Tokenized RWAs as DeFi Collateral: The $2.7B That Actually Moves

The second act nobody covers

Almost everything written about tokenization is about issuance. A fund launches on-chain, a total value figure gets quoted, the sector is declared to be growing.

Issuance is the easy half. The question that determines whether tokenization is a settlement upgrade or a genuine financial primitive is what happens to the token after it exists. Can it be posted as collateral? Borrowed against? Composed into something the underlying asset could never do?

In April 2026 Dune Analytics published the first serious answer, and it was not the one the sector's marketing suggests. Of roughly $27 billion in tokenized real-world assets, approximately $2.7 billion — about ten percent — was actually deposited into DeFi lending protocols.

Ninety percent sits still.

What the $2.7 billion counts, precisely

Before drawing any conclusion from that number, the methodology needs stating, because it is narrower than "RWAs in DeFi" implies.

The figure counts RWA tokens deposited or supplied into lending protocols — collateral and vault supply only. It excludes borrowed amounts, and it excludes stablecoins supplied as lending liquidity. That is a deliberately conservative definition, and it is the right one for this question: it measures tokenized real-world assets being used as balance-sheet inputs, not the total notional flowing around them.

We are citing Dune's analysis as reported in Crowdfund Insider's April 2026 coverage alongside Dune's own publication. In August 2026 Dune launched a dedicated RWA dataset specifically to make post-issuance activity observable, which suggests the measurement problem this article describes is one the data providers have also noticed.

The inversion at the centre of the sector

Here is the finding that should reframe how you read every tokenization headline.

Credit instruments were roughly 17% of tokenized assets under management but around 80% of DeFi deposits.

Tokenized treasuries dominate the totals, generate the press releases, and attract the institutional names. They are also, overwhelmingly, inert. Tokenized private credit is a much smaller share of the market and does nearly all of the work.

The reason is unglamorous arithmetic. Looping only pays when the yield on the collateral exceeds the cost of borrowing against it. A tokenized treasury yielding a few percent leaves almost nothing after borrow costs, so there is no trade. Credit instruments yield materially more, and the spread is what makes the position worth opening.

Put plainly: the assets that move are the ones where someone is being paid to take risk. That is not a criticism, but it should adjust your reading of "RWA adoption" statistics, which almost always lead with the treasury number.

For context on the underlying instruments, our guide to tokenized private credit covers where the legal claim actually sits and what has already gone wrong in the category.

Where the active capital sits

Four venues hold most of the composable RWA capital, across Ethereum, Solana and various L2s.

Figures as reported from Dune's April 2026 analysis.

On Kamino, we are declining to pick a number. Two accounts of the same Dune analysis give $476 million and $587 million, a difference of over twenty percent. We do not know which is correct or whether they are measuring on different dates. Reporting the disagreement is more useful than choosing the figure that makes a cleaner table, and if you need Kamino's exposure for a real decision, query the source data rather than trusting either.

Morpho's trajectory is worth noting because it shows how fast this went from nothing to something. RWA deposits on Morpho grew from near zero at the start of 2025, to roughly $400 million by the end of Q3 2025, to $957 million by April 2026. Small in absolute terms, but that is a category being built rather than a category being talked about.

One instrument worth singling out: Maple's syrup tokens reportedly passed $1 billion across Aave and Kamino on four chains. These sit between an RWA and a stablecoin, and the fact that the largest single composable position is a hybrid rather than a pure tokenized asset is itself informative about what the market actually wants.

How the loop actually works

The dominant strategy is straightforward enough to describe in four steps, which is part of why it has attracted capital.

1. Acquire a yield-bearing tokenized credit position.
2. Post it as collateral on a lending venue such as Morpho or Aave.
3. Borrow stablecoins against it at the venue's borrow rate.
4. Use the borrowed stablecoins to acquire more of the same position, and repeat.

While the credit yield exceeds the borrow cost, each turn of the loop adds spread. A position yielding around six percent against a stablecoin borrow cost around three percent produces a positive carry, and leverage multiplies it.

This is a real strategy with real economics, and it is why the composable tenth exists at all. It is also leveraged exposure to illiquid private credit, and the sentence describing it should always contain the words "leveraged" and "illiquid" together, because the risks are not the ones the yield number suggests.

What you are actually underwriting

This section is the longest in the article deliberately. The mechanics above are simple; the risks are not, and they compound rather than sitting side by side.

