A concentrated market carries concentrated failure modes. In DeFi, a single lending market holds roughly a third of all onchain lending, and a single liquid staking provider holds close to half of its category.
Meanwhile, the habit of building protocols on top of one another means the token issued by one becomes collateral in the next and a trading pair in a third. The failure of a dominant venue therefore does not stay contained within it.
This article sets out where a concentrated and interconnected DeFi breaks. It works through the three ways a dominant protocol can fail, the liquidation cascade that carries the damage outward, the diversification that fails to diversify, the metric that misleads an allocator into the wrong decision, and the structural fragility that concentration cannot remove.
A Single Name's Failure Becomes the Category's Failure
Concentration turns a single-name problem into a category-wide one, because once a dominant protocol's positions are reused across the rest of the stack, a failure at the source reaches everything built on it.
These failures fall into three types, and each has already appeared in some form onchain.
Technical Failure
A contract bug or an oracle error at a dominant protocol expands beyond it. Aave holds around one third of lending TVL and Lido holds almost half of liquid staking capital; and the positions built on both extend across the stack. So, a failure at the source propagates to every venue that relies on it.
The clearest demonstration came from an exchange in 2023, when a bug in a smart-contract compiler used by Curve allowed several of its pools to be drained. The damage did not end at the pools, because the founder of Curve held large loans backed by the protocol's token across multiple lending markets, and as the token's price fell on the news those loans moved toward liquidation.
For a period, a compiler bug in one exchange threatened cascading liquidations across several lenders that had no direct connection to the exploit, which remains the clearest illustration of how a technical failure at one composable venue becomes a solvency question for the protocols built around it.
Economic Failure
Economic failures spread through prices rather than through code, and the sharpest version in a concentrated stack is a liquid staking token losing its peg to the asset it represents.
When a staking token trades below the asset it stands for, every lending market that accepts it as collateral marks those positions down at once, loans backed by it move toward liquidation together, the liquidations force selling, and the selling deepens the very discount that began the cycle.
During the 2022 stress, the largest staking token traded at a discount to ETH as a large forced seller exited, and the discount fed concern across the lending markets that held it. The redemption mechanism introduced the following year tightened the link between the token and the underlying asset, since the token can now be redeemed instead of only being sold, which reduces the structural risk. However, it doesn’t remove it, because redemption runs through a queue and a queue is slower than instant settlement in a fast-moving market.
A large staking token under stress therefore remains a channel through which one category's problem becomes another's.
Why Liquidations Turn a Local Problem Into a Cascade
The liquidation is the mechanism that ties these failures together. Lending markets protect themselves by selling collateral when a loan falls below its required backing, and in calm conditions this proceeds in an orderly way.
Yet, under stress it does not:
A falling collateral price triggers liquidations.
The liquidations sell into a falling market.
The selling drives the price down further and sets off the next round.
Concentration determines how widely that cascade spreads, since when the collateral is a token tied to a dominant protocol, every market that accepts it liquidates at the same time, into the same thin liquidity, against the same falling price.
The design that keeps a single loan safe is the one that synchronises the damage once the collateral is shared across the system, so a varied set of collateral assets would absorb a shock locally while a concentrated one transmits it. This is the precise point at which the efficiency of shared, reusable collateral becomes fragility, through a single asset performing a single job across many venues at once.
Governance Failure
The third type of failure is slower and harder to price. A dominant liquid staking provider concentrates more than capital, since it also concentrates validator influence over the network it stakes into and governance control over its own parameters. At close to half of its category, a single provider's share of the underlying network's staking approaches the level at which questions arise about who can influence the chain's operation, and that concern has driven a long-running debate about whether the largest provider should limit its own growth and about how its governance distributes control between its token holders and its stakers.
None of this requires bad intent to matter, because the concentration of influence is a risk in itself, narrowing the set of parties whose failure or coordination could affect the network that everyone else depends on.
Holding the Top Five Is a Single Correlated Bet
Spreading capital across the five largest protocols in a category has the appearance of diversification while delivering very little of it, because the largest protocol holds between a third and a half of the category and the names tend to move together.
An allocation spread across the top names of one category, or across several categories that all depend on the same market, holds far less diversification than the number of positions implies. Genuine diversification in a concentrated and composable system has to run across categories and across chains, and even that does not fully insulate capital when the whole market turns together, because the correlation that stays invisible in calm conditions is the only correlation that matters in a drawdown.
A Fragmented Category Carries Its Own Risk
Fragmentation lowers a category's headline concentration while leaving its risk in place, relocating that risk from a single protocol to the chains the category spreads across.
A chain-local exchange depends on the chain beneath it, so if that chain halts, as several have for hours at a time, the leading exchange on it halts as well, regardless of how sound the exchange's own code may be. An allocator who holds the leading exchange on each of several chains holds several chain risks in place of a diversified book of exchanges, because the fragmentation moves the risk without removing it.
The Wrong Measure Produces the Wrong Decision
Concentration data is only useful alongside the right denominator, and the wrong one produces confident mistakes.
Reading exchange health from locked capital misprices the category, because its business is fees while its locked capital is merely inventory. An allocator relying on that figure would have read the recent drawdown as an exchange decline when the category was holding its share of the system.
Reading liquid staking from dollar value confuses two separate things, because that value falls with the price of the staked asset whether or not participation changes. A category can appear to shrink while its actual usage holds simply because the asset it is denominated in has fallen. The data itself is accurate, and the error enters through the choice of measure.
The Fragility Concentration Cannot Remove
The deepest constraint is structural and so far unresolved, because concentration is counter-cyclical and reaches its height when conditions are at their worst, since stress drives capital into the largest venues at the very moment a failure in one of them would do the most damage. The qualities that make the leaders the safe choice in calm markets, their depth and their integration, are the same qualities that make them the largest single points of failure in stressed ones.
The qualities that make the leaders the safe choice in calm markets are the same qualities that make them the largest single points of failure in stressed ones.
A system can measure this, monitor it, and allocate around it, yet no current design removes it, since for now the pull toward safety and the buildup of systemic risk are the same motion. That is why concentration has to be managed instead of assumed away.
None of these failure modes amounts to a case against using DeFi, and together they make a case for pricing concentration honestly. A position in a dominant protocol carries the depth and the integration that make it attractive alongside the reach that makes its failure everyone's problem, and the two arrive together and cannot be separated.
An allocator who understands this holds the same exposure as everyone else, with the single difference that they have measured it and can act before a market turns instead of after.
About the data. Every figure in this article is drawn from the Sentora Crypto Dominance Dashboard, which tracks the top five protocols across lending, decentralised exchanges, and liquid staking, alongside each category's share of total DeFi TVL. The dataset is maintained by Sentora Research and refreshed regularly.
