Vaults: The Infrastructure Layer for Institutional Finance Onchain

Vaults are becoming DeFi's default entry point for institutions

This article is an excerpt from our research on Vaultification of Finance.

Download the complete report here:
https://docsend.com/view/r3278hmn8d6tqyqn

Categorising Vaults

This section of the report presents a quantitative analysis of the vault landscape to provide a comprehensive picture of the sector and its evolution. We analyse the ecosystem by category, tracing TVL shifts across vaults and curators.

We break down curator concentration and provide an outlook of the main flows, contextualising the structural shifts that defined vaults during this year.

Vaults should not be considered a single, encompassing market but rather evaluated by their various implementations, each with different parameters, risk vectors, and responses to stress tests. Aggregated stats only provide a small part of the picture, with much-needed nuances.

Before starting the analysis, it is important to define the term “vault“ as the basis of our methodology.

Our definition is based on the deployment path. Vaults are categorised as “instruments through which users access active yield-generating strategies”. Any asset that is purely a wrapper of an offchain instrument is excluded from our analysis.

Maple’s syrupUSDC qualifies as a vault: users deposit stablecoins into the protocol, which lends them to institutional borrowers and accrues APY through credit activity on the issued token.

Lido stETH is a vault: users deposit ETH, and the protocol earns staking yield, which is distributed through the rebasing token.

Centrifuge JAAA is a vault: users access AAA-rated CLO yield through a tokenised wrapper that generates yield via its credit positions.

BlackRock’s BUIDL does not qualify as a vault by this definition: it is a direct token issuance representing a 1:1 claim on an off-chain Treasury fund.

We apply this lens to define eight structural categories:

1. Lending vaults

2. Liquid staking

3. Restaking

4. Risk-Curated vaults

5. Vault Infrastructure Providers and Yield Optimisers

6. RWA credit vaults

7. Perpetual LP vaults

8. Options vaults

For the purposes of this analysis, we categorise risk-curated vaults as a standalone category to better understand their dynamics and growth.

Before diving into the categories individually, we highlight the vaults’ overall performance.

State of the Vault Ecosystem

The net TVL across all defined vault categories combined is $120.4 billion, down ~50% from its peak of $241 billion around last October. The downtrend after the October peak was driven by the October Liquidation Event, which triggered a cascade of liquidations across DeFi.

The vault TVL number is higher than the current DeFi TVL (~$86 billion) due to overlap. For example, Liquid Staking protocols like Lido issue stETH, a rebasing asset representing the yield on staked ETH, which is used as collateral in Lending protocols like Aave and Morpho.

The overall picture changes drastically if we move to a category-level analysis. Recent incidents led to TVL outflows and prompted a broader reality check (and, hopefully, a shift toward a security-first approach) on security and risk management across the sector.

Categories such as Lending, Liquid Staking, and Restaking suffered the most, as they have the most exposure to onchain assets and enable the onchain economy, while RWA vaults continue to showcase uncorrelated growth due to non-crypto asset exposure. Categories like Option vaults peaked in April 2022 and have been struggling ever since. Risk-Curators-managed vaults took an equal hit to other major categories due to the October Liquidation Event. Their TVL peaked around the end of October, then declined following the Stream Finance blowup.

Three events from October 2025 to May 2026 (the Stream Finance, Resolv, and the Kelp hack) represent a good stress test window, as these blowups/exploits had cascading effects across all of DeFi. In the chart below, we highlight the TVL history across these categories in this exact period. As mentioned above, most categories performed poorly, except the RWA vaults, which gained 37.8% during the same period, while the others experienced significant drawdowns.

Next, we move on to analysing per-vault category growth and highlighting recent trends and shifts.

Lending Vaults

Lending is the largest vault category and accounts for the majority of DeFi TVL. Last year marked a broad shift toward curated vaults, driven by products like Morpho, which helped scale the trend. On Morpho, curators can create their own vaults that can have exposure to multiple markets and earn yields for depositors. These vaults can eventually be curated by any provider, including TradFi institutions. Morpho’s recent Vaults V2 upgrade enabled even more functionality for curators, including the ability to embed approved adapters to access yields from multiple sources, granular risk control (such as setting absolute or relative caps on vault exposure), built-in KYC controls, and other features.

