roughly six times the token's entire market value. 7 |
| Second security incident | Fetch.ai: $1.53M on Ethereum, plus NuNet: $462,730 6 | One valid conversion signature drained Fetch.ai's converter; the same wallet took the NuNet mint. 8 |
| Binding governance outcome | Arbitrum bans Good Entry, Limitless and APX Finance 9 | Each vote carried with more than 163.8M ARB in favour against fewer than 16,000. |
| Highest displayed APY | Aerodrome Slipstream USDC-PROS: 51,640.60% at $871,576 TVL 10 | Reward emissions supply 51,629.28% of it; base swap fees supply 11.32%. |
The gain came from the ether price
The daily series shows the same relationship at higher resolution. TVL drifted from $87.351B on 14 September down to $86.008B on 16 September, recovered to $88.124B by 18 September, and then jumped $4.673B in a single day to $92.797B on 19 September. 1 Ether moved from $2,446.14 to $2,611.56 on that same day. 2
Loading chart…
| Metric | 14 September | 20 September | Change |
|---|---|---|---|
| Total DeFi TVL | $87.351B | $93.120B | +6.60% 1 |
| Ether price | $2,476.68 | $2,632.01 | +6.27% 2 |
| Bitcoin price | $76,819 | $81,236 | +5.75% 11 |
| DeFi TVL measured in ether | 35.270M ETH | 35.380M ETH | +0.31% 1 |
TVL values token balances at current prices, so an asset's own price move raises the reported total while the underlying quantities stay where they were. A protocol that lost depositors can still print a higher balance in a week like this one. Reading the protocol and chain tables in ether rather than dollars is what separates the two effects.
Where the tracked book sits
The chain endpoint totals $95.916B across 467 tracked networks at its latest reading, with Ethereum holding $54.163B of that, or 56.469%. 12 The endpoint returns current balances by chain, so it shows where the book sits rather than how it moved during the window.
| Chain | Current TVL | Share of tracked total |
|---|---|---|
| Ethereum | $54.163B | 56.469% 12 |
| Solana | $6.454B | 6.729% 12 |
| Base | $6.135B | 6.397% 12 |
| BSC | $5.875B | 6.125% 12 |
| Tron | $5.659B | 5.900% 12 |
| Bitcoin | $4.629B | 4.826% 12 |
| Arbitrum | $1.471B | 1.534% 12 |
| Hyperliquid L1 | $1.305B | 1.361% 12 |
| Monad | $1.044B | 1.089% 12 |
| Robinhood Chain | $998M | 1.040% 12 |
| Polygon | $796M | 0.830% 12 |
| Plasma | $550M | 0.574% 12 |
Solana and Base take the next two places, with BSC and Tron behind them. The five largest chains account for 81.620% of the tracked total, and Base holds its lead over BSC by $260.8M.
Protocol ranking: a repricing, and two withdrawals on Ethereum
The screen covers non-exchange protocols with at least $50M in current TVL. Before balances are back-calculated as current TVL divided by (1 + change_7d / 100), and dollar moves are rounded to the nearest $0.1M. 3
Top protocol dollar gainers
| Protocol | TVL before | Current TVL | 7-day change | Dollar move | What the feed shows |
|---|---|---|---|---|---|
| Lido | $24.336B | $26.707B | +9.74% | +$2,371.0M | Staked ether balances repriced across five chains. 3 |
| Aave V3 | $17.464B | $18.668B | +6.89% | +$1,203.8M | Deposits across 21 deployments moved with the market. 3 |
| Morpho Blue | $9.845B | $10.947B | +11.19% | +$1,101.9M | Ethereum rose from $4.499B to $5.007B; a new Arc deployment appeared on 17 September. 13 |
| SSV Network | $13.191B | $14.293B | +8.35% | +$1,101.8M | The Ethereum validator book repriced. 3 |
| SparkLend | $4.916B | $5.808B | +18.13% | +$891.2M | Ethereum rose from $4.774B to $5.591B, and borrows on the same chain grew 15.3%, from $2.333B to $2.691B. 14 |
