Tokenized Treasury Assets Reach $16B, Most Sit Idle

Tokenized US Treasury funds have piled up roughly $16 billion in distributed value, and the industry keeps celebrating that figure. The harder question is what those assets actually do after issuance. For most of them, the answer is not much.

Most of those tokens get held, then eventually redeemed. They are not posted as collateral in lending markets, and they are not wired into structured positions. Economically they are dormant. The gap between the tokenization pitch and the delivered reality is wide, and the industry is slowly waking up to it.

The liquidity problem is concrete. A holder with $100 million in tokenized bonds who needs cash has one real option: redeem the tokens and exit the position. The smarter version, the one the industry promises, would let that holder borrow stablecoins against the tokenized bonds while the underlying asset keeps paying yield. The position stays open, and cash arrives without a sale.

A token used this way becomes a working part of the financial system instead of a faster way to move paper. Traditional finance solved this problem decades ago with repo markets and securities lending. The tokenization pitch is that those mechanisms become programmable and transparent onchain. The catch is that tokenized assets don't behave like their traditional counterparts, and the difference matters when collateral is on the line.

Ethereum, for example, can be liquidated almost instantly in a continuous market. A tokenized credit portfolio cannot. The underlying loans and bonds trade during standard market hours. Net asset values do not update continuously.

The mismatch creates real risk for any lending protocol that tries to plug a tokenized fund into its liquidation math. The math won't work until the asset itself changes.

Issuers now face a different design question. An issuer cannot ask only whether an asset can be tokenized. The real question is whether it can function inside an onchain financial system, which means it has to meet standards for pricing, redemption speed, liquidity, and legal structure that collateral use actually demands. The bar for that is much higher.

mWIN, launched in August 2026, is probably the clearest attempt so far to clear that bar from the start. Midas issues the token. Wellington Management runs the credit strategy, and Northern Trust is the custodian. The strategy was built for onchain issuance, not retrofitted from an existing fund, and it covers investment-grade CLOs (collateralized loan obligations) and other asset-backed credit with a yield around 6.9%.

Mechanics matter here. mWIN can be minted and redeemed daily on a T+1 basis, drawing on multiple liquidity sources. It supports loans in PayPal's PYUSD. Sentora sets the loan parameters from historical data, so loan-to-value limits are calibrated and positions can be sold before the collateral loses value. The architecture is what makes the token usable rather than just issuable.

Most tokenization projects start with the asset and figure out utility later, if at all.

mWIN started with the question of how the token would function inside a lending market and built backward from there.

Industry metrics still miss the point. Total value of assets issued onchain is the headline everyone cites, and it doesn't say how much of that value is actually working. A more honest measure would count tokenized collateral securing loans and stablecoin liquidity generated against those positions. The picture looks different, and less flattering, than the $16 billion headline.

An internet analogy fits. Digitizing a document changed little on its own. The change came from building networks and workflows around those documents, making them functional instead of merely digital. Tokenization is on the same path.

Parts of that infrastructure already exist. Figure PRIME's lending book on the DeFi protocol Morpho has crossed $200 million. Aave's Horizon, launched in August 2025, lets institutions borrow stablecoins against tokenized assets and has passed $250 million in total value locked.