With the same context, Aave has also launched its V4, which introduces specialised spokes and a unified liquidity hub. Spokes provide greater functionality with custom risk parameters, isolated collateral types, and Oracle configurations per market. It differs from Morpho’s curator-led model in that Aave’s governance still reviews and approves the implementation of these Spokes, whereas Morpho is permissionless. This is Aave’s shift from Monolithic to Modular Lending.

The curator model led Morpho to reach over $7.5 billion in TVL concentrated on the Ethereum mainnet and Base. Base has strongly contributed to Morpho’s growth, going from $604 million to over $2.8 billion. This shows the power of the distribution partnerships Morpho has been pursuing, such as the one with Coinbase: currently, about 40% of TVL in USD terms is cbBTC, while it helped facilitate over $1 billion in loans for Coinbase users.

In response to the curator model finding PMF among institutional investors, Aave is competing on institutional rails with Horizon, which has accrued over $350 million in TVL since its launch.

Additionally, in the last few months, Aave has undergone many changes, including service providers like BGD and ACI leaving Labs, and the announcement and approval of the “Aave will Win” framework, which routes all revenue from different Aave products to its holders.

None of these incidents had much direct impact on Aave users. The only impact of these incidents was on the Aave token’s price performance, but the recent KelpDAO attack changed things: Aave lost over $12 billion in TVL, moving closer to its competitor, Morpho, in terms of TVL. The ratio between Aave TVL and Morpho TVL used to be in the 5-6x range but has now fallen below 2 because of this incident.

Spark, a lending protocol part of the Sky ecosystem, has been among the ones benefiting the most from inflows post-rsETH hack.

The image below provides a picture of the TVL change in the protocol:

Most notably, Bitcoin supply nearly tripled, stablecoin borrows increased by 78% to $752m with utilisation rates remaining contained, and WETH borrows increased by 44.1% to 325k WETH borrowed.

Fluid’s unified liquidity layer also introduced a different approach to liquidity design, in which lending, borrowing, and DEX share the same capital. User collateral acts as an LP in Fluid’s DEX, earning trading fees, while borrowed funds are deployed as Smart Debt into DEX pools, earning fees that offset borrowing interest costs.

Another interesting approach from Fluid is its collaboration with protocols such as Jupiter and Venus, through which it has launched white-label products like JupLend (Solana) and Venus Flux (BSC), each of which currently sits at a TVL of $926m and $21m, respectively. This comes as a broader positioning by Fluid to work with major players across chains and gain more market share, with those players sharing fees with Fluid.

Honourable mention to Kamino’s vaults, the main lending stack on Solana, with over $1.6 billion in TVL. The protocol experienced notable growth through their K-Lend model, the Morpho equivalent on Solana. This allows Kamino to collaborate with established curators such as Gauntlet and target institutional integrations.

The largest vault on the platform is currently Sentora’s PYUSD, with over $219 million in TVL, the second is RockawayX’s RWA USDC vault, at just $33 million, indicating there’s still a lot of room to grow for both Kamino and Solana as a whole.

Liquid Staking and Restaking

Liquid Staking and Restaking both account for a large share of the TVL in vaults, at $42.4 billion and $20.6 billion, respectively.

Major players in Liquid Staking are Lido ($21.8 billion), Binance Staked ETH ($8.9 billion), Rocket Pool ($1.2 billion), and Coinbase cbETH ($320 million).

Lido has, over time, maintained its dominance, and its issued asset, stETH, is highly composable across DeFi. At the same time, Lido’s dominance also signals concentration risk. They have expanded their offerings by introducing their Earn product, which serves as an aggregator layer, depositing users’ funds across DeFi to earn yield. However, this product took a hit after the recent Kelp DAO hack due to its exposure to $rsETH.

Binance Staked ETH has grown by 121.8% since last year by leveraging Binance’s user base.

For others and the category as a whole, growth has been slow and has come at the cost of staking yield dilution, which currently stands at ~2.5%.