| Binance staked ETH | $9.349B | $10.181B | +8.90% | +$832.4M | Liquid staking repriced. 3 |
| WBTC | $9.080B | $9.874B | +8.75% | +$794.3M | Wrapped bitcoin balances moved with BTC's 5.75% weekly gain. 3 |
| Coinbase Bridge | $7.947B | $8.647B | +8.81% | +$700.4M | The Bitcoin leg rose from $7.653B to $8.001B. 15 |
Top protocol dollar losers
| Protocol | TVL before | Current TVL | 7-day change | Dollar move | What the feed shows |
|---|---|---|---|---|---|
| Grove Finance | $1.971B | $1.275B | -35.35% | -$696.8M | Ethereum stepped from $1.418B on 14 September to $816.4M on 17 September and $722.8M on 18 September; Base, Plume and Avalanche stayed flat. Cause unresolved. 4 |
| Anemoy Capital | $725.1M | $350.2M | -51.69% | -$374.8M | Ethereum stepped from $650.0M on 17 September to $350.1M on 18 September. Cause unresolved. 5 |
| Lisk Bridge | $219.4M | $59.1M | -73.06% | -$160.3M | Ethereum swung from $227.5M on 14 September to $83.5M on 20 September, unwinding the migration-driven inflows of late August. 16 |
| HyperLend Pooled | $538.7M | $452.9M | -15.93% | -$85.8M | Supply on Hyperliquid L1 swung between $450.3M and $532.4M inside the window. 17 |
| Nexus BTC | $306.7M | $242.6M | -20.92% | -$64.2M | Bitcoin balances stepped from $277.1M on 14 September to $212.7M on 16 September. 18 |
| Exponent Yield Exchange | $113.0M | $75.5M | -33.20% | -$37.5M | Solana balances fell from $111.8M on 16 September to $70.1M on 18 September. 19 |
SparkLend is the one large gainer whose expansion outran the ether move for a reason the chain data can show. Outstanding borrows on its Ethereum deployment rose 15.3% over the window, faster than ether's 6.27%, so the collateral growth there reflects new borrowing demand on top of the repricing. 14
On the loss side, Grove Finance and Anemoy Capital both moved as step functions rather than drifting with prices, and both steps landed between 16 and 18 September on Ethereum alone. 45 Both sit as observed balance changes, with the cause unresolved pending a statement from either protocol.
Security: a manipulated oracle on Starknet, and one key that reached two converters
DeFiLlama recorded six incidents inside the window, totaling $6,141,006. 6 Oracle manipulation accounted for the largest single loss, and a compromised key produced the second- and fourth-largest losses.
| Protocol | Chain | Loss amount | Attack vector | Remediation status |
|---|---|---|---|---|
| Nostra Money Market | Starknet | $3,500,000 6 | Spot price oracle manipulation. 6 | Market paused; lending, borrowing, withdrawals and liquidations frozen while positions are reconciled. 7 |
| Fetch.ai | Ethereum | $1,530,000 6 | Private key compromise. 6 | Affected wallets and contract deactivated with SingularityNET; the team's on-chain analysis is preliminary. 8 |
| Flamincome | Ethereum | $595,000 6 | Spot price oracle manipulation. 6 | Undisclosed. 6 |
| NuNet | Ethereum | $462,730 6 | Private key compromise on the token's deployer account. 6 | Same exploiter wallet as Fetch.ai; NTX fell more than 70% and touched an all-time low. 8 |
| Nimiq | Polygon | $50,400 6 | Signature verification flaw. 6 | Undisclosed. 6 |
| Startale | Ethereum | $2,876 6 | Improper access control. 6 | Undisclosed. 6 |
