The On-Chain Yield Curve

Bruce Bent's footnote

In 1971 a former banker named Bruce Bent, with a partner, Henry Brown, had an idea that sounded faintly absurd: pool together short-term government paper, sell shares in the pool at a stable dollar apiece, and pass through whatever the paper earned. They called it the Reserve Fund, and it was the first money market fund. The pitch was irresistible and, for thirty-seven years, essentially true — a Treasury yield in a form as spendable as cash, at a price that never moved. Savers got the return of government debt with the convenience of a checking account. The industry Bent started grew into the trillions and became so much a part of the plumbing that people stopped thinking of it as a product at all.

Then, on the morning of September 16th, 2008, the Reserve Primary Fund — Bent's own fund — told its investors that its shares were worth ninety-seven cents. It had been holding Lehman Brothers commercial paper, and Lehman had filed for bankruptcy the day before. Breaking the buck, the industry called it, in the tone you'd reserve for a phrase you hoped never to say out loud. A run followed within hours; the Treasury guaranteed the entire industry within days to stop it from spreading.

The lesson of that morning is the most useful thing I know about yield, and it is this: a yield number tells you what you will earn, and says nothing whatsoever about what you can lose. For thirty-seven years the money market fund quoted the first figure in enormous type and the second in a footnote nobody read. The return sat on one axis. The risk sat on another. And the industry only learned to draw the second axis on the single morning it needed it.

I have been thinking about Bruce Bent, because crypto has spent the past two years rebuilding his invention — and is quoting it exactly the way he did.

A yield, reborn

The thing the industry now calls a tokenized Treasury is, mechanically, a money market fund with a blockchain where the transfer agent used to be. BlackRock's BUIDL, Circle's USYC, Ondo, Franklin Templeton, WisdomTree, Spiko — all of them do what Bent did, which is hold short-dated government paper and pass the coupon through. As of this writing the category holds roughly $16B, up from about a hundred million dollars at the start of 2024, and it pays four to five percent, which is simply the Treasury bill rate wearing new clothes.¹

And it is, genuinely, one of the better things to happen to crypto in years. Not because the yield is high — it isn't — but because the dollar that carries it is now composable. A program can hold it; a protocol can settle in it; a lending market can accept it as collateral. Bent gave savers a Treasury they could write a cheque against. This gives software a Treasury it can build against. When people say the risk-free rate has come on-chain, this — not some token emission — is the real thing they are pointing at, and it deserves the enthusiasm it gets.

The shape of the curve

Above the floor, the curve steepens, exactly as a curve should. The whole tokenized real-world-asset market — Treasuries, private credit, gold, a first sliver of equities — comes to about $28B once you set stablecoins aside. Two of those tiers pay a yield you are underwriting risk to earn: the Treasuries at the floor, and on-chain private credit above them. Read the issuers and you can see the institutions arriving in order of caution — a sovereign-grade money fund from BlackRock at the top, a credit book from Maple a little lower, tokenized gold off to the side paying nothing at all.²

The largest on-chain real-world-asset issuers, by value locked (live, DefiLlama). Treasuries and money-market funds dominate the top of the book; a single private-credit manager, Maple, and two gold tokens round out the leaders. Stablecoins excluded.

The more telling picture is the trajectory. Two years ago essentially none of this existed on-chain; the risk-free floor was a rounding error and private credit was a handful of experimental pools. What followed is the cleanest hockey stick in crypto that isn't a memecoin — and, unusually, one drawn by treasurers rather than degens.

On-chain value of the two yield-bearing tiers, month-end, for the major issuers (live, DefiLlama). Treasuries (dark) were near zero until 2024; private credit (light) collapsed in the 2022 credit unwind, then rebuilt. Together the tracked issuers now carry roughly $18B — most of the market's ~$21B yield-bearing base.

The steepening

Sit above the Treasuries and you find on-chain private credit — Maple lending to trading desks, Centrifuge against receivables, Huma against invoices, Goldfinch, in its day, into emerging markets — paying eight to fifteen percent.³ It is the on-chain echo of the loudest trade in traditional finance, where private credit has swelled toward two trillion dollars and made giants of Apollo and Blackstone. The crypto version is smaller — a few billion of on-chain value, though the outstanding loan principal it has originated is several times that — but it has grown large enough to matter.⁴

The on-chain yield curve, indicative net yields, mid-2026 (researched — see sources). The floor is the tokenized T-bill rate; the premium is what private credit pays for taking real borrower risk. The step between them is not a free lunch — it is the price of the risk you have agreed to hold.

