Vaults need quality RWAs

The vault manager’s cockpit

You run a vault. Sixty million dollars in stablecoins, your name on the allocation, one number published every day where anyone can line it up against every other vault on the same screen.

Come in low and the deposits leave by Friday. Reach too far and you become the thread someone screenshots next quarter, the curator who chased an extra point of yield into a market that should have been left alone. That number has to come from collateral, and the collateral you can actually get your hands on is the part nobody mentions when they sell you on the magic of on-chain yield.

The safe floor underneath you is a tokenized Treasury paying around 3.5% in mid-2026. Your depositors want more than that, or they would have kept their cash in a money-market fund and skipped the smart-contract risk. So the spread you owe them, the entire reason they chose your vault, comes down to what you lend against and how well you understand it. Matching that floor is easy. Beating it, quarter after quarter, without taking the risk that ends a career, is the entire job.

For most of the last two years, the honest answer to what you can lend against has been a short list: ether, bitcoin, a handful of liquid staking tokens, and US Treasuries. Good collateral, all of it. It is also the same collateral everyone else in your seat is using. Which means beating the screen comes down to two questions almost nobody answers honestly: where does the extra yield actually come from, and how is it built so you can lend against it safely?

Every vault is fishing the same pond

Start with the water. There is more idle capital sitting on-chain than at any point in this market’s history. Stablecoin supply sits at around 300 billion dollars, near its all-time high, held across more than 260 million wallets. Most of it earns nothing where it sits. The job of putting it to work has consolidated into one place: the curated vault.

That layer grew up quickly. A depositor places USDC into a vault, and a professional risk team, Steakhouse, Gauntlet, MEV Capital, decides which lending markets the money enters and at what size. Morpho‘s curated-vault model became the dominant home for this capital, with close to seven billion dollars in deposits. The serious money noticed. Morpho closed a 175 million dollar round in June 2026, co-led by Paradigm, a16z crypto, and Ribbit, with Apollo, VanEck, and Circle taking part. Apollo, which manages more than 900 billion dollars, separately agreed to acquire up to nine percent of Morpho’s token supply over four years. Coinbase routes its USDC lending product through a Steakhouse-curated Morpho vault and runs more than 1.6 billion dollars of collateral through the same plumbing. Kraken launched its own version in January 2026.

Here is the part that does not appear in any of those announcements. Almost all of that capital lends against the same narrow set of collateral, the ether, the bitcoin, the staking tokens, the Treasuries. Blue-chip stablecoin vaults cluster between 4% and 8%, and the safe end of that range is anchored to the same Treasury yield you started with. Thousands of vaults, billions in deposits, all fishing the same pond. When everyone competes for return out of the same shallow water, two things follow. Returns compress. And the curators who refuse to accept the compression go looking for deeper water in places they do not understand.

When the pond runs dry: the Goldfinch lesson

Goldfinch is the cleanest illustration of what happens next, and it is recent enough that the ink is still wet. The protocol launched in 2021 on a thesis that was, and remains, correct: real yield exists in emerging markets, and the world has a vast unmet need for capital the existing system does not reach. Andreessen Horowitz and Coinbase Ventures backed it on a pitch of roughly ten percent yields drawn from genuine economic activity. The thesis was sound. The architecture was not.

Goldfinch lent on an undercollateralized basis to off-chain borrowers, with credit decisions delegated to a loose group of auditors rather than underwritten in-house, and with recovery dependent on legal systems where collateral is slow and expensive to chase. A five million dollar loan to a Kenyan motorcycle financier went bad after the borrower diverted funds to a struggling parent company in breach of its terms. A twenty million dollar facility to a credit fund left roughly seven million impaired. A Singapore borrower repaid about four million of a ten million dollar loan and defaulted on the rest. Cumulative losses passed eighteen million dollars, and the token lost almost all of its value. In June 2026, holders voted to wind the protocol down and shut its newer Prime product, which had never reached the scale to justify the spend.

