Same-Pool Funding Risk: Why Pooled Lending Mispriced LST Concentration

The Pattern, Stated First

A pooled-liquidity money market like Aave V3 takes deposits in an asset (say, WETH) and lends them out against many different collateral types. The borrower pays an interest rate that varies with utilization. The lender earns the same rate regardless of what collateral the WETH was lent against. This is the elegance of pooled liquidity: depth on day one, simple UX, deep markets. It is also the failure mode the April rsETH event surfaced.

When 98.5% of WETH borrow collateral is sourced from ETH LSTs, ETH depositors aren't financing a diversified book. They're financing a concentrated carry trade that rents their capital to LST holders looping leverage. The lender earns a single pool APR that reflects the average risk of the book; the actual realized risk distribution is sharply bimodal, with depositors on the wrong side of the tail. The pooled architecture compresses materially different risk profiles into a single rate, so depositors are opted into financing the strategy at a price that does not reflect the tail risk they are underwriting.

This is not a bug in Aave's contracts. The contracts behaved as designed throughout the incident. The mispricing is structural to the pooled-rate framework. What April surfaced was that the cross-subsidy was already large enough to be visible: when LST collateral concentrates above ~95% of WETH borrow demand, the pool APR a depositor accepts is roughly the right price for a diversified borrowing book and roughly the wrong price for the actual LST-loop concentration sitting in the contract.

Three Architectures Through the April Event

The April incident is the cleanest natural experiment to date on what pooled-liquidity lending does and does not handle well. Three of the largest DeFi lending venues experienced the same shock with three different architectural responses.

The most informative cross-sectional finding is the Aave-vs-SparkLend split. Both protocols share the same monolithic-pool architecture. SparkLend captured the largest share of the post-incident outflows from Aave, despite having essentially the same architectural exposure to LST concentration. The data point that resolves the apparent paradox is that SparkLend had proactively deprecated rsETH in January 2026, three months before the event, when Aave's risk framework still listed it as an eligible collateral at competitive LTVs.

Stated bluntly: architecture alone doesn't determine trust, underwriting discipline does. Capital rotated to whichever protocol was perceived to be managing risk most credibly during the moment of stress, and architectural similarity to Aave was not the disqualifier. Spark's parameter conservatism and pre-incident deprecation decision were the differentiators. This complicates the simple "modular will replace monolithic" thesis.

How the Cross-Subsidy Actually Works

The mechanics of the pooled-rate cross-subsidy are worth tracing carefully because they're what make the architectural critique precise rather than rhetorical.

In Aave V3, a WETH supplier earns a single supply APR that's determined by the WETH market's overall utilization. The borrowers in that market are a mix: some are borrowing WETH against rsETH at 95% LTV (a high-leverage LST carry trade), others against wstETH at 93% LTV (a similar but lower-yielding LST loop), still others against weETH (LRT, with EigenLayer slashing risk on top), and a small minority against non-LST collateral like ETH itself or stables. Each of those borrowers pays the same WETH borrow APR per unit of utilization.

The economic substance is that the WETH supplier's capital is being deployed into the highest-leverage strategy first, because that strategy is the most rate-insensitive borrower. A WETH borrower earning 12% on a leveraged LST loop will pay any reasonable WETH borrow rate to keep the trade on. A WETH borrower with a less-levered structural reason to be short ETH will exit at lower rate levels. So utilization in WETH naturally concentrates toward the high-leverage segment, which is exactly the segment that carries the most tail risk. The supplier earns the same rate as if they were funding a diversified book.

The April incident made this concrete. When Aave's Risk Steward cut the WETH Slope2 rate from 10.5% to 3% during the crisis (to prevent a rate spike from accelerating bad debt), the immediate effect was to subsidize the looped LST holders even further: their funding cost dropped exactly when their position was under stress. Stable suppliers who borrowed dollars against now-illiquid aWETH for synthetic exit pushed stablecoin utilization to 100% with rates spiking toward 15%, displacing risk into a different market without eliminating it. Emergency parameter changes contained immediate contagion but redistributed it.

Aave V4 as the Structural Response

Aave's own answer to this critique is V4. The V4 design includes two architectural changes that directly target the cross-subsidy problem:

If V4 delivers on risk-premium pricing in production, it removes the cross-subsidy that made the rsETH incident's ETH depositor impact so broad. The structural ETH supplier would still face tail risk on LST concentration, but they would be compensated for it through a higher rate, and the highest-risk borrowing concentration would migrate into its own isolated spoke. Execution on V4 migration becomes the critical forward variable for the bull-case framing of Aave's post-incident moat repair: faster migration equals faster moat repair, slower migration equals continued exposure to the V3 structural critique.

The V4 mainnet launched March 30, 2026, just under three weeks before the April 18 incident. The migration of high-volume markets from V3 spokes to V4 spokes is ongoing as of this writing. Whether the migration absorbs enough of the LST-concentration risk before the next stress event is a question only time and a second stress test can answer.

The Modular Alternative: Why Morpho Wasn't Exposed in the Same Way

Morpho's architecture takes the structural answer to the pooled-rate cross-subsidy further. Morpho Blue is a permissionless lending primitive where each market is defined by five immutable parameters: collateral asset, loan asset, oracle, liquidation LTV, and interest rate model. Markets do not share liquidity by default. Curators (Steakhouse, Gauntlet, Re7, others) build vaults that allocate capital across Morpho markets per their own risk frameworks, and depositors choose which curator's vault to use.

