Borrowing Crypto in 2026: What Loans on Aave, Morpho and Spark Taught Me About Health Factor

Three years of borrowing against ETH have taught me one thing: the loan doesn't care about my thesis. I borrow stablecoins against my ETH to fund positions without selling, but the protocol holds exactly one opinion — a number called health factor. Above two, the position is mine to manage. Sliding toward one, it belongs to whichever liquidation bot gets there first.

This is my borrower's playbook for 2026, built around the positions I monitor on Aave, Morpho and Spark. It covers health-factor math, LTV vs liquidation threshold, a $20,000 ETH worksheet I reuse, pooled versus isolated markets, and the April 2026 rsETH event that changed several of my rules. I won't invent a dramatic personal liquidation — it didn't happen — but watching roughly $193 million of borrowed WETH leave Aave in 46 minutes is an education you can take for free.

The one number I check every morning

The health factor (HF) answers one question: can my collateral still cover my debt after the protocol's safety margin? Per Aave's own FAQ, it is collateral value multiplied by the weighted-average liquidation threshold, divided by total debt. Morpho uses different terminology — its liquidation docs define a position as healthy above HF 1 and liquidatable once it meets the market's LLTV.

Two forces move that number against you around the clock: falling collateral shrinks the numerator, and accruing borrow interest grows the denominator. A loan opened at 1.4 on Monday can drift to 1.25 by Friday even if price does nothing, so the displayed HF is a live reading, not a setting. Liquidation is permissionless — no call, no grace period — and the liquidator takes collateral at a discount, around 5% on large-cap markets.

LTV is the doorway; the liquidation threshold is the cliff

Borrowers consistently confuse these two, and the gap between them is where careless loans die.

• Loan-to-value (LTV) caps what you may borrow when opening the position. If ETH has an 80% LTV, the app lets you borrow up to $0.80 per dollar of ETH.
• Liquidation threshold (LT) is where liquidation becomes legal. It always sits above the LTV — say 83% — and it, not the LTV, feeds the health factor.
• The buffer between them is the protocol's room for price noise. Borrowing at the LTV cap parks you right above the cliff with only that thin gap.

Correlated modes compress that buffer on purpose. Aave's E-Mode lets ETH-correlated collateral borrow ETH assets at elevated ratios — in April, rsETH in E-Mode ran a 93% LTV and a 95% threshold: efficient for looping, brutal when the collateral misbehaves. Spark's liquidation glossary uses the same formula; learn it once and every Aave-derived market reads identically.

My $20,000 ETH worksheet

I run this before every loan. Illustrative assumptions: $20,000 of ETH, 80% LTV, 83% liquidation threshold, USDC debt with zero price drift. Risk-adjusted collateral is $20,000 × 0.83 = $16,600.

Read the last row twice. The app happily lends $16,000 because that is what 80% LTV permits — and a routine 3.6% ETH candle then brings the liquidators. The max-borrow slider measures permission, not safety. My rule is to borrow no more than half the displayed maximum against volatile collateral, here $8,000–10,000 at HF 1.7–2.1, and to set alerts at the liquidation price shown in the UI.

The health-factor bands I actually use

Thresholds vary with collateral and sleep schedule, but these are my bands:

Three protocols, three risk machines

Aave is a pooled money market: every whitelisted asset shares one deep liquidity layer, with governance-set parameters, supply and borrow caps, a Guardian that can freeze reserves, and a Risk Steward that tunes rate models. The depth is why I borrow stablecoins there. The tradeoff is shared fate — one risky collateral listing can pressure the pool I borrow from.

Morpho Blue inverts that. Each market is isolated — one loan asset, one oracle, one rate model and an immutable LLTV chosen by its creator — so a bad market's losses stay put, and curators do the vetting Aave governance does. The tradeoff is on me: a high-LLTV market or careless curator is my mistake, and thin markets can be illiquid exactly when I want out. For side-by-sides I keep our Aave vs Morpho and Aave vs Spark comparisons bookmarked.

Spark runs Aave-derived markets with a conservative, Sky-aligned asset list and deep USDS/ETH liquidity: fewer experiments, faster delists. April proved the value — Spark deprecated rsETH as collateral in late January 2026, three months before the exploit, and was untouched while drawing outflows. Our ranking of the best lending protocols tracks how these tradeoffs evolve.

April 18, 2026: the day a pool borrowed itself dry

This event belongs in every borrower's risk folder. A forged message on Kelp DAO's LayerZero bridge — a 1-of-1 verifier — released 116,500 rsETH, roughly $292 million, with no backing. The attacker parked it as collateral on Aave, Compound and Euler and borrowed real assets. Per the LlamaRisk incident report, 89,567 rsETH went into Aave V3 alone; seven addresses borrowed 82,650 WETH plus 821 wstETH, about $193 million, at up to 93% LTV in E-Mode, with health factors of 1.01 to 1.03.

The oracle kept quoting rsETH's canonical ETH rate, so automatic liquidations never triggered; the tokens were forged, not mispriced. What hit ordinary borrowers was the liquidity side. As news spread, everyone ran for the WETH door: Core WETH available liquidity fell from $689 million at 17:00 UTC to $1.5 million by 19:00, utilization pinned at 100%, and reported WETH borrow rates jumped from roughly 2.3% to about 8.7% APR — inside LlamaRisk's documented pre-adjustment range of 8.5–10.5% at full utilization. Suppliers couldn't withdraw, rates multiplied, and the Risk Steward later cut Slope 2 while guardians froze reserves.

