DefiingerMulti-Chain DeFi Data, News & Research

Weekly DeFi Intelligence — 28 September–4 October 2026

A five-question briefing on 28 September - 4 October 2026: Networks, Stablecoins, DeFi, Yields and Outlook, each backed by a data table, plus a risk register and watchlist.

Weekly DeFi Intelligence2026-10-0411 min readDefiinger Research Desk2321 words

This briefing answers the same five questions every week, in the same order, so you can scan the ecosystem in two minutes or read it properly in ten. This edition covers the seven days ending Sunday 4 October 2026.

The value of a fixed frame is comparison: when an answer changes from last week, the change is the story. TVL, volume, stablecoin and fee figures come from DefiLlama's public API; yield levels are recomputed for this specific week from per-pool historical data, so they reflect the week named rather than whatever happens to be quoted today.

Quick answer

What is the one-sentence state of DeFi for 28 September - 4 October 2026? The week consolidated: fees held at $243.8M for a second week while volume stayed soft, TVL dipped 0.4% after six up weeks, and the stablecoin float rebounded 0.5% after three declines — while off the charts, NEAR Intents turned a $3.865m exploit into the rarest outcome in this series, a 100% recovery in roughly 36 hours.

1 Networks — how active is each chain?

Activity and fee generation by chain, week ending 2026-10-04.
ChainTVLW/WFeesW/WTurnover
Ethereum$53.3B-0.5%$83.9M-2.7%0.18
TRON$5.7B+0.4%$1.8M+0.0%0.05
BNB Chain$5.8B-0.2%$25.3M+27.8%1.27
Solana$6.5B-1.5%$111.7M-1.8%2.50
Base$6.4B+1.6%$12.9M-7.2%1.30
Arbitrum$1.4B-2.8%$5.8M-19.4%0.76
Optimism$0.5B+0.0%$2.4M+200.0%1.69

The fee level is the story: $111.7M at Solana (-1.8%) and $83.9M at Ethereum (-2.7%) kept the aggregate at $243.8M, a second consecutive week at the elevated level. Last week's question — which side of the divergence reverts — has an answer: neither, yet. BNB Chain was the week's exception at +27.8% to $25.3m, landing with its volume +13.8% after two weeks of unwinding.

Deposits tipped gently. Two chains grew TVL — Base +1.6% to $6.4B and TRON +0.4% to $5.7B — Optimism was flat, and the other four eased between 0.2% and 2.8%, leaving the aggregate down 0.4% at $79.6B, the first decline since the week ending 16 August. Turnover held at 0.55x — essentially flat, pausing the five-week slide.

2 Stablecoins — how are stablecoins performing?

Stablecoin supply ranked by size.
ChainStablecoinsShareW/W
Ethereum$146.9B51.6%-0.4%
TRON$94.3B33.1%+1.6%
BNB Chain$16.9B5.9%+0.4%
Solana$16.8B5.9%+0.5%
Base$5.2B1.8%+2.2%
Arbitrum$4.1B1.4%+7.3%
Optimism$0.6B0.2%+1.8%

The float rose 0.5% to $284.8B, the first weekly increase in four weeks and the end of the sample's longest drain. TRON swung from -1.3% to +1.6% ($94.3B), doing the most good this week after doing the most damage last week. Arbitrum added 7.3% — the largest float gain of any chain — Base 2.2%, and Ethereum was the only decliner at -0.4%.

One week answers a trend question in neither direction, but it reverses the burden of proof: the September drain never broadened into a structural story, and this week six of seven chains grew float. Regulatory news set the boundary conditions: the Fed's proposed GENIUS Act rules would put a two-business-day ceiling on issuer redemptions, with comments due 30 November — while the Andersen Institute's $76bn count of exchange-held stablecoins marks the layer the rule does not reach. Ethena, meanwhile, ended ENA incentives — a subsidy removal that the yield column will measure directly.

3 DeFi — what happened this week?

Liquidity and trading activity by chain.
ChainTVLShareDEX volumeTVL W/W
Ethereum$53.3B67.0%$9.4B-0.5%
Solana$6.5B8.2%$16.3B-1.5%
Base$6.4B8.0%$8.3B+1.6%
BNB Chain$5.8B7.3%$7.4B-0.2%
TRON$5.7B7.1%$0.3B+0.4%
Arbitrum$1.4B1.8%$1.1B-2.8%
Optimism$0.5B0.6%$0.8B+0.0%

Volume reshuffled rather than moved: the aggregate slipped 0.8% to $43.6B while the composition swung. Base jumped 28.1% to $8.3B — above even its late-August surge peak — BNB Chain rose 13.8%, and Optimism exploded 260.9% on a $0.83bn base, the same week Velodrome and Aerodrome announced their merger into Aero ahead of a 21 October launch. The three largest venues all fell: Arbitrum -34.0%, Solana -11.1%, Ethereum -10.3%.

