The leveraged ETH restaking carry trade — once one of DeFi's most popular institutional strategies — has officially entered negative carry territory. As of February 20, 2026, the debt-weighted average cost of borrowing ETH stands at 3.40%, exceeding the staking yields offered by every major liqui...
"The depth of the drawdown and the degree of leverage reset have made the current price washout increasingly attractive for building positions on a one- to two-year view." — Matthew Sigel, Head of Digital Asset Research, VanEck
The leveraged ETH restaking carry trade — once one of DeFi's most popular institutional strategies — has officially entered negative carry territory. As of February 20, 2026, the debt-weighted average cost of borrowing ETH stands at 3.40%, exceeding the staking yields offered by every major liquid staking token (LST) and liquid restaking token (LRT) tracked by Sentora's coverage universe.
This is not a temporary inversion. The 7-day moving average of ETH borrow rates has climbed steadily since late January 2026, rising from approximately 2.0% to over 3.4% by mid-February, while staking and restaking yields have remained flat or declined. A standard 5x leveraged wstETH position now generates an estimated -1.90% annualized return. During peak utilization events — when borrow rates spike above 6% — annualized losses widen dramatically to -15% to -18%.
The structural inversion has arrived alongside two severe WETH utilization spikes on Aave V3, with the second pushing utilization to 99.85% on February 15 — the most extreme lending market stress event in DeFi since the Curve pool crisis. Meanwhile, EigenLayer's EIGEN token has collapsed 96% from its all-time high, restaking TVL has contracted from $20 billion to roughly $7 billion, and the broader February deleveraging event has cleared $3–4 billion in liquidations across crypto markets. The restaking carry trade isn't just underperforming — it's becoming a source of systemic fragility.
The restaking carry trade was DeFi's answer to the Japanese yen carry trade — a leveraged arbitrage strategy that borrows at a low rate to invest in a higher-yielding asset. In this case, institutional and sophisticated DeFi participants would:
At its peak, this recursive leverage strategy promised compounded yields of 8–15% APY when native staking yield was around 3.3%, additional restaking rewards provided 2–4%, and ETH borrow costs remained at 1.5–2.5%. The spread was positive, the leverage was available, and billions flowed in.
Ether.fi alone accumulated $5.7 billion in TVL. EigenLayer surpassed $20 billion. The liquid restaking token market approached $8 billion. It became arguably the single most capital-efficient trade in DeFi — until it wasn't.
The inversion stems from a fundamental asymmetry: staking yields have a natural ceiling, while borrowing costs are theoretically unbounded.
Ethereum's consensus-layer staking yield is determined by protocol emissions and validator participation. With over 34 million ETH staked (approximately 28% of total supply), validator yields have compressed to roughly 2.3% on Lido and 3.3% for solo validators. These figures move slowly. There is no mechanism for staking yields to spike in response to market demand.
Borrowing costs, by contrast, are governed by lending market utilization — and utilization has surged. As DeFi lending markets matured through early 2026, total active loans across Ethereum-based protocols surpassed $28 billion. Aave, Compound, and Morpho all experienced elevated utilization rates, driven not just by restaking carry traders but by broader market dynamics including collateral liquidations during the February crash and increased demand for ETH borrowing for short-selling.
The result: ETH borrow rates climbed from ~2.0% in late January to 3.40% by February 20 as a debt-weighted average — with peaks far exceeding that level during stress events. Meanwhile, the all-in yield from staking + restaking has not meaningfully increased, creating a negative spread that punishes every leveraged position in the ecosystem.
The most dramatic evidence of the carry trade's structural breakdown came from Aave V3's WETH reserve, which experienced two extraordinary utilization spikes in February 2026:
First Spike — February 6-7, 2026:
Second Spike — February 14-15, 2026:
At a 9.01% borrow rate, a 5x leveraged restaking position was hemorrhaging approximately -30% annualized — before accounting for any price depreciation in the underlying collateral. These weren't theoretical risks; they were realized losses for every position that remained open through the spike.
The restaking sector's flagship protocol tells the story in its token price:
The fundamental problem is structural, not merely cyclical. EigenLayer's TVL contracted from over $20 billion at its peak to approximately $7 billion by late 2025, a 65% decline that accelerated after the protocol activated slashing on April 17, 2025. The slashing launch triggered a rational repricing of interconnected risk — once economic penalties became real, the risk-adjusted return of restaking collapsed.
Adding to the pressure: on February 1, 2026, 36.8 million EIGEN tokens were unlocked, worth approximately $12.3 million at the time. This represented over 2% of total supply and 6.75% of the then-market capitalization — a dilution event that overwhelmed thin market demand. EigenCloud, the protocol's AI-focused product extension, generates only $5.31 million per month, and there is still no live mechanism to accrue value to EIGEN holders through fee-sharing or buybacks.
An Incentives Committee was announced for early 2026 to implement ELIP-12, which would direct emissions and route fees to EIGEN buybacks. But the proposal remains pre-implementation — a roadmap item while the token trades at pennies.
The most dangerous aspect of the restaking carry trade is not that it turned negative — it's that unwinding it is structurally constrained.