This architecture has two consequences directly relevant to the April-event analysis:

• Cross-subsidy is structurally absent. A WETH supplier in a Morpho vault is funding a specific cohort of borrowers determined by the curator's allocation choices. If the curator is allocating only to non-LST collateral types, the supplier is not financing levered LST loops at all. If the curator is allocating to LST loops, the supplier knows which loops at what LTV and can choose accordingly. The pooled-rate compression that bit Aave V3 doesn't happen here because there is no shared rate across collateral classes.
• Lenders required higher APRs for equivalent LST-backed loops. On Morpho, an LST-collateralized loop pays a higher rate than a diversified market with the same loan asset, because the curator and the rate model price the segment-specific risk. This naturally limited the scale of LST loops on Morpho relative to Aave V3, which is structurally why Morpho was less exposed to the April outcome at the time it happened.

The cost of this architecture is the cost of any modular system: less shared liquidity per market, more curator decisions for the user to evaluate, more discrete risk surfaces. Morpho's curator concentration (Steakhouse + Gauntlet hold ~73% of vault TVL combined) creates its own risk class. The April-event lesson that mattered for Morpho was the March 22 Resolv / Morpho cascade, where an oracle-latency failure on a synthetic stable plus public-allocator stress non-conditionality drove ~$3.8M in Morpho-market bad debt. Different stress shape, same architectural lesson: curator-level discipline is the differentiator, not the pooled vs. modular split alone.

What This Synthesis Actually Argues

The conventional read on the April rsETH event is "Aave got hit, the bridge had a single-DVN config, the smart contracts worked." All three of those statements are accurate and incomplete. The structural read is more interesting:

Architecture is necessary but not sufficient.

Pooled-liquidity lending is not architecturally broken. SparkLend uses the same architecture as Aave and was the largest beneficiary of the post-incident flows. The architecture sets the ceiling on how badly cross-subsidy mispricing can scale; underwriting discipline determines where on that ceiling a given protocol actually lives. A protocol with the right discipline on a "wrong" architecture outperforms a protocol with weak discipline on a "right" one at the moment of stress.

Risk-premium pricing is the structural fix.

If a protocol prices borrows by the actual risk class of the collateral, and lenders earn rates that reflect what their capital is funding, the cross-subsidy is removed at the source. V4's Risk Premiums architecture is one expression of this; Morpho's per-market rate models are another. The conceptual answer is the same regardless of which architecture implements it: depositors should not be opted into financing concentrated tail risk at a rate calibrated for a diversified book.

The April event was a public test of an existing structural exposure.

The 98.5% LST share of WETH collateral was visible on-chain for months. Anyone who computed it before April could have surfaced the cross-subsidy critique. The reason the critique only landed publicly after the rsETH incident is that the visible material event made the abstract risk concrete. For TI's framework, this implies that the relevant on-chain risk metric is the share-of-borrow-collateral that comes from a single tail-risk asset class, not the overall TVL or the headline LTV. Tracking that metric across the major DeFi lending venues is the leading indicator of which protocol gets hit first in the next analogous event.

What to Watch

Five concrete watch items that follow from the structural analysis:

1. Aave V4 migration share by TVL. The cleanest single indicator of whether the protocol's structural response is landing. If V4 absorbs the high-volume markets (especially WETH) within 6 months of mainnet, the cross-subsidy critique is being addressed at the framework level. If migration stalls and the bulk of TVL stays on V3, the structural exposure persists despite V4's architectural availability.
2. Aave WETH LST share trend. The 98.5% number is the reading at the time of the incident. If the share drops materially (below 80%) without a corresponding drop in TVL, it means the protocol has rebalanced toward more diversified borrowing. If the share stays above 95%, the structural concentration that bit in April is unchanged.
3. SparkLend deposit-share trajectory. The post-incident +$1.8B inflow surge was a reaction to one event. Whether SparkLend retains those flows over multiple cycles is the test of whether "underwriting discipline" actually compounds into a durable moat or whether it's a temporary preference reversion that fades when crisis-memory fades.
4. The next protocol to deprecate an LST proactively. Spark deprecated rsETH in January 2026, three months before the event made it correct. Watching which other major lending venues take similar pre-incident actions on weETH, EBTC-class assets, or any other tail-risk collateral is the leading indicator of which protocol earns the next round of post-incident inflows.
5. Cross-architecture stress event. The April incident hit pooled lending hardest. The next analogous structural shock might land on the modular architecture (curator concentration, oracle-latency cascades like the Resolv event). The cross-architecture comparison only completes when both architectures have absorbed a public stress at scale.

Closing Thoughts

The April rsETH event is being read by most coverage as a bridge security incident with downstream impact on Aave. That reading is true and underclaims the importance of the on-chain analysis that surfaced afterwards. The bridge attack was the ignition; the underlying fuel was the LST-concentration mispricing in pooled lending. The same fuel exists today, less visibly, in any pooled-liquidity money market that has not addressed cross-subsidy in its rate framework. The next stress test will look different in its surface mechanics (different bridge, different protocol, different collateral) but the structural shape will be recognizable.

For TokenIntel users tracking DeFi lending exposure, the practical translation is: read aggregate TVL and headline LTV with the share-of-borrow-collateral-by-class number sitting next to it. A protocol with $20B TVL and 98% of one borrow market funded by a single collateral class is structurally exposed in a way the headline numbers do not surface. A protocol with $5B TVL and a diversified-by-class borrow book is structurally less exposed for the same reason. The relevant metric was hard to find before April; it will be common terminology by the end of 2026.

The architecture choice between pooled and modular lending is genuine and the tradeoffs are real. The April event did not settle that question. What it did was disqualify the version of the pooled architecture that lets cross-subsidy mispricing scale unchecked. Aave V4 is one structural answer to that disqualification. The market will judge whether the answer is enough.