A crypto market maker needs a $40 million working-capital line to fund inventory across exchanges. Its balance sheet is audited, its books are real, but banks don't understand the collateral and traditional syndication takes months. Half a world away, a freight forwarder waits 60 days for invoices to clear while payroll is due weekly. Both need credit — the kind banks have increasingly stopped extending.
On-chain private credit exists to fill that gap with stablecoin capital. In January 2026 CoinDesk called it the potential breakout use case for tokenization, and the growth numbers since make the case hard to ignore. But the sector also carries the scars of real defaults. This piece explains how the machinery works, how the two leading names — Maple and Centrifuge — differ, and what you are actually lending against when you deposit.
TL;DR. On-chain private credit channels stablecoins into actual loans: a protocol onboards a vetted borrower, structures the loan through a bankruptcy-remote entity, and tokens represent lender claims. The market is still small relative to the $1.5–2T global private credit industry the Financial Stability Board measured in May 2026 — about $7.0B distributed on-chain by rwa.xyz's July 2026 count ($18.9B under broader methodology, ~$33.7B all-time originated). Maple runs an overcollateralized book for crypto-native trading firms (~$2.5B TVL, ~4.7–5.6% yield via syrupUSDC); Centrifuge is institutional infrastructure powering JTRSY Treasuries, JAAA CLOs (~6.5%), Apollo credit and high-yield funds ($1.6–1.8B TVL). Senior money earns 5–12%; anything paying more is usually taking junior or first-loss risk. Defaults happened in 2022–2023 — tokenization reveals them faster; it never prevents them.
What is private credit — and why did it move on-chain?
Private credit is lending that bypasses public bond markets: a lender extends a loan directly to a company and holds it, privately negotiated, with no exchange ticker and usually no credit rating. Since the 2008 financial crisis, banks retreated from parts of direct lending while non-bank managers filled the gap — the FSB put the global market at roughly $1.5–2 trillion in its May 6, 2026 vulnerabilities report, cautioning it has never been tested in a severe economic downturn.
Blockchain adds three things to this model: settlement in stablecoins at any hour, fractional lender participation so a $200 depositor can sit beside a $20M fund, and transparent, timestamped reporting of loan performance. It does not add credit enhancement. The borrower can still default, the collateral can still be misvalued, and the token is only as strong as the legal claim behind it.
How does an on-chain loan actually get made?
The standard structure runs through four steps. A pool delegate or asset manager sources and underwrites a borrower. The loan is signed off-chain with a bankruptcy-remote special purpose entity, isolating the asset from the protocol. Tokens are issued against the pool — ERC-7540 asynchronous vaults, co-authored with Centrifuge and now merged into OpenZeppelin, are becoming the standard. As the borrower repays interest, token holders accrue; if the borrower misses payments, the default appears on-chain and workouts proceed through the legal documents, not through code.
Senior, junior and first-loss capital
Pools typically tranche risk. Senior capital is paid first and earns the lowest yield; junior or mezzanine capital absorbs losses above a threshold for a higher coupon; a first-loss position — often the pool sponsor's own capital — takes the initial hit. When you see "12% on-chain private credit," the crucial follow-up question is which tranche earns it. A senior yield near the secured-policy rate is a different product from a junior yield promising double digits.
Maple: a rebuilt lender for crypto-market infrastructure
Maple launched in 2021 as an undercollateralized corporate-credit venue and nearly collapsed in the 2022 contagion: Babel Finance and Orthogonal Trading defaulted on nine-figure obligations, and lenders took real losses. Rather than close, Maple restructured around overcollateralized, fixed-term facilities for crypto-native market makers and trading firms with audited balance sheets, with qualified custody through Zodia and direct deals via Maple Direct.
The model has since scaled: platform figures cited by AltStreet put Maple at roughly $2.54B TVL, about $4.59B in broader AUM and $11.27B originated across roughly 60 borrowers. FinanceFeeds reported $2.4B in active loans and syrupUSDC transfer volume of $4.98 billion by January 2026. Its core rate was about 4.70% APY per Maple's own API, and the permissionless syrupUSDC wrapper accretes roughly 5.6% per year. That is secured-financing economics, not distressed-debt yield — appropriate for the collateral.
The underwriting criteria are public in outline — audited financials, defined minimum net assets, operational track record and collateral posted into qualified custody — and the fixed-term structure matters more than it first appears. Because facilities mature on schedules, the pool can return capital as paper rolls off rather than depending on a new lender arriving to take out the old one. That is the structural lesson the platform extracted from 2022: a credit book can only promise liquidity its asset maturities actually support, and transparency about term lengths is preferable to a false promise of instant exits.
Centrifuge: the infrastructure layer behind the fund brands
Centrifuge does not run one lending book; it builds the rails asset managers use. Its homepage reported $1.8B-plus TVL and 1,768 tokenized assets as of September 2026. The flagship products are a directory of regulated finance: JTRSY, a tokenized Treasury product that crossed $1B in Q1 2026 and carries S&P's AAAf fund rating; JAAA, the first fully on-chain AAA-rated collateralized-loan-obligation fund managed by Janus Henderson, yielding about 6.5% and the fastest tokenized fund to reach $1B; ACRDX, Apollo's diversified credit fund enabled as Morpho collateral; and HYB, New York Life's first tokenized high-yield corporate strategy.
Centrifuge's older real-economy pools financed trade invoices, consumer credit, emerging-market fintech debt and real-estate bridge loans — the "lending to real companies" part of the sector's promise. In 2026 the company also became a Sky protocol agent, selected for Obex's inaugural cohort with a $2.5B mandate, and a preferred Coinbase tokenization infrastructure. More detail on our Centrifuge review.
Maple vs Centrifuge: the side-by-side
| Factor | Maple | Centrifuge |
|---|---|---|
| Role | Direct lender / credit book | Asset-management infrastructure |
| Borrowers | Crypto market makers, trading firms | Asset-manager funds, real-economy pools |
| Collateral model | Overcollateralized, fixed-term facilities | Varies by pool; SPV-isolated assets |
| Scale (2026) | ~$2.5B TVL / ~$4.6B AUM | ~$1.6–1.8B TVL |
| Retail access | syrupUSDC, permissionless | Wrapped deRWA tokens, select pools |
| Typical yield | ~4.7–5.6% | ~5–12% by asset and tranche |
| Default history | 2022 Babel / Orthogonal; restructured | 2023 Harbor Trade Credit; queues |
Worked example: lending $10,000 through the year
What does supplying stablecoins actually look like in practice? Say you deposit $10,000 USDC into Maple's secured facilities via syrupUSDC, accreting at roughly 5.6% per year — around the product's stated rate — with no additional contributions. Monthly balances, rounded:
| Month | syrupUSDC value | Cumulative interest |
|---|---|---|
| 1 | $10,047 | $47 |
| 3 | $10,141 | $141 |