The extra points are not free money; they are the credit-risk premium, which is a polite way of saying you are now lending to someone who can fail to pay you back. And some already have. The category had small, private versions of 2008 in 2022 and 2023, when real-world borrowers stopped servicing their loans and their on-chain lenders rediscovered an ancient fact: a loan that has been tokenized is still a loan, and a loan that can go bad still does. The receipts are in the data — Goldfinch, once a flagship of the category, now shows barely more than a million dollars of value where it once carried hundreds of millions; the pool did not fail loudly, it simply stopped being paid back.²

The second axis

Lay the tiers side by side — four percent here, twelve there — and the market reads them as a menu, and picks by appetite. But that is reading one axis. Draw the other.

Every point on this curve is carry: you are being paid to hold a risk. And on-chain, the risk has a particular and under-appreciated shape. The Treasuries at the floor are safe in themselves — but the entire reason to tokenize them was to make them composable, and the instant a composable asset is posted as collateral in a lending market, it acquires a liquidation price. The private credit above them carries default. Search the whole curve, top to bottom, and you will not find a single point that pays you and protects you. The yield is plotted in fine detail; the fragility is not plotted at all.

It is Bent's fund again — the return in large type, the loss in a footnote — except the footnote is now a smart contract that sells your collateral at a pre-agreed number, or a borrower an ocean away who simply stops answering.

None of this is an argument against the curve. The money market fund was a good invention that ran for thirty-seven years before its footnote came due, and it is still, net, one of the great conveniences of modern finance. The point is narrower and, I think, more useful: the on-chain version has reproduced the yield with remarkable speed and has not yet reproduced the risk machinery — the stress tests, the sponsor backstops, the hard-won conventions — that the money-fund industry spent four decades and one bankruptcy building around it. We have the return. We have not had our September.

What comes next

So where do real-world assets go from here? The easy answer — more assets, more yield, equities and commodities and structured credit and everything else that isn't nailed down — is probably correct and probably beside the point. The yield problem is close to solved. Bent's invention has been ported cleanly, the second tier is filling in, and the third is coming.

Picture the curve honestly, then, on two axes instead of one. The horizontal axis is the one the whole market already draws — return, running from the four-percent floor to the low teens of private credit. The vertical axis is the one nobody has drawn: downside protection, running from fully exposed at the bottom to protected at the top. Plot today's instruments on that plane and the same thing jumps out every time. Every one of them sits on the bottom edge. The Treasuries are exposed the moment they become collateral; the credit is exposed the moment a borrower stops paying. The market has spread itself out beautifully along the horizontal and left the vertical entirely blank.

The same market, drawn on two axes instead of one — conceptual, not measured. The horizontal axis (return) is real and populated; the vertical axis (downside protection) is empty. Every instrument on today's on-chain yield curve sits on the bottom edge. The upper-right quadrant — yield with a protected floor — is unoccupied.

That empty upper-right is the whole story of what comes next. It is empty not because it is impossible but because no one has built it yet, and there are, in theory, only a few ways to fill it. You can tranche — slice the loss and sell it to whoever wants it, so the senior holder keeps a yield with a real floor beneath them; the mechanism that built structured credit in traditional markets, for better and for worse. You can insure — spend a slice of the yield to buy an explicit floor, the way an option does, so the protection is a line item rather than a hope. Or it is something not yet named. What every version has in common is that the protection stops being implicit. It stops being a footnote and becomes a term of the instrument — priced, visible, and paid for out in the open.

This is, in the end, the road the money market fund already walked, only compressed. Bent's industry spent thirty-seven years behaving as though the vertical axis did not exist, learned on a single morning in 2008 that it did, and then spent the years after building the machinery it had skipped — floating valuations for the riskier funds, liquidity gates, redemption fees, capital buffers — an entire apparatus for the loss it had never plotted. On-chain has ported the yield in two years. It has not ported the apparatus. It will build it either before its own morning, or after.

The unsolved problem is the one the money market fund needed thirty-seven years and a bankruptcy to confront: the risk axis. The most interesting question in this whole corner of crypto is no longer how to manufacture a yield on-chain — that is finished work. It is whether anyone will build the point on the curve that Bent's savers never knew they wanted until the morning they suddenly did: a yield you can hold without the footnote. Whether that turns out to be protected credit, or tranching that puts the loss somewhere explicit, or something not yet named, is genuinely open.

What is not open is that the curve, as drawn today, is missing an axis. Someone is going to draw it in. The only question is whether they do it before the buck breaks, or after.