The week's defining security event ended better than any in this series. NEAR Intents' Omni-layer bug produced about $3.865m of USDT outflows across five main withdrawals straddling 30 September and 1 October — after $10 and $11 test transfers the night before, per Bitquery. SHIELD monitoring flagged the activity; the patch landed within an hour; 11 networks stayed paused about 12 hours. General manager Alex Shevchenko then said 'We have identified you, sir', published return addresses on Bitcoin, EVM chains and Solana, and set 48 hours. ZachXBT's tracing through KuCoin and a bridge into Bitcoin was already public. The attacker opened negotiations with a 1 BNB message — 'Willing to cooperate, reply with your Signal so contact is possible' — returned roughly 34.59 BTC (about $2.95m) between 14:31 and 15:05 UTC on 2 October, and closed with an on-chain message at 16:15 UTC acknowledging being 'in the wrong' and recommending bug bounties. Shevchenko confirmed the full return and closed the investigation that afternoon, 2 October.

Around it: MetaMask's staking unit began exiting about 17,000 validators holding roughly 523,000 ETH after 18 of 19 block-proposing validators sent fee rewards to a Tornado Cash-funded address — 0.36 ETH diverted, withdrawal keys untouched, exits completing by 7 October across a ~45-day cycle. September's trackers closed at roughly $766m (+462% from August), and Bitget — September's dominant term — completed its phased withdrawal restart on schedule and restored its protection fund to $309m, including 3,705 BTC.

4 Yields — what is happening in the yield market?

Yield levels read from DefiLlama's yields endpoint and TVL-weighted where aggregated. Levels are a snapshot; the spread over a risk-free reference is the part we would act on.
StrategyRepresentative levelNote
Stablecoin lending, Ethereum4.30%TVL-weighted across all pools
Stablecoin lending, Base4.79%L2s currently pay a premium
Stablecoin lending, Arbitrum5.63%Highest of the large L2s
LST (ETH)2.19%Pure staking; lending collateral excluded
LST (SOL)4.91%Nominal yield above ETH staking
Blended stablecoin4.14%Across all tracked chains
LP (major pairs)variableIL risk dominates the outcome

Ethereum stablecoin lending sits at 4.21% TVL-weighted against a tokenized T-bill reference of 3.45% — a premium of 0.76pp. Every core aggregate moved four basis points or less, the quietest week in the archive: last week's re-widening held instead of extending, and the reference rate's third consecutive rise did the only visible work.

Arbitrum holds the widest spread at 1.75pp — in every week of the archive now — and Solana staking at 4.91% pays 1.46pp over the reference. ETH liquid staking at 2.19% remains the below-reference outlier. The structural item to watch is Ethena's ENA incentive stop: the first weeks without the subsidy will show whether sUSDe's rate was carried by the strategy or paid for in tokens.

5 Outlook — what to watch next week

Five items, each with the specific evidence we would treat as decisive.
What to watchWhyTrigger
Whether the fee level survives a third weekTwo consecutive readings near $243m against soft volume now read as a new operating level rather than event trafficA third consecutive week near $243m — or a clear break back toward the pre-spike range
Whether the stablecoin rebound continuesThe float rose 0.5% after three weekly declines, led by the same TRON that led the drainA second consecutive rise, or a resumption of the weekly contractions
MetaMask Staking's exit cycleFinal affected validators exit by 7 October; the full exit-and-reentry cycle takes about 45 daysThe post-mortem naming the entry path, and whether any slashing lands
NEAR Intents' promised post-mortemThe full amount was recovered, but the Omni interaction bug that allowed the loss has not been explained in publicPublication, and whether intent-based peers adopt the same patch class
The Aero merger launch on 21 OctoberOptimism's volume spiked +260.9% in the announcement week on a small baseOptimism and Base volume after launch, measured against this week's levels

The single most informative number is the fee level. Two consecutive weeks near $243m have converted last week's spike question into a level question: a third week near $243m would make the new operating level hard to dispute, while a break back toward the pre-spike range would return the divergence to unexplained. The second is the stablecoin float — whether the rebound was a one-week reversal or the September drain's end.