The instructive part is what the failure was not. It was not the blockchain. The rails moved the credit exposure exactly as designed. What broke was the difficulty of pricing and enforcing credit when the collateral is thin and the borrower is far away. The people who run the demand side of this market said as much. Aave’s founder, long skeptical of Goldfinch’s emerging-market model, was blunt that the closure should not be read as proof the model is broken. This doesn’t mean that undercollateralized onchain lending doesn’t work, he wrote, expecting better underwriters to step in with stronger models.

There is a second failure that looks like the opposite of Goldfinch and comes from the same root. When transparent, high-quality collateral is scarce, some curators reach for yield in synthetic, opaque places instead. When the Stream Finance xUSD market unwound late in 2025, several well-known curators who had allocated to it despite public warnings took losses, while the disciplined shops that stayed away were untouched. One failure came from lending against real assets that were poorly underwritten. The other came from lending against engineered yield no one could see into. Both trace back to a single variable: the quality and the curation of whatever sits underneath.

The rails are live. The RWA shelf is thin.

While that was playing out, the demand side quietly built the piece that had always been missing. Tokenized real-world assets used to sit on-chain doing nothing, exposure without utility. That has changed.

Aave launched Horizon, a market where qualified institutions post tokenized real-world assets as collateral and borrow stablecoins, while anyone at all can supply stablecoins to that market and earn the yield those institutional borrowers pay. It is live and built to scale past a billion dollars. Morpho runs the same idea from the other direction. A private-credit fund is tokenized, Apollo’s diversified credit through Securitize, Fasanara’s vehicle, Pareto’s, then supplied as collateral on Morpho and financed with stablecoins drawn from a Steakhouse-curated vault. A static yield product becomes productive, composable collateral. The ambition is no longer quiet. Morpho now describes itself as building the open credit network for the world, and one of its backers noted that trillions of dollars move through global credit every day on infrastructure that has barely changed in decades.

So the rails are finished, institutional-grade, and hungry. Now look at what is actually eligible to run on them. Horizon opened with collateral that was, every piece of it, US paper: a Superstate Treasury fund, Circle’s money-market token, a Centrifuge AAA collateralized loan obligation[1] fund, a VanEck Treasury fund. Now look at the shape of that shelf:

• Tradable tokenized real-world assets: about $31 billion.
• Of that, tokenized Treasuries: roughly $14 billion, close to half, paying about 3.5%.
• Actually usable as collateral inside DeFi lending today: only a few billion.

The plumbing was finished before the inventory arrived, and you can watch it on rwa.xyz in real time. What sits on the shelf is thin, and almost all of it comes from one country.

Where the real yield lives

Come back to your sixty million for a moment. The spread you owe your depositors is real, and it exists. It is sitting in assets that have not reached the rails yet.

Concretely, the assets that beat the floor look like this:

• Saudi REITs[2] pay high-single-digit yields on rental income that is steadier than most equity dividends.
• Tokenized private credit, where it exists, pays between 8% and 12%, and some structured credit already on-chain pays more.
• Bank and telecom dividends across emerging markets sit comfortably above the Treasury floor you are anchored to.

None of it is exotic. It is the ordinary investable surface of most of the world’s economies, and emerging markets are the larger half of this opportunity, the part that drives most of the world’s growth yet is almost none of what is tokenized today.

Hundreds of US-listed names are live on-chain through platforms like Ondo. The number of emerging-market listed equities tokenized for these markets is, for practical purposes, zero. The first emerging-market credit is starting to appear, Brazilian agricultural receivables, a handful of emerging-market portfolios, but it is a sliver. The tradable shelf is US Treasuries, gold, and US stocks. The rest of the world is missing.

The reason it is missing is worth stating precisely, because it is not the reason most people assume. The technology is solved. Horizon and Morpho proved that an asset can be tokenized, posted as collateral, financed, and made composable. What is missing is everything that has to happen before the token exists: sourcing an asset worth lending against, qualifying it, structuring it so a vault can absorb it, and wrapping it in a form a regulated institution can hold. That work, the unglamorous origination layer, is the bottleneck. It is also exactly the layer Goldfinch skipped.