Source: https://dune.com/queries/3980225/6698235

On the other hand, Restaking and Liquid Restaking, as a category, grew to boost yields earned from Liquid Staking. Kelp DAO was a Liquid Restaking protocol, and its hack and the broader DeFi cascade highlighted the composability risk these assets pose, as they are accepted as collateral across DeFi, making it more of a bug than a feature in this incident. The major players in restaking are EigenCloud ($7.8 billion), EtherFi ($5.7 billion), Kelp DAO ($1.6 billion), and Renzo ($270 million).

The restaking products, such as EigenCloud and EtherFi, expanded over time to include additional services.

EigenCloud’s 2025 rebrand helped position them more in the AWS category, pushing toward verifiable execution. EigenDA, Eigen’s data availability layer, is used by multiple L2s, including MegaETH, Mantle and Celo. The data posted on EigenDA surpassed 1.8 TB and generated ~$90k in total fees. EigenCloud TVL had remained flat in ETH for a long time but recently dropped after the Kelp hack, as users tend to withdraw their funds during periods of uncertainty.

Similarly, EtherFi expanded into a neobank and has thousands of active card users who have cumulatively spent ~$440 million through its offerings. Moreover, they have a Liquid product (let’s not forget that EtherFi first launched as a liquid staking protocol) that supports multiple strategies to boost yields across DeFi. One of its top ETH yield vaults has a TVL of $177.5 million.

Risk Curated Vault

Risk Curated Vaults is one of the fastest-growing categories, reflecting the shift from monolithic to modular lending. The managed vaults they provide on platforms like Morpho earn them performance and management fees, similar to how TradFi funds operate, deploying users’ capital across strategies to generate returns.

The current TVL of this category is ~$6.5 billion, of which 75% is held by three curators: Sentora ($1.85 billion), Steakhouse ($1.63 billion), and Gauntlet ($1.5 billion), suggesting less competition in this category.

These risk curators charge lower fees than TradFi hedge funds and venture funds, which generally charge a management fee (~1-2% of the entire AUM) and performance fees (~10-20% of the interest earned). For example, Steakhouse Financial, the top curator by revenue, generates $3 million in annualised revenue on an AUM of $2.13 billion (0.14% of total AUM). These curators usually charge only performance fees and, in some cases, management fees, but these fees are currently significantly lower. This is the result of the competitive landscape, as curators race to offer the lowest fees to attract the most TVL.

But still, the risk curators are concentrated at the top, with dominance split among three providers, which is better than in Liquid Staking, where Lido leads by a wide margin.

Moreover, what does this concentration signify? This is what the Steakhouse team had to say about it: “Concentration will likely follow power laws found in the traditional asset management analog, e.g. ETFs, where most of the AuM concentrates around the leading managers. This does not have to be a bad thing and is a reflection of scale and trust compounding to larger managers competing on performance, product range and fee load. The good thing about DeFi is that the arena is open. Anyone can come in and compete. We expect concentration to persist at the top tier, with healthy competition at the margin and room for specialisation.”

The concentration dynamics changed very recently after the Stream Finance Incident, before which MEV capital and Re7 also had strong representation, peaking at $1.49 billion and $830 million, respectively. Later, they shrank, and Sentora grew to become the second largest curator.

Additionally, following the KelpDAO hack, the impact on risk curators is evident, but a few winners, such as KPK (+159.6%) and Gauntlet (+42.7%), have seen net positive flows. For KPK, this growth came from their recent Morpho V2 vault launch, which attracted deposits from ensdomains, CoWSwap, NexusMutual, and others. They integrated agent-powered automation for rebalancing and vault exits, improving their risk management. For Gauntlet, growth came from its BSC chain expansion and its collaboration with the Lista DAO Lending protocol, which attracted new inflows.

As Juan Pellicer from Sentora points out, “DeFi cover is also becoming a real part of the institutional picture. The ability to offer economical cover changes the calculus for a treasury or asset manager who has to answer to an investment committee, it’s a structural unlock”.

