The chart didn't lie. It just showed a number the bull market narrative prefers to ignore: 85%. Nobody wants to hear that during a bull market.
Dune Analytics, commissioned by 1inch, sliced 2026 H1 on-chain data across seven chains. The finding lands like a hammer on the concentrated liquidity thesis. Four out of five dollars locked in CLMM positions never earned a single fee. 29.5% of all deployed capital sat entirely outside its active range โ zero fee accrual, full impermanent loss exposure. Abandoned fees across the seven chains? $150 million. Annualized.
That's not friction. That's a structural leak in the most celebrated capital-efficiency innovation of the last DeFi cycle.
I've spent enough years watching order books and LP positions to recognize what this pattern actually is. It is not market noise. It is a usage failure. The model is fine. The users โ and the interfaces guiding them โ are the bug.
Concentrated liquidity arrived in 2021 with Uniswap v3, promising a revolution: instead of parking capital across an infinite price curve, LPs could concentrate dollars into tight bands around the current price. Same capital, 10x to 100x more efficiency. Uniswap v4 extended the concept with hooks โ programmable liquidity that executes strategies directly inside the pool. The theory was elegant. Capital would flow to where it worked hardest.
Elegant theory. Messy reality.
The Dune dashboard 1inch commissioned covers the 2026 H1 period, and the chain data tells a different story than the whitepapers. Across Ethereum and the L2s, the median LP position behaves nothing like the optimized market-making strategy the model assumes. Most LPs set a range, pray, and disappear. When price exits their band โ and it always does โ the position sits dormant. No fees. No rebalancing. Just impermanent loss compounding quietly.
I bought the pixel, not the promise when I tested v3 ranges back in 2021. Turns out most retail LPs bought the promise without ever watching the pixel.
Let's decompose the headline, because aggregate stats hide the real pathology.
First, the 29.5% figure โ capital entirely outside the active range. This is the pure-loss category. No fee income. Full IL exposure. These LPs are effectively paying to provide liquidity: their capital is locked, depreciating relative to holding the assets outright, while the pool does nothing for them. On-chain evidence suggests most of these positions trace back to default parameter settings. Retail LPs open a position at the current price tick, pick a โsafeโ width, and never touch it again. The market moves. The range breaks. The LP forgets.
This is where my execution-risk background kicks in. A market maker who leaves a quote alive overnight without monitoring deserves what they get. A retail LP who does not even understand the quote is live deserves better protocol design. The gap between those two realities is the 85%.
Second, the fee math makes the inefficiency even uglier. If roughly 15% of capital captures virtually all of the $150 million in fees, the active dollar is working about 5.7 times harder than the theoretical average. That is not a marginal optimization problem. That is a signal to every quantitative fund in crypto that the spread between lazy capital and active capital is the widest it has been since the DeFi summer of 2020.
Third, the distribution inside that active 15% is unlikely to be democratic. My read of the data suggests a brutal Pareto split: a small cohort of professional market makers and sophisticated bots harvests the majority of fees, while a long tail of retail positions dilutes the averages. The โaverageโ LP is not a professional market maker. The average LP is someone who watched a tutorial, clicked โAdd Liquidity,โ and went back to their day job.
The yield farming experiment I ran in 2020 taught me this lesson early. I spun up local nodes, verified transaction finality, and measured gas costs manually. Most DeFi users will never do that. They will never rebalance. They will never check whether their tick range is still alive. The result: their capital becomes exit liquidity for everyone else.
Every candle tells a story of fear. In CLMM, every out-of-range position tells a story of neglect.
Now the uncomfortable part.
This research was commissioned by 1inch. 1inch is an aggregator. Their entire business model depends on routing trades through the most efficient liquidity sources available. When they publish a study showing direct DEX market making is catastrophically inefficient, the implicit pitch is obvious: use the aggregator, not the pool.
I don't discount the data. I do discount the framing.
The methodology has transparency gaps. What exactly counts as โidleโ? Does the dashboard classify both tails outside the range as idle, or does it make finer granular distinctions? What criteria selected the seven chains? These definitions materially shift the 85% number. Without the underlying SQL and parameter definitions publicly audited, the headline is a directional signal, not a scientific conclusion. No peer review. No independent reproduction. Treat it accordingly.
And here is the counter-intuitive angle: idle liquidity is not purely waste. It is a free option. LPs who set wide ranges and walk away are effectively selling tail-risk protection to the market. They earn nothing in quiet periods, but their capital acts as a backstop when price makes a violent move. The $150 million โabandonedโ is the premium they are not charging for that insurance. Calling it pure inefficiency is like calling an options seller stupid for letting theta decay โ it depends entirely on what the seller is hedging.
Liquidity vanishes when the music stops. But the LPs who left their ranges wide open? They are the ones still there when the tape breaks. That has value. It just does not show up in a fee-efficiency dashboard.
Code is law, until it isn't. And right now, the law says $150 million a year is sitting on the table for anyone who can solve range management at scale.
The obvious answer is AI-agent-driven liquidity management โ automated rebalancing bots that keep positions in-range without human attention. I have been backtesting exactly this since early 2025, and the edge is real. But the deeper question is whether protocols should design for the user they have, not the user they want. Dynamic ranges that track volatility. Hooks that self-heal. Defaults that auto-roll instead of rotting.
The $150 million will be harvested. The only open question is whether it goes to the LPs who earned it, or the bots smart enough to take it.