Nostra paused its Starknet money market on 17 September after a manipulated NSTR price feed let one account borrow about $3.5M against NSTR collateral. 7 The attacker pledged NSTR and drew ETH, STRK, USDC, USDT, WBTC and DAIv1 from the market, and PeckShield tracked $1.92M of the proceeds bridged to Ethereum as 234.57 ETH and 1.3M DAI. 7 NSTR's entire market value stood at about $546,751, which puts the borrow at roughly six times the size of the collateral token, and DeFiLlama's record for the protocol fell from about $4M on 16 September to $710,632 at the time of the incident. 7 The protocol loss figure remains unpublished, and Nostra has warned users about impersonators during the reconciliation. 7
Starknet has now seen two lending-market oracle failures this month. On 4 September a bad price from the oracle provider Pragma's publishing pipeline triggered 47 liquidations across 42 borrower wallets on the money market Vesu; Pragma reported 95% asset recovery on 13 September. 7 Pragma's failure came from its publishing pipeline, while Nostra's came from someone deliberately moving a collateral price. 7
On 19 September an exploiter drained roughly $1.56M of FET from Fetch.ai's token converter on Ethereum, using a valid conversion-authorizer signature to call
conversionIn on the TokenConversionManagerV3 contract. 8 SlowMist's reading of the contract is that conversionIn leaves a single-EOA ECDSA signature as its only authorization check: the checkLimits(amount) modifier that guards conversionOut is missing from it, and the call verifies no on-chain allowance. 8 The same wallet received an NTX mint from NuNet's deployer account that Blockaid valued at roughly $452,000, taking the combined haul to about $2.01M; DeFiLlama carries that mint at $462,730. 8 The attacker swapped the proceeds for 546.36 ETH, worth about $1.44M at the time. 8 NTX fell more than 70% over 24 hours to an all-time low of $0.000328, while FET fell about 5% in a broader market decline. 8Governance: three permanent bans, a new security council, an RSR unlocking pair
Snapshot's index lists dozens of proposals that closed inside the window. The ones below changed protocol parameters, treasury signers, or program eligibility. Snapshot votes are off-chain decisions, and each one takes effect through the relevant DAO's own execution path.
| Protocol (space) | Proposal | Outcome | For / Against | Quorum | What it changes |
|---|---|---|---|---|---|
Arbitrum Foundation (arbitrumfoundation.eth) | Ban Good Entry | Passed | 163,882,585.95 / 7,518.70 9 | Not exposed | Bars Good Entry from future ArbitrumDAO programs. |
| Arbitrum Foundation | Ban Limitless | Passed | 163,829,365.05 / 15,649.00 20 | Not exposed | Bars Limitless from future ArbitrumDAO programs. |
| Arbitrum Foundation | Ban APX Finance | Passed | 163,897,837.23 / 3,386.46 21 | Not exposed | Bars APX Finance from future ArbitrumDAO programs. |
Reserve Protocol (reserveprotocol.eth) | RSR Unlocking: The Burn Gauge (1 of 2) | Passed | 1,100,538,734.71 / 136,200,070.01 22 | 1,000,000,000 | Changes how the RSR burn gauge allocates. |
| Reserve Protocol | RSR Unlocking: The Milestone Plan (2 of 2) | Passed | 833,278,316.45 / 711,423,924.11 23 | 1,000,000,000 | Ties RSR unlocks to completion milestones. |
1inch DAO (1inch.eth) | 1IP-106: Security Council and treasury multisig rotation | Passed | 28,052,420.45 / 0.11 24 | 10,000,000 | Creates a DAO security council and ratifies a rotation of the treasury multisig's signers. |
Spark Protocol (sparkfi.eth) | RLUSD force-deallocate penalty update, SAEP-21, SAEP-22, SAEP-23 | Passed | 610,238,145.31 / 0 25 | Not exposed | Adjusts the RLUSD force-deallocation penalty and three operational artifacts covering delegate compensation, risk curation and subproxy management. 26 |