Vault Infrastructure Providers and Yield Optimisers

Yield optimisers, as a category, are maturing and seeing a ton of new entrants. As the yield sources onchain increase, the optimisation or aggregation model would serve as a better vault model to provide depositors with the best yield across the board.

Protocols like Veda ($1 billion), Upshift ($380 million), and Fluid Lite Vault ($164 million) lead the category overall. Everyone serves a different model, but the goal remains to seamlessly integrate vaults which optimise yield and provide the best available yield across DeFi to their depositors. They are currently well below their peak due to ongoing market drawdowns and the stress period from October last year.

It’s helpful to think of providers like Veda and Upshift not as an aggregator but as infrastructure for creating isolated yield products. Upshift uses its own policy engine to enforce the vault mandate and self-custodiality by restricting deployments to whitelisted chains/protocols/tokens/smart contract calls. Additionally, Upshift is better categorised as a multi-strategy vault because its vaults provide exposure to strategies across DeFi, including Lending, basis trading, carry trading, LPing, RWAs, etc.

Veda leverages a modular architecture, separating operations into a “boring” vault, whose only purpose is to hold assets, while any specialised tasks are executed by external modules. The protocol uses Merkle trees to enforce permissions by whitelisting for specific vault actions.

Infrastructure providers make it extremely easy for institutions to start with a single integration, allocating to one lending protocol, and add more sophisticated strategies as the product offering expands for higher yield and deeper liquidity.

Other products, such as Fusion from IPOR ($30 million) and Gearbox ($29 million), also serve as yield-optimisation layers. The main goal of Fusion, for example, is onchain vault infrastructure that enables independent entities such as curators and asset managers to build and operate yield strategies like leveraged looping and carry trades.

Every Fusion vault is unique in terms of curation, strategy, and allocation. Automations are built at the strategy level, with different triggers for optimisation, leverage maintenance, liquidation risk management, routing, and more. Examples include swapping in the case of a negative spread, using a flash loan for cross-market migration of a leveraged position, or exiting in the event of risk events. As the Fusion team points out, “This automation was key in the latest rsETH/Aave crisis, where the IPOR DAO stETH looping vault on mainnet was among the first to fully cut exposure to Aave v3 core. Automation and execution in general allow curators to risk manage quickly when rapid action is needed most”.

Leveraged Looping represents by far the highest protocol-managed value, standing at ~$80 million. This is higher because TVL is an insufficient metric for yield optimisers. Instead, these providers should be analysed by their Assets Under Management (AUM), since they allocate funds to other protocols, so their TVL doesn’t reflect the true growth.

Gearbox launched a vault architecture targeting both passive lenders and active borrowers.

At its core, the protocol focused on enabling access to leveraged or delta-neutral exposure to farming or liquidity provision strategies. While most vault mechanisms are built around asset management for curators, Gearbox focuses on risk management infrastructure for lenders.

Borrowers can open Credit Accounts to interact with outside protocols from Gearbox while funds remain non-custodial. V3 introduced strategy-level firewalling, protecting the protocol in the event of a failure of a Credit Account or strategy. In the event of incidents, they cannot drain the shared liquidity pool beyond the amount allocated to it, thereby protecting passive lenders from contagion.

Recently, the protocol also announced a focus on RWA-looping vaults.

RWA Vaults

RWA Vaults has witnessed consistent growth over the last 5-year period, with a CAGR of 231.3%, reflecting increasing interest in RWA yield exposure among both retail and institutional investors. Even after the recent Resolv and Kelp exploits, the RWA vault category has been sticky and has not moved much due to limited exposure to onchain assets.

The largest players in this category are Maple Finance ($2.1 billion), Centrifuge ($1.6 billion), Anemoy Capital ($1.1 billion), Re ($263 million), and more.

Maple Finance grew rapidly over the last year, with TVL climbing almost 10x from the start of 2025. This growth can be attributed to various factors, including the launch of Syrup, a part of the protocol’s transition from an institutional-only model. This launch opened the door to retail flow-through products like syrupUSDC and syrupUSDT, which are highly composable in DeFi. DeFi composability and deep liquidity enable assets to be leveraged by looping through lending protocols and integrating with products like Pendle, thereby contributing to the growth flywheel. Reflecting demand for the product, the platform’s current active loans total ~$1.7 billion. These loans are dominated by USDC, representing ~75% of total active loans, followed by USDT, which accounts for the remainder.