The LlamaRisk scenarios are what I quote to anyone who thinks over-collateralization guarantees safety: $123.7 million if the 15.12% depeg was socialized across all rsETH, and $230.1 million if losses stayed on L2 rsETH, backed at only 26.46% after the drain. The propagation mechanics are the same ones in our liquidation cascade explainer. My borrower takeaways: pooled contagion is real, maxed-out correlated loans are a non-starter, and "protocol solvent" does not mean "I can withdraw today."

The second loan: rate and utilization risk

Every variable-rate loan carries a quieter second loan: the risk that your rate doubles when the pool gets crowded. Rates follow a kinked curve — gentle below optimal utilization, near-vertical above it — and April was that curve in fast-forward, though ordinary versions happen whenever a looping trade goes viral. I read current utilization and the rate at 100% before borrowing, and avoid leverage in reserves already above the kink; the math is in our guide to interest-rate models and the utilization kink.

Looping turns this into a reflexive loop — borrow WETH against an LST, buy more LST, redeposit, repeat — farming the spread between staking yield and borrow cost. It is how ETH-correlated collateral came to dominate Aave's WETH book, and why one LRT shock nearly drained the reserve. I size looped positions smaller and start them at HF 2.2+, never the 1.03 the attacker used; our liquid staking looping guide shows when the carry clears the risk.

My monitoring rules for positions I monitor

• Open at HF ≥ 2; for looped or LST collateral, 2.2. Accrued interest and a normal wick should never threaten the loan.
• Never borrow above half the displayed max against volatile collateral. The slider measures permission, not safety.
• Two alerts per position: HF 1.5 ("pay attention") and HF 1.3 ("act now"). Plus a price alert at the liquidation price.
• Keep rescue funds on the same chain. An optimistic-rollup withdrawal can take seven days; a stablecoin top-up sitting on L1 is useless during an L2 liquidation.
• Check utilization and the max borrow APR before opening. Above the kink, I either wait or use a quieter market.
• Know the collateral's history. I avoid freshly bridged restaking tokens and recent freeze or delisting drama; conservative collateral is a feature.
• Treat governance alerts as position news. Freeze notices, rate-model changes and steward votes are the canary, not background noise.
• De-risk in calm markets, not during the spike. Partial repayment at HF 1.5 costs nothing; amid 100% utilization it may be impossible.

Borrowing well is mostly about staying bored: conservative entry, a number checked daily, a rescue plan funded early. The protocols are impressive; risk management is the borrower's job.

Sources and further reading

• Aave FAQ — official definition and formula for health factor and liquidation threshold.
• Morpho Docs — Liquidation — LLTV, health factor and liquidation incentive mechanics in isolated markets.
• Spark Glossary — Liquidation (DeFi) — worked health-factor example and the liquidation process.
• LlamaRisk / Aave service providers — rsETH Incident Report (April 20, 2026) — attacker positions, defensive actions and the $123.7M–$230.1M bad-debt scenarios.

Frequently asked questions

What is a safe health factor for a crypto loan?

My rule is to open every volatile-collateral loan at a health factor of at least 2, treat 1.5 as the monitoring line where I start checking the position daily, and act (repay or add collateral) by 1.3. A health factor below 1 means the position is liquidatable. The attacker accounts in the April 2026 rsETH event were running health factors of just 1.01 to 1.03, which is why a single collateral reprice would have liquidated them instantly.

What is the difference between LTV and liquidation threshold?

LTV is the maximum you can borrow when you open the position; the liquidation threshold is the collateral ratio at which liquidators can seize your collateral. The threshold is always higher than the LTV, and the gap is your buffer. On Aave-style protocols the health factor is calculated with the liquidation threshold, not the LTV: health factor equals collateral value times liquidation threshold divided by total debt. Morpho calls the equivalent parameter LLTV, the liquidation loan-to-value.

How close did the April 2026 rsETH exploit come to creating Aave bad debt?

The Aave service providers' LlamaRisk incident report modeled bad debt of $123.7 million under uniform socialization of the 15.12% rsETH depeg, and $230.1 million if losses were isolated to L2 rsETH, which was backed at only 26.46% after the bridge drain. The attacker had supplied 89,567 rsETH and borrowed 82,650 WETH plus 821 wstETH on Aave V3 at up to 93% LTV in E-Mode, with health factors of 1.01 to 1.03.

Is borrowing on Morpho safer than borrowing on Aave?

Different, not strictly safer. Aave is one pooled market where every listed collateral shares liquidity, so one bad asset can pressure the whole pool, as rsETH did in April 2026, but it also offers deep liquidity, governance risk parameters and guardian freezes. Morpho Blue uses isolated markets with immutable LLTVs, so a failure is confined to that market, but risk shifts onto the oracle, interest-rate model and curator vault you choose, and liquidity is shallower. Spark runs conservative Aave-derived markets and had deprecated rsETH as collateral in January 2026, which is why it was unaffected by the event.

What happens to my loan if a lending pool reaches 100% utilization?

Two things: borrow rates climb above the kink toward the rate model's ceiling, and suppliers may temporarily be unable to withdraw until utilization falls. During the rsETH event, Aave's Core WETH available liquidity fell from $689 million to $1.5 million in two hours, utilization hit 100%, and reported WETH borrow rates moved from roughly 2.3% to about 8.7% APR before the Risk Steward cut Slope 2. Existing safe borrowers were not liquidated by this alone, but their interest costs jumped and rescue liquidity vanished.

Keep reading: Interest-Rate Models and the Utilization Kink, The DeFi Liquidation Cascade Explained and Liquid Staking Looping.