Grove (grovefinance.eth) | Diamond PAU rate limits; USDS burn and USDC-to-USDS swap limits; Base vault onboarding | Passed | 84,883,502.56 / 0 27 | Not exposed | Raises the recorded rate limits on Grove's Ethereum allocator module, sets its USDS burn and USDC-to-USDS swap limits to unlimited, and onboards the Grove x Steakhouse USDC vault on Base. 2829 |
Arbitrum's three votes are the first permanent exclusions to come out of the DAO's Watchdog Program. The committee flagged 457,553 ARB of misuse across STIP and LTIPP grants involving Good Entry, Limitless and APX Finance, gave the projects until 10 September to answer the findings, and opened the three Snapshot votes when none of them replied on the forum thread. 30 Voter numbers ran into the hundreds: 1,002 on the Good Entry vote, 991 on Limitless and 1,003 on APX Finance. 26
Voting power on those three votes was heavily concentrated. The largest single voter on the Good Entry ban cast 32,676,435.31 ARB, and the twelve largest supplied 134,794,465.18 of the 163,882,585.95 in favour. 26
1inch's 1IP-106 clears its space's 10,000,000 quorum on a Yes total of 28,052,420.45 against 0.11 for No, and the five largest voters supplied 27,946,513.35 of that. 26 The Reserve pair was the week's most contested: the Burn Gauge carried 1,100,538,734.71 For against 136,200,070.01 Against, while the Milestone Plan carried 833,278,316.45 For against 711,423,924.11 Against, with 40,931,004.42 recorded as Abstain. 2223 Both cleared their 1.0B quorum on total votes cast, and on Snapshot's basic voting type an Abstain ballot counts toward quorum rather than against it. 31
Yield screen: emissions still carry the four-figure prints
Sixteen pools with at least $500,000 in TVL print APYs above 1,000%, and fourteen of the sixteen sit on Base. 10 Splitting each rate into base swap fees and reward emissions shows where the return actually comes from.
| Pool | Chain | TVL | Total APY | Base APY | Reward APY | 7-day base | 30-day mean | Risk profile |
|---|---|---|---|---|---|---|---|---|
| Aerodrome Slipstream USDC-PROS | Base | $871,576 | 51,640.60% | 11.32% | 51,629.28% | 9.48% | 84,050.22% | Emissions supply 99.98% of the rate; the 30-day mean sits above today's print. 10 |
| Aerodrome Slipstream CTR-USDC | Base | $558,726 | 42,907.35% | 3.35% | 42,904.00% | 8.80% | 51,352.57% | Base fees cover 0.008% of the headline figure. 10 |
| Aerodrome Slipstream USDC-VELVET | Base | $1,054,675 | 34,054.52% | 1.39% | 34,053.13% | 1.53% | 113,717.99% | The weakest base-fee contribution of the set, with a 30-day mean more than three times the current print. 10 |
| Aerodrome Slipstream WETH-CBBTC | Base | $1,668,133 | 12,490.95% | 421.72% | 12,069.23% | 74.29% | 2,883.05% | $226.7M of daily volume generated the 421.72% base rate; the pair carries two volatile assets. 10 |
| Aerodrome Slipstream USDC-MSTRC | Base | $987,578 | 8,752.00% | 279.49% | 8,472.51% | 76.51% | 3,131.19% | Real fee income on a small book, with emissions still at 96.81% of the total. 10 |
| Aerodrome Slipstream AVNT-USDC | Base | $702,686 | 4,513.92% | 0.80% | 4,513.12% | 0.69% | 7,253.94% | Swap fees are effectively absent; the rate is the reward token. 10 |
The pools where fees carry the return look entirely different, and they are the larger ones. Aerodrome's USDC-CBBTC pool holds $7.26M and prints 528.49%, of which base fees are 515.88%, on $12.59M of daily volume; the deeper WETH-CBBTC pool holds $10.13M and prints 366.41% on $60.19M of daily volume, with base fees at 108.46%. 10 Those rates come from trading activity rather than token distribution, which changes the risk farmers are actually underwriting.