Structuring RWAs a vault can use

DeFi did not replace the machinery of finance. It sits on top of it. Before a real-world asset can become collateral a vault will accept, the same unglamorous work still has to happen that has always made lending safe:

• someone independent has to value the asset,
• an auditor has to check the books,
• a trustee has to be able to step in and take the collateral if a borrower stops paying.

The blockchain settles and moves the instrument. It does not value it, and it cannot chase it down.

The instrument that turns a real-world asset into something a vault can hold is, in the end, a familiar one. It is a note backed by the asset’s cash flows, issued through a regulated structure in Luxembourg[3] that ring-fences each deal, so that one deal’s investors are protected even if another runs into trouble, and the structuring always comes first, before any token exists. That is the same kind of instrument a vault already accepts. The AAA collateralized loan obligation fund that Aave Horizon took as collateral at launch, now several hundred million dollars in size, is a close cousin of it: a pool of real loans, packaged so the safest slice can be lent against. Luxembourg is not incidental. It is one of a handful of triple-A-rated countries, and it has spent two decades as Europe’s hub for exactly this kind of structuring. The difference between the CLO already on Horizon and an emerging-market note is not the structure. It is the source.

Because the structure is flexible, the note can be shaped to fit what a vault actually needs. It can be collateral-backed, with real first-loss protection and assets a curator can underwrite, or its return can be linked to the performance of the underlying. Either way it can reach investors on two tracks: as a security carrying an ISIN[4] for the institutions that need one, and as a token for the vaults that do not. The output is the thing the shelf is missing: an instrument that earns more than Treasuries and is built to be lent against.

A curation problem, not a coding problem

None of this can be automated, and that is the heart of it. A model can price a tokenized Treasury, because a Treasury is a known quantity. It cannot tell you:

• whether a sponsor in a frontier market will actually repay,
• whether the collateral can be enforced in that jurisdiction,
• how a deal should be structured to protect investors if the cash flows slip.

That is judgment, and it is done by people. It is exactly the work Goldfinch handed to a loose group of outside auditors and got wrong.

So the edge in the next phase of on-chain credit belongs less to the best engineers than to the teams that can originate genuinely new real-world assets and stand behind the diligence. The blue-chip managers already on-chain brought what they had: US Treasuries and US credit. The assets that are missing, emerging-market credit, real-asset income, the listed companies of the rest of the world, need originators who can source them, structure them to an institutional standard, and turn down most of what they see. As we have argued, the real bottleneck is not the rails, it is curating real assets from every market. That is a different skill than building the rails, and it is the one that is now scarce. It is the work we do at T-Blocks: building Internet Capital Markets for the parts of the world the existing rails never reached.

The prize is not small. The sustainable yield a curated vault can offer today clusters around 4% to 6%, anchored to the roughly 3.5% a tokenized Treasury pays and to the same blue-chip collateral every vault already holds. The structured real-world assets that have not reached the rails yet, emerging-market credit, real-asset income, REIT distributions, earn 8% to 12% at the asset level, and some structured credit already on-chain pays more. That gap is the headroom.

As structured collateral becomes a larger share of what vaults lend against, the blended yield available on-chain can move from the low-to-mid single digits toward the high single digits. Not through leverage or synthetic engineering, but because the underlying genuinely earns more, and only if the discipline holds: over-collateralization, first-loss protection, independent valuation, curated risk parameters, the things that let a vault reach the higher band without becoming the next Goldfinch. The ceiling on sustainable on-chain yield was never the Treasury rate. It is the yield of the real economy, properly underwritten.

The pond was never the real constraint. The water is real, and most of it is somewhere the rails have not reached yet. Bringing it there, one curated asset at a time, is the whole opportunity.