Liquidity-to-TVL ratios are dangerously thin. For certain liquid restaking tokens, on-chain secondary market liquidity represents a fraction of outstanding supply. WeETH, ether.fi's liquid restaking token, has an estimated liquidity-to-TVL ratio of approximately 0.035%. This means that large exits through secondary markets would trigger severe slippage — potentially double-digit percentage losses on top of the negative carry.
Native redemptions are slow. Exiting through the protocol requires Beacon chain withdrawals, which currently take approximately 9 days under normal conditions. During the July 2025 deleveraging event, the validator exit queue expanded to 19 days — the longest in Ethereum's history. Borrowers remain exposed to fluctuating interest rates throughout this multi-week exit window. If borrow rates spike during the exit period, losses compound on positions that cannot be closed.
Concentration risk amplifies the problem. A large portion of ETH borrowing across Aave, Compound, and Morpho appears concentrated in carry trade strategies. If multiple institutions attempt to unwind simultaneously, utilization rates spike, borrow costs escalate further, and what begins as orderly deleveraging becomes a self-reinforcing liquidation cascade.
The restaking carry trade's breakdown does not exist in isolation. It intersects with the broader February 2026 deleveraging event, which has cleared $3–4 billion in liquidations across crypto markets, compressed BTC futures open interest by 20% (from $61 billion to $49 billion), and pushed Bitcoin below $61,000.
Pathway 1: Lending market utilization spirals. As carry traders unwind, they must repay borrowed ETH. Selling LRTs for ETH on thin markets depresses LRT prices, triggering additional collateral calls, which force more selling. Aave's WETH reserve has already hit 99.85% utilization once — a second event during synchronized unwinding could push borrow rates above 15%, making any leveraged position untenable.
Pathway 2: LRT depeg cascades. Liquid restaking tokens are designed to trade at or near par with ETH. Under severe redemption pressure, LRTs can depeg — trading below the value of their underlying staked ETH. A depeg triggers collateral revaluation across every lending protocol that accepts LRTs, potentially cascading into liquidations far beyond the restaking sector.
Pathway 3: Validator exit congestion. If restaking positions unwind at scale, the resulting validator exits create queuing delays on the Beacon chain. During the delay, exiting validators continue to bear staking risk without earning adequate returns, while the queue itself signals network stress — potentially depressing ETH price further in a reflexive loop.
Ethereum co-founder Vitalik Buterin has repeatedly warned that restaking protocols could pose "significant systemic risk" to the network. The February 2026 data suggests those warnings were prescient.
The ETH restaking carry trade has turned structurally negative, with borrow costs (3.40%) exceeding all-in staking and restaking yields across every major LST and LRT as of February 20, 2026.
Aave V3's WETH reserve hit 99.85% utilization on February 15, with borrow rates spiking to 9.01% — a stress event that exposed limitations in Aave's Slope2 Risk Oracle, which failed to escalate for 43 hours during extreme conditions.
EigenLayer's EIGEN token has collapsed 96.5% from its all-time high to $0.20, with TVL contracting from $20B+ to ~$7B after slashing activation. No live value-accrual mechanism exists for token holders.
Liquidity-to-TVL ratios for major LRTs are dangerously thin (as low as 0.035%), meaning large-scale exits through secondary markets would trigger severe slippage and potential depegs.
The restaking carry trade's breakdown is not isolated — it intersects with the broader February deleveraging event ($3–4B in liquidations) and creates three distinct contagion pathways through lending markets, LRT pricing, and validator exit queues.
The fundamental asymmetry remains unresolved: staking yields are bounded by protocol emissions, while borrowing costs are theoretically unbounded during high utilization — a structural mismatch that makes leveraged restaking strategies inherently fragile.
The restaking carry trade was DeFi's most popular institutional strategy for 18 months. At its zenith, it represented everything the ecosystem aspired to be: capital-efficient, composable, and yield-generating without direct token emission subsidies. It attracted billions in TVL, spawned an entire liquid restaking token economy, and became the primary use case for EigenLayer — a protocol once valued in the tens of billions.
Now the trade is broken, and the unwinding has begun.
The February 2026 data reveals that the strategy's profitability was never structural — it depended on a narrow borrow-staking spread that existed only under favorable utilization conditions. Once those conditions deteriorated — driven by the broader market crash, increased ETH borrowing demand, and lending protocol maturation — the carry turned negative and the entire architecture became a liability.
What makes this moment dangerous is not the losses on existing positions, which are manageable. It is the potential for synchronized deleveraging in a system where exit liquidity is thin, redemptions are slow, and borrowing costs escalate precisely when participants need to leave. The restaking stack has created a chain of interconnected financial claims — ETH → stETH → eETH → collateral → borrowed ETH — where stress at any link propagates across all of them.
The economic-value-first lens reveals a familiar pattern: another DeFi vertical that attracted tens of billions in capital while generating minimal fee revenue, sustained primarily by token incentives and speculative positioning rather than durable cash flows. EigenLayer's $7 billion in remaining TVL produces a fraction of the revenue needed to justify its capital base. The EIGEN token, stripped of its airdrop-fueled premium, has repriced to reflect this reality.
The restaking experiment is not over. But the carry trade that funded it is. What comes next will determine whether restaking evolves into genuine shared-security infrastructure — or joins the long list of DeFi innovations that created more leverage than value.