Aave's two tracked Ethereum V3 stablecoin backstop pools contracted by $12.4 million between Sept. 17 and Oct. 17, 2026, even as the extra reward rates paid to their depositors increased. The $USDC and $USDT positions together held $116.09 million at the later observation, down 9.67% from $128.52 million a month earlier.

Both positions sit within Umbrella, Aave's mechanism for deploying staked assets to absorb lending shortfalls. Depositors continue to earn the underlying lending yield and collect supplementary rewards in exchange for accepting the possibility that their stakes will be drawn upon to cover losses.

The figures represent dollar-valued stakes. Because their movement reflects asset valuation as well as changes in the amount deposited, a falling balance alone cannot distinguish withdrawals from valuation effects.

Tracking the decline

Multiplying each position's tracked dollar value by its displayed extra reward rate implies roughly $1.994 million annualized at both endpoints. A higher percentage rate therefore coincided with less tracked protection capital, while the indicative aggregate reward total barely moved.

DefiLlama's $USDC history shows a decline from $58,739,617 to $53,437,993, while its $USDT history records a fall from $69,783,937 to $62,655,694. Those are reductions of $5.30 million and $7.13 million respectively, together totaling $12,429,867.

The comparison uses the Sept. 17, 2026 observation at 23:01 UTC and the Oct. 17, 2026 observation at 15:01 UTC — timestamps slightly less than 30 days apart. The window opens before the Sept. 25 TokenLogic reward proposal described below and closes after it. The rates referenced below are the additional reward components displayed at those observations.

The tracked positions are wrapped-aToken Umbrella stakes on Ethereum V3. DefiLlama labels them stkwaEthUSDC.v1 and stkwaEthUSDT.v1, and their token identities match the current Aave address book.

The $USDC extra rate rose by 0.16 percentage points and the $USDT rate by 0.17 points. Lending yield moved independently: the $USDC base rate climbed from about 3.60% to 3.77%, while the $USDT rate increased from about 3.02% to 3.12%. The base component reflects lending returns, whereas the extra component compensates for staking and slashing risk; only the latter belongs in this safety-incentive calculation.

How rates can rise without larger payouts

For each observation, the calculation multiplies the dollar value by the displayed extra reward percentage, divides by 100, and sums the two stakes. The September values produce $1,993,923.12 annualized; the October values produce $1,993,859.28. Both round to $1.994 million.

These amounts are indicative annualized run rates derived from point-in-time snapshots. Actual accrual and payments depend on the path of rates, stake values, reward configuration and participation over time.

Umbrella's reward-controller documentation describes a mechanism through which the percentage rate can increase without a larger aggregate distribution. Its emission curve ties the amount of rewards distributed to the amount staked relative to a configured target.

Aggregate emissions peak at the target. Between the target and 120% of target, emissions taper toward 80% of that maximum. At or above the 120% boundary, aggregate emissions are flat — meaning fewer stakes can share the same reward stream, lifting the rate per unit of capital. Below target, the curve raises rewards per unit to encourage participation, while aggregate emissions remain beneath their peak. The plateau thus offers one route by which a smaller denominator can command a higher percentage rate. Establishing whether it applied to these pools at both observations would require matched contract settings and balances; the histories alone leave that specific explanation unresolved.

A Sept. 25 proposal from TokenLogic sought to renew the allowances used to pay Umbrella rewards while leaving emission rates, target liquidity, cooldown parameters, deficit offsets and the Dec. 2, 2026 distribution end unchanged. The proposed allowances covered rewards already accrued but unclaimed — including amounts owed to former stakers — plus remaining emissions through that distribution end, and they replaced the leftover allowances. Reading those amounts as a fresh annual incentive budget would therefore blend past obligations with prospective payments. If adopted as proposed, the configuration would hold fixed through the Dec. 2, 2026 end date, giving later readings of the two pools a stable parameter set against which to compare balance and rate movement.

What shrinking balances mean for protection

Aave's documentation limits staking risk to the relevant asset and network: $USDC stake covers $USDC deficits, while $USDT stake serves its own reserve. The combined total measures the scale of the two positions, but each reserve's protection depends on its own allocated stake.

Slashing follows a realized reserve deficit that exceeds the applicable deficit offset. LlamaRisk's coverage explanation describes that offset as a DAO-funded first-loss layer deployed before Umbrella stake is touched.

Exit rules also matter to stakers weighing the additional return. Aave documents a 20-day cooldown followed by a two-day withdrawal window. Rewards keep accruing during cooldown, and the stake remains exposed to slashing while the holder waits. Those rules also frame month-over-month comparisons: stakes that entered cooldown shortly before the October snapshot would still count toward the measured balance, and any withdrawal completed between the two observations would have required an exit initiation at least 20 days earlier.

The DefiLlama adapter values provider-reported total assets using an associated price. Wrapped aTokens carry an exchange rate, so valuation forms part of the measured balance.

The same distinction applies to Aave's separate V4 coverage plans. CryptoSlate's September reporting on the V4 proposal concerned proposed targets for different protection modules, which are distinct from the observed V3 balances.

At the October observation, extra rates were higher while the two tracked balances were smaller: the incentive offer and the capital behind it had moved in opposite directions.