6 What would change our mind

Named conditions, so the calls above can be checked rather than argued.
Our current readEvidence that would overturn it
The fee level is durableA third week near $243m, or a fall back toward $220m with volume unchanged
The stablecoin drain was not structuralA second consecutive weekly rise, or a return to declines
The Optimism volume spike is merger-adjacent, not merger-provenVolume holding above $0.5bn into the Aero launch, or fading to baseline
Arbitrum's float growth is capital arriving ahead of tradingVolume recovering while float holds — or float following volume down
The NEAR Intents recovery closes the incidentThe post-mortem showing the bug was architectural, implicating peer protocols

7 Risk register

Judgements, not measurements — but they follow where this week's data points.
RiskLikelihoodImpactWhat would change our read
Layer-junction bugs in intent and bridge plumbingdemonstrated — NEAR Intents' Omni interaction bug cost $3.865m before a full recoverymediumThe post-mortem on how the bug reached production, and whether other intent bridges share the vulnerable pattern
Operator staking infrastructure compromisedemonstrated — a fee-recipient redirect forced ~17,000 validators (~523,000 ETH) into precautionary exitsmediumMetaMask's entry-path disclosure, and slashing outcomes across the ~45-day exit-and-reentry cycle
Issuer redemption gaps outside the new ruleformalised — the Fed proposes a two-business-day issuer ceiling while $76bn sits on exchanges governed by venue termsmediumThe final rule text after the 30 November comment deadline, and whether venues align their terms
Subsidy-dependent yieldstested — Ethena ends ENA incentives from 30 SeptembermediumsUSDe-linked rates over the coming weeks without the emissions
Fee level fragilityunresolved — two weeks at $243m-244m with no identified driverlowWhether the level holds a third week, or breaks down toward the pre-spike range

The register's newest line is also its most instructive: NEAR Intents' incident demonstrated that a layer-junction bug with a fast patch, wide pause, public identification and a traceable exit route can end with 100% recovered — an outcome this series had not recorded before. The MetaMask Staking case adds the operator-config variant: the compromised layer sat one step away from custody, so the cost was 45 days of foregone rewards rather than principal.

8 Formulas used in this briefing

Turnover = weekly DEX volume / TVL
Fee yield = weekly fees / TVL
Chain share = chain metric / ecosystem total
Yield spread = strategy APY - tokenized T-bill APY

9 The numbers behind the five answers

Every figure here is recomputed for this week and shared with the data report published alongside it.
QuestionKey metricValueDirectionNote
NetworksWeekly fees$243.8M+0.2%Second week at the elevated level
NetworksBlended turnover0.55x-0.5%Essentially flat; five-week slide paused
StablecoinsTotal supply$284.8B+0.5%First rise in four weeks
DeFiTotal TVL$79.6B-0.4%First decline in seven weeks
DeFiWeekly DEX volume$43.6B-0.8%Reshuffle inside a flat total
YieldsStablecoin lending, Ethereum4.21%+0.04ppQuietest week in the archive
YieldsTokenized T-bill funds3.45%+0.02ppThird consecutive weekly rise

10 How to read this briefing

  • The five questions never change, so a changing answer is the signal — read this week against last week rather than in isolation.
  • Every number comes from the same dataset used by the data report and news review this week; no column can disagree with another.
  • Yield levels are recomputed for the specific week, not quoted from today's snapshot.
  • Named judgments are labelled as such. Everything else is arithmetic on public data.
  • Where we hold an opinion with real uncertainty, the condition that would overturn it is stated in the table above rather than omitted.

Key Takeaways

  • Fees held at $243.8M for a second week while volume eased 0.8% — the fee/volume divergence resolved by not resolving; TVL slipped 0.4%, the first decline in seven weeks.
  • NEAR Intents lost about $3.865m to an Omni-layer bug, patched within an hour, paused 11 networks ~12 hours — and recovered 100% within roughly 36 hours of the theft, about 15 hours into a public 48-hour ultimatum.
  • MetaMask began precautionary exits of ~17,000 validators (~523,000 ETH) after a 0.36 ETH fee-recipient redirect; withdrawal keys and wallets were untouched.
  • Stablecoin float rebounded 0.5% to $284.8B after three weekly declines; TRON flipped from -1.3% to +1.6% and Arbitrum added 7.3%.
  • Yields went quiet: ETH lending 4.21%, premium 0.76pp, T-bill 3.45% on a third straight rise.
  • The Fed proposed a two-business-day issuer redemption ceiling while $76bn sits on exchanges outside the rule; Bitget completed its restart on schedule and restored its fund to $309m.

