The State of Lending on Ethereum — July 2026

Halved growth, deeper concentration

Aave V3 was the only major DeFi lending protocol on Ethereum that grew in July. Its net constant-price deposit flow was +$649M. The other five covered protocols (SparkLend, Morpho V1, Fluid, Compound V3, Euler V2) combined for a net −$96M in the same window. The sector's aggregate +$553M inflow was Aave V3, with a small amount left over. Without Aave, the sector would have contracted.

That is the shape of a depositor market running out of participants. July halved June's aggregate growth (from +$1.33B to +$553M) while curator concentration on the Morpho vault surface climbed for the second consecutive month (HHI 2,095 → 2,337), and the sector's real yield on stablecoin supply deepened to −75 bps below the 4-week T-bill (from −37 bps at the June close). Fluid's June rate leadership on USDC (the sector's only positive real-yield spread over T-bill at the June close) narrowed from +281 bps to −107 bps as its USDC supply APY fell from 6.41% to 4.66%.

Aave V3 carried the sector

Aave V3's +$649M July flow was collateral-driven, not stablecoin-driven. USDC on Aave V3 shed $20M across the month, a much smaller drain than June's −$162M but still a decline. USDT shed another $62M, and WETH shed $29M. The net inflow came through the LRT and BTC-variant reserves, with weETH alone contributing +$285M. This is the second consecutive month of the same pattern: collateral in, base stablecoin and ETH reserves flat to negative. Aave V3's growth mechanism in this cycle is depositors supplying collateral for borrow capacity, not depositors chasing supply APY.

Curator concentration continued climbing

Morpho V1 + V2 combined curator concentration read HHI 2,337 at the July 31 close, up from 2,095 at June 30 (+242 index points, +11.6%). Top-three curator share moved from 74.6% to 77.9% (+3.3 pp). Underneath that aggregate view, the composition rotated. Sentora grew from 27.65% at May 31 to 33.56% at July 31, overtaking Steakhouse Financial to become the largest single curator on the Morpho vault surface. Steakhouse held roughly flat across the two months (31.08% → 31.31%). Gauntlet declined for the second consecutive month, from 18.59% to 13.02%. The "concentration climbing" story is really "Sentora rising, Gauntlet declining, Steakhouse steady."

Real Yield Spread deepened

The sector's blended stablecoin supply APY minus the 4-week T-bill closed July at −75 bps, down from −37 bps at June 30. Real yield on stablecoin supply worsened relative to T-bills by 38 bps across the month. The mechanism is the same as June: soft stablecoin borrow demand while T-bill rates held around 3.60%.

Fluid's USDC supply APY fell from 6.41% at June 30 to 4.66% at July 31. Fluid still leads the sector on USDC supply, but the gap to Aave V3 (3.27%) narrowed from 322 bps to 139 bps. Fluid was the sector's only positive real-yield spread over T-bill at the June close (+281 bps); at the July close it sits at −107 bps below the T-bill.

The sector's borrow book barely moved

Sector loan-to-deposit ratio at July 31 was 40.63%, essentially unchanged from June's 40.99% (−36 bps). Borrow book utilization is not the July story. The July story is on the depositor side and the curator layer: growth halved, concentration climbed, real yield deepened, and one protocol absorbed almost all of what growth remained.

By the numbers

Snapshot date: July 31, 2026, 23:59 UTC. All month-over-month comparisons against June 30, 2026.

Net constant-price deposit flow, July

Balance sheet by protocol, July 31

USDC supply APY across the pool-based markets, July 31

SparkLend's USDC supply APY (3.47%) is set by a Sky-governance-linked interest rate model, not by pool utilization, so it is not a like-for-like comparison with the four pool-based markets above.

Rate compression across the sector

Fluid held the sector's only positive real-yield spread on USDC at the June close, paying depositors 281 bps above the 4-week T-bill. By July 31 it paid 107 bps below the T-bill. Fluid lost more than half its rate lead in a single month while the sector's real-yield deficit deepened by another 38 bps. Stablecoin depositors on DeFi lending markets are being paid less than a T-bill everywhere in the sector, and Fluid's June rate premium is being priced away.