Other products also witnessed great growth. Centrifuge, for example, positioned itself as a private credit infrastructure protocol. Its partnership with Anemoy led to $1.1 billion of T-bill pools running on Centrifuge infrastructure. Centrifuge has also been recently chosen by Coinbase as its tokenisation partner.

Products like Re brought in reinsurance underwriting risk onchain exposing users to real-world yield more broadly. Apart from this, Upshift USDC vaults lend to overcollateralised institutional funds, giving their depositors exposure to institutional lending.

Even with all the growth RWA has witnessed in DeFi, it still represents only a fraction of the value tokenised onchain. Currently, the active RWA DeFi TVL is about 1/10th of the total RWA value. This huge difference between the two values is because these assets fall into a different category that goes beyond general considerations for normal assets, as it includes redemption periods, compliance, and liquidity issues in certain cases.

For any asset to scale in DeFi, it needs active redemption and secondary liquidity, as users might need to sell these assets to regain liquidity, or in the instance of a lending protocol, liquidators repay loans and sell the assets close to the mark price for a profit, but with all the caveats that RWA come with, most of this becomes harder to implement. Additionally, yield-bearing assets like RWA have another important part of their growth flywheel: looping.

RWA looping borrows stablecoins against tokenised treasuries and repeatedly redeploys them into the yield-bearing vault. A 4-5% base treasury yield, levered 2-3x, can produce 7-12% in returns but only when borrow costs stay low (~1%). Onchain stablecoin rates are volatile and can significantly compress this spread. Leverage used to perform such a trade amplifies Liquidation and Oracle risk, and the strategy depends on the RWA collateral being stable in value. With this, redemption issues also play their role as some RWAs settle T+1 and some T+5.

To solve this, currently there are a few solutions:

1. ERC-7540: Introduce an asynchronous ERC-4626 vault so that users can use their redemption claim as liquidity while the underlying assets settle offchain. Centrifuge is one of the most important examples of ERC-7540 in production, using synchronous deposits and asynchronous redemptions, solving the tension between DeFi and TradFi T+ settlement. These hybrid vaults are becoming a template for any vault touching offchain assets.

2. Securitize Vault Registrar: This ERC maps each investor to their identity when using RWA in DeFi, ensuring protocols comply with all the regulations and requirements the asset requires.

3. Redstone Liquidation Flow: They perform RWA liquidation by introducing auction-based liquidations and connecting positions to KYC-verified solvers, who take on the underlying asset offchain and close the position onchain.

4. Upshift Clear: Upshift is releasing its new product with Superstate to enable instant RWA redemption, allowing users to swap their RWAs for USDC at the current reported price, subject to a 5bps redemption fee.

Another protocol in this category is 3F, a platform for leveraging RWAs onchain. It currently has $7 million in TVL and approaches the problem of RWA assets in DeFi differently than other solutions. It tries to externalise different factors, including Bridge Facilitators and Liquidity Integrators. The former provides upfront liquidity to complete the exposure the user intends to take on their base capital. For example, a user targeting $3 million in exposure and with $1 million to deposit can obtain the remaining $2 million in liquidity from a bridge facilitator, enabling the entire position at 3x leverage. Similarly, when the user intends to unwind, the facilitator provides the required liquidity, solving the redemption delay problem. The latter, liquidity integrators provide instant liquidity in the instance a user wants to exit immediately. Because even with a Bridge facilitator, the user has a $1 million deposit that must go through the entire redemption process, these integrators provide the much-needed liquidity.

Both of these approaches borrow efficiency from the market, just as liquidation works in lending, with motivated onchain actors filling the required gap in RWA looping for profit. Over time, systems like this become easier to scale because every participant has something to gain from the process: loopers get a smooth exit, and facilitators earn a profit by providing liquidity and faster redemption for users.