Four of the week's outliers pair stablecoins against tokenized single stocks. Aerodrome opened tokenized-equity pools on 24 August for Nvidia, Apple, Meta and Alphabet shares, which carry a "c" suffix and represent a 1:1 claim on shares held in regulated custody through Alpaca under the Abu Dhabi Global Market's regulatory umbrella; the pools drew about $103M of volume within days of launch. 32 DeFiLlama's pool list now pairs USDC with tickers including MSTRC, NVDAC, AAPLC, TSLAC, MSFTC and AMZNC, and their yields follow the same pattern as the rest of the screen: USDC-NVDAC holds $2.20M and prints 2,045.96%, with emissions at 94.73% of it, and USDC-MSTRC prints 8,752.00% with emissions at 96.81%. 10
Trader and yield farmer action checklist
- Price the exit before entering any four-figure pool. Base swap fees supply under 12% of the headline rate on USDC-PROS, CTR-USDC, USDC-VELVET and AVNT-USDC; the rest is emissions, and realized return depends on the reward token's liquidity when you sell. 10
- Read the protocol tables in ether, not dollars. A balance that rose less than ether's 6.27% this week lost ground in ether terms, so compare each protocol's move against that figure rather than against zero. 3
- Re-underwrite any position that depends on Arbitrum grant programs. Good Entry, Limitless and APX Finance are now barred from future ArbitrumDAO programs, and any vault, market or gauge whose economics assumed those incentives needs a fresh look. 9
- Check the collateral side of every oracle-priced market. Nostra's loss came from a deliberately moved price feed rather than a contract bug, and the same chain had a second oracle failure on 4 September. 7
- Audit allowances held by token converter and queue contracts. Fetch.ai's converter authorized a full inventory drain from one external signature with no amount limit in the path, which is the same shape of exposure that legacy withdrawal queues carry. 8
- Watch the two Ethereum step-downs for follow-through. Grove Finance and Anemoy Capital both lost balance in a single step between 16 and 18 September, and neither has published a cause; a second step in the coming week would confirm withdrawals rather than a revaluation. 45
References
- 1DeFiLlama historical chain TVL API
api.llama.fi
- 2CoinGecko Ethereum market chart API
api.coingecko.com
- 3DeFiLlama protocols API
api.llama.fi
- 4DeFiLlama Grove Finance protocol API
api.llama.fi
- 5DeFiLlama Anemoy Capital protocol API
api.llama.fi
- 6DeFiLlama hacks API
api.llama.fi
- 7Altcoin Buzz on the Nostra oracle borrow
altcoinbuzz.io
- 8BeInCrypto on the Fetch.ai and NuNet exploit
beincrypto.com
- 9Snapshot vote: ban Good Entry
snapshot.box
- 10DeFiLlama yield pools API
yields.llama.fi
- 11CoinGecko Bitcoin market chart API
api.coingecko.com
- 12DeFiLlama chain TVL API
api.llama.fi
- 13DeFiLlama Morpho Blue protocol API
api.llama.fi
- 14DeFiLlama SparkLend protocol API
api.llama.fi
- 15DeFiLlama Coinbase Bridge protocol API
api.llama.fi
- 16DeFiLlama Lisk Bridge protocol API
api.llama.fi
- 17DeFiLlama HyperLend Pooled protocol API
api.llama.fi
- 18DeFiLlama Nexus BTC protocol API
api.llama.fi
- 19DeFiLlama Exponent Yield Exchange protocol API
api.llama.fi
- 20Snapshot vote: ban Limitless
snapshot.box
- 21Snapshot vote: ban APX Finance
snapshot.box
- 22Snapshot vote: RSR burn gauge
snapshot.box
- 23Snapshot vote: RSR milestone plan
snapshot.box
- 24Snapshot vote: 1IP-106
snapshot.box
- 25
- 26Snapshot GraphQL API
hub.snapshot.org
- 27Snapshot vote: Grove Diamond PAU rate limits
snapshot.box
- 28Snapshot vote: Grove Ethereum DPAU rate limits
snapshot.box
- 29Snapshot vote: Grove Base vault onboarding
snapshot.box
- 30The Defiant on the Arbitrum Watchdog bans
thedefiant.io
- 31Snapshot documentation on voting types
docs.snapshot.box
- 32PrimeXBT on Aerodrome's tokenized stock pools
primexbt.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.