The underlying asset is illiquid and the token is not. A tokenized private credit position can be transferred in seconds. The loan book behind it cannot be sold in seconds, or in days. Everything else in this list follows from that mismatch. When a token trades continuously against an asset that does not, the token's price is a claim about liquidity that has never been tested at scale.

Pricing depends on an oracle, not a market. Liquidation thresholds are enforced against an oracle price. For a private credit instrument, that price is not discovered by a deep order book; it is reported. If the reported value lags reality during a credit deterioration, the protocol will not liquidate when it should, and when it does liquidate the collateral may be worth materially less than the last posted mark. Our explainer on the RWA oracle problem covers why this is structurally harder than pricing a liquid token.

Redemption is not guaranteed, and terms do not change because it is a token. Private credit has a term structure and issuer-set redemption windows. Under stress, the exit you assumed may be a queue.

Liquidation assumes a buyer. The mechanism that protects lenders requires someone willing to take the collateral at auction. For a tokenized credit position in a stressed market, that buyer may not exist at any price a liquidator will accept, which converts a collateral shortfall into a protocol loss. Our guide to DeFi lending liquidation warning signs covers what precedes these events.

Leverage compresses the margin for error in all of the above. A loop that survives a ten percent mark-down unlevered may not survive a three percent one at four turns.

And the correlation is unhelpful. The conditions that impair private credit — rising defaults, tightening liquidity — are also the conditions that raise stablecoin borrow rates and reduce liquidator appetite. These risks arrive together, not independently.

None of this makes the strategy unreasonable for someone who understands it and sizes it accordingly. It makes the yield number a poor summary of what is being taken on. If you are new to the underlying mechanics, our explainer on how DeFi lending works is the right starting point.

A note on figures still circulating

While researching this article we read the rwa.xyz tokenized treasuries registry directly on 6 September 2026. It showed $15.86 billion in distributed value, down 2.22% over the preceding thirty days, across 25 treasury fund assets held by approximately 69,025 holders, at a 7-day APY of 3.42%. The largest positions were BlackRock's BUIDL at $2.79 billion, Circle's USYC at $2.62 billion, Ondo's USDY at $2.20 billion, iBENJI at $1.71 billion and WisdomTree's WTGXX at $1.22 billion.

We note this because considerably larger treasury figures from mid-2026 coverage are still being quoted as current, and because the thirty-day direction is down. A category that only ever gets reported when it grows will be systematically misrepresented.

As always in this sector, check whether a figure describes distributed value, meaning what holders actually hold on-chain, or represented value, meaning the notional value of the underlying. Publishers quote whichever is larger without saying which. We use distributed value and date every reading.

What would move the other 90%?

Three things, in ascending order of difficulty.

Redemption terms that survive stress. Not faster redemption in calm markets, which several issuers already offer, but terms that hold when everyone wants out simultaneously. Until an issuer has demonstrated this through an actual stress event, it is an untested promise.

Oracle pricing an institutional risk committee will underwrite. The current arrangement, where a reported mark drives a liquidation engine, is workable at $2.7 billion and would not be at $27 billion.

Venues with counterparty and compliance controls that clear institutional review. Permissionless composability is the feature that makes this interesting and the property that makes it unusable for regulated holders. The venues resolving that tension will capture the flow, if anyone does.

Absent those, the ninety percent will keep sitting still — and it is worth saying clearly that this is not a failure. A tokenized treasury held by a corporate treasurer for faster settlement is doing exactly its job. Composability is a capability, not an obligation. The mistake is reading tokenization totals as though they measure DeFi activity, when the actual overlap is one asset in ten.

Conclusion

Tokenization's first act worked: assets are on-chain, at meaningful scale, with real institutional names attached. The second act is where the claims get interesting and the evidence gets thin.

On the evidence available, about ten percent of tokenized assets do anything at all, that tenth is dominated by credit rather than the treasuries that dominate the headlines, and the strategy driving it is a leveraged carry trade whose risks are correlated and whose collateral is priced by report rather than by market.

That is a real, functioning, early market. It is not the frictionless composability story, and the gap between the two is the most useful thing to understand about this sector right now.

This analysis is built from Dune Analytics' April 2026 composable RWA data as published and as reported by Crowdfund Insider, from Dune's August 2026 RWA dataset announcement, and from our own reading of the rwa.xyz tokenized treasuries registry on 6 September 2026. We did not run these strategies, hold positions in any instrument named, or receive briefings from these protocols. Where sources disagree, as with Kamino's deposit figure, we report the disagreement rather than selecting a number. On-chain values change continuously and this data is several months old in places; verify before relying on any figure here. This is not investment advice.