As mentioned in the section above, Gearbox is also planning to launch “Retokenisation”: a feature that allows the infrastructure to natively support leveraged minting and redemption of non-atomic tokenised assets, without requiring secondary liquidity or incurring redemption delays. In practice, Gearbox’s contracts will operate as one with the RWA issuer’s contracts, creating a seamless, composable system for RWA leverage directly at the issuer level, making Gearbox the only EVM protocol offering native leverage for RWAs.

Perpetual LP Vaults

Perpetual LP Vaults is represented by Jupiter Perps ($715 million), Hyperliquid HLP ($396 million), Drift ($256 million, down after recent hack), GMX ($242 million), and Ostium ($51 million).

Jupiter’s JLP remains by far the largest perp vault by TVL, but has lost more than half its value since October last year, due to the Liquidation event.

HLP has performed better in terms of value retention and is down 30% from its $600 million peak last September. Hyperliquid’s vault experiences constant ups and downs, often driven by the HLP variable yield, which is influenced by its structure and market conditions. Hence, high-yield cycles attract capital, while low-yield or loss periods push it out.

One of the major loss incidents occurred in March 2025, when a trader opened a massive short position on the Jelly token and then withdrew their margin, triggering a forced liquidation and prompting HLP to take over the position. Losses like these to the vault create a structural bias for depositors, often categorising HLP as a riskier vault, but Hyperliquid reduced the allowed leverage on such tokens to avoid such situations, thereby amplifying losses.

Products like Ostium OLP provide exposure to RWA perps and offer a different configured yield to their users, but their TVL is down ~50% from its peak. This drawdown was the result of broader market moves and Ostium yield cycles.

Additionally, Ostium recently introduced architectural changes, making OLP a senior tranche and an intraday settlement layer that never takes the first risk. This is the opposite of the HLP model: depositors who previously wanted the directional exposure OLP provided might leave, but at the same time, it becomes a passive source of yield for depositors with reduced risk in this new model.

Options Vaults

DeFi Option Vaults (DOVs), as a category, faded over time, peaking in 2022. DOVs provided exposure to strategies like covered calls and cash-secured puts but lacked capital efficiency, carried higher risk, and, over time, attracted a smaller audience, as crypto users tend to gravitate toward perpetuals. But option vaults have recently been improving and solidifying their use case, at least for more savvy users.

Option vaults no longer exist in the same format as before. Instead, they are architecturally distinct and more user-friendly, delivered via products such as Derive and Rysk. These days, option vaults are executed via a Request for Quotes (RFQ) system, with market makers running the background processing.

Derive is an options and perpetual exchange, and with its V2 launch in March 2025, it witnessed accelerated growth due to feature expansion, such as using CLOB and enabling institutional-grade features like off-exchange custody and support for multiple collateral types, processing $12 billion and $16 billion in perps and options volume, respectively. Derive V1 had vaults that are still active, giving users exposure to different strategy options and creating a delta-neutral position for its depositors, maximising APY. These vaults currently hold ~$2.4 million in TVL.

On the other hand, products like Rysk provide retail exposure to options through covered calls and cash-secured puts. It launched on Hyperliquid with a focus on covered calls on HYPE and currently has ~$56 million in TVL and processed $975 million in options notional volume. With this, they also offer Rysk Premium, a flagship product featuring as a vault for sophisticated allocators deploying funds across different options strategies and generates a consistent yield for depositors.

New vault implementations focus on solving some of the previous issues with incumbents. These included poor strategy design, with timeframes as short as 7 days; the execution of trades at fixed intervals, creating opportunities for front-running; and customizable designs that let users align their own size, strike, or expiry. Option vault providers are now much more attuned to the market pulse on which assets to list to capitalise on new windows of opportunity in yield-bearing assets.

Find the complete report here:

https://docsend.com/view/r3278hmn8d6tqyqn

written by @francescoweb3 and @noveleader ✍️

Every week for the last 3 years, we have shared our research for free, directly in your email. Not a subscriber yet? Let’s fix it:

If you are more of a Telegram guy, you can read all of our research without the noise on our TG channel:

https://t.me/castlelabsreads