The sector's blended stablecoin supply APY closed July at 2.85%, against a 4-week T-bill of 3.60%. Real Yield Spread: −75 bps. That is deeper than June's −37 bps close and marks two consecutive months of materially negative RYS after May sat close to parity at −0.3 bps.

Fluid's rate leadership narrowed

Fluid's USDC supply APY fell from 6.41% at the June 30 close to 4.66% at July 31, a decline of 175 bps in a month. Aave V3's USDC supply rate moved 8 bps across the same window (3.19% → 3.27%). The Fluid compression was idiosyncratic, not sector-wide. Fluid's lead over Aave V3 narrowed from 322 bps at June 30 to 139 bps at July 31.

What the compression means

Two forces are converging on the sector's stablecoin markets. First, T-bill yields held steady around 3.60% through July, providing a hard floor for any risk-free comparison. Second, stablecoin lending rates across the pool-based markets compressed toward one another and toward that floor. At the June 30 close the spread from top to bottom of the pool-based USDC field was 322 bps (Fluid 6.41% to Aave V3 3.19%). At July 31 it was 157 bps (Fluid 4.66% to Euler V2 3.09%). The dispersion halved.

Issue 003 closed with RYS at −37 bps and framed it as a compression signal. Issue 004 closes with it at −75 bps. Compression is now a trend, not a data point.

Where the July money went

The July flow map has one dominant venue, one clear loser, and four protocols that broadly held or added modest ground. Aave V3 alone accounted for 117% of the sector's net inflow, with the other five protocols combining for −$96M. SparkLend was the largest single-protocol contraction of the month.

Aave V3's composition

Aave V3's July composition mirrored June. USDC contracted by $20M across the month. The remaining +$669M came from collateral inflows across the LRT and BTC-variant reserves, with the largest additions in weETH (+$285M), sUSDe (+$145M), and WBTC (+$124M). The observation from Issue 003 continues to hold: Aave V3's growth mechanism in this cycle is collateral supply attracted by borrow-book depth, not stablecoin depositors chasing rates. Two consecutive months of the same pattern is now a trend, not a June anomaly.

SparkLend's outflow

SparkLend's July −$203M net constant-price flow was the sector's largest single-protocol contraction. The month contained two outsized single-day outflows that shape the total: July 12 (−$315M) and July 28 (−$173M). The other 29 days combined for approximately +$285M of offsetting inflows. The July 28 outflow followed the July 23 Atlas Edit spell that cut SparkLend's stablecoin borrow rate by 18 bps. The July 12 outflow does not align with any known Sky governance action.

The mixed middle

Morpho V1 added +$135M despite the curator layer's continued consolidation into V2 vaults through the month. Fluid added +$26M, a near-flat outcome: rate-leadership narrowing did not cause aggregate outflows at the protocol level. Compound V3 and Euler V2 each shed roughly $25M to $30M, modest contractions on protocols with smaller absolute TVL bases. The July flow map is not "everyone shrank." It is one protocol growing large, one shedding material capital, and four broadly holding — more consistent with a sector picking a winner than one broadly leaving.

Per-protocol coverage

Aave V3

Aave V3's July constant-price net deposit flow was +$649M, down from June's +$845M but positive for the second consecutive month — the only major covered protocol to have grown in each of the last two months. Available pool liquidity at July 31: $11.44B; borrow book $8.85B, the deepest in the sector; LDR 43.63%. Net stablecoin position grew by roughly +$145M, but the growth landed in the USDe complex (Ethena's synthetic dollar and its Pendle-tokenised forwards) rather than in USDC or USDT — a different depositor composition than early 2026.