11 In brief

The data week was a consolidation, and every line confirmed it. Fees landed at $243.8M for a second consecutive week (+0.2%), volume eased 0.8% to $43.6B, and blended turnover held at 0.55x. Total TVL slipped 0.4% to $79.6B — the first weekly decline in seven weeks — and the stablecoin float did the opposite of its three preceding weeks, rising 0.5% to $284.8B.

The security answer was a first for this series: a complete arc. NEAR Intents lost about $3.865m of USDT to a bug where its Omni deposit-and-withdrawal system met its smart contract on BNB Chain, patched within an hour, paused 11 networks for about 12 hours, publicly identified the attacker, gave them 48 hours — and had every dollar back within roughly 36 hours of the theft, with an on-chain message recommending bug bounties. Alongside it, MetaMask began precautionary exits of about 17,000 Ethereum validators (roughly 523,000 ETH) after a fee-recipient redirect diverted 0.36 ETH of rewards, and September's loss trackers closed the month at roughly $766m, up about 462% from August, with Bitget's breach doing most of the damage.

Yields went quiet. Every core aggregate moved four basis points or less: Ethereum stablecoin lending at 4.21% against a tokenized T-bill reference of 3.45% — a third consecutive weekly rise — leaving the premium at 0.76pp, exactly where last week's re-widening left it.

12 Weekly scorecard

A compressed read; each row is expanded below.
DimensionReadDirectionOne line
NetworksThe level heldmixedFees flat at $243.8M, volume -0.8%, turnover 0.55x; TVL -0.4%
StablecoinsReboundingupFloat +0.5%, first rise in four weeks; TRON flipped to +1.6%
DeFiReshuffledmixedBase +28.1% and Optimism +260.9% volume; Ethereum, Solana, Arbitrum down
YieldsNoise weekflatETH lending 4.21%; premium 0.76pp held; T-bill third straight rise
SecurityA complete arcupNEAR Intents refunded 100% in ~36 hours; September total ~$766m
DE
Defiinger Research Desk

The Defiinger Research Desk compiles multi-chain DeFi data and commentary from public on-chain sources and vetted industry publishers. Our editorial process prioritizes verifiable figures and clearly dated references.

Sources & Methodology

  1. DefiLlama — Total Value Locked series across the seven chains tracked.
  2. DefiLlama — DEX spot volume series across the seven chains tracked.
  3. DefiLlama — stablecoin supply across the seven chains tracked.
  4. DefiLlama — network fees across the seven chains tracked.
  5. DefiLlama yields endpoint — per-week TVL-weighted lending, staking and stablecoin rates.
  6. NEAR Intents — statements by general manager Alex Shevchenko on 2 October and the investigation closure the same afternoon.
  7. Bitquery — on-chain measurement of the NEAR Intents outflows. ZachXBT — the tracing through KuCoin into Bitcoin.
  8. MetaMask and Lido — the staking incident notice, the validator exits and the 7 October exit deadline.
  9. Federal Reserve — the GENIUS Act implementation proposals announced 24 September and published in the Federal Register 29 September.
  10. PeckShield and CertiK — September monthly loss totals. Ethena — the ENA incentives change announced 26 September and effective 30 September.
  11. Jeeves, Walapay and PANews — the week's funding announcements.
  12. Chainlink and Swift — the settlement-pilot announcement of 28 September. CoinShares — the weekly flows report.
  13. Bitget — the withdrawal-restart completion and the 30 September protection-fund replenishment.

Headlines and figures on this page are drawn from the outlets listed above; commentary is clearly labelled opinion and is not investment advice. Last reviewed 2026-10-04.

Frequently Asked Questions

Why does this briefing answer the same five questions every week?
A fixed frame makes week-to-week comparison trivial and lets AI assistants quote consistent answers, which is the point of GEO-friendly structure.
Is the outlook a prediction?
No. It is a list of items worth monitoring with explicit triggers, not a forecast and not advice.
How do you decide direction labels?
From the same shared dataset used by the data report: up or down on week-over-week change, flat when the move is inside the trailing four-week range.
Are the yield levels live?
Yes. Yield levels are read from DefiLlama's public yields endpoint, and the activity figures come from the same weekly series the data report uses. See the Methodology page for filters and weighting.
Does this briefing cover incidents outside the seven tracked chains?
Yes, when they are material — NEAR Intents, MetaMask Staking and Bitget's follow-through this week — and we say explicitly when a figure sits outside our published dataset.
Is this investment advice?
No. Every judgment above is labelled, and each carries the condition that would overturn it.