SparkLend

SparkLend's July flow was −$203M, the largest single-protocol contraction of the month. Five of SparkLend's eighteen reserves price their borrow rate off a single external anchor rather than off pool utilization; all five are stablecoins, and all five stepped together on July 23. At 14:43 UTC that day, Sky's weekly Atlas Edit executive spell executed on-chain, cutting the Sky Base Rate (the wholesale cost of capital SparkLend pays for its USDS inventory) from 3.90% to 3.72%. SparkLend's on-chain USDC borrow rate followed the same day, stepping from 4.42% to 4.28%.

SparkLend is therefore not a single-mechanism protocol. It is a hybrid: a Sky-linked stablecoin book (DAI, USDC, USDT, USDS, PYUSD, all pointing at a single Sky-linked rate source) bolted onto a standard Aave-style collateral book of thirteen reserves that clear on utilization curves. Comparing Fluid's 4.66% supply APY on USDC to SparkLend's 3.47% is not a fair like-for-like: the two rates are set by different mechanisms responding to different inputs.

Morpho

Morpho's July constant-price net deposit flow was +$135M at the protocol level — the second-highest positive flow among covered protocols after Aave V3, even as the vault surface consolidated into fewer curator hands. Combined V1 + V2 curated TVL at July 31: $2.21 billion, of which V2 accounted for 71.76%. The three largest curators: Sentora (33.56% share, $741M, 3 V2 vaults), Steakhouse Financial (31.31%, $691M, 18 vaults), and Gauntlet (13.02%, $287M, 22 vaults, V1-heavy). Morpho also shipped Midnight, a fixed-rate and fixed-term lending product on Morpho V2, on July 24 — a categorical change for a depositor market, though not observable in July's flow data yet.

Fluid

Fluid's July constant-price net deposit flow was +$26M — essentially flat in a month when its USDC rate leadership narrowed materially. Available pool liquidity at July 31: $464M; borrow book $571M; LDR 55.19%, second only to Euler V2 and well above the pool-based cluster. If Fluid ran on a pure yield-chasing depositor base, a 175-bps rate compression would typically produce a visible outflow. It did not — a data point about depositor stickiness that August will test further.

Compound V3

Compound V3's July flow was −$23M, a modest contraction. Available pool liquidity at July 31: $908M; borrow book $460M; LDR 33.48%. Compound V3's USDC supply APY at July 31 was 3.19%, only 10 bps above the sector's laggard (Euler V2) and 8 bps below Aave V3 — not a rate-competitive position for a protocol that historically led on USDC yields in earlier cycles.

Euler V2

Euler V2's July flow was −$31M, the second-largest single-protocol contraction after SparkLend, on the smallest available-liquidity base among the covered protocols ($99M; borrow book $308M; LDR 74.86%, the highest in the sector). The vault surface is highly concentrated: Sentora runs 72.2% of Euler V2's curated TVL through five vaults, and the combined V1+V2 HHI is 5,512 — materially higher than Morpho's 2,337.

What August tests

Issue 004 documents a sector at half speed. Growth halved from June, concentration on the curator layer climbed for a second consecutive month, the real-yield deficit deepened, and one protocol carried most of the aggregate flow. Whether these are trend prints or one-month artifacts is what August resolves.

• Does one protocol continue to carry the sector? Two consecutive months of Aave-dominant growth (June +$845M, July +$649M) is not a base case for a healthy multi-protocol sector.
• Does curator concentration keep building? A third consecutive month of climbing Morpho HHI hardens the two-curator dominance framing; even a modest reversal reads as "consolidation peaked in July."
• Does the real-yield deficit deepen further? RYS went from −37 bps in June to −75 bps in July. Another 30+ bps deepening in August pushes the deficit toward levels not seen since March.

A month in which Aave V3 goes flat or negative while another protocol absorbs material inflow would falsify the concentration-in-flow trend. Without that, the base case for Issue 005 is: same shape, deeper.

Source: on-chain reads against each protocol's core contracts, DefiLlama, and the U.S. Treasury 4-week T-bill via FRED. Compound V3 and Euler V2 readings substitute on-chain Comet and EVK aggregations for DefiLlama's chain-TVL figures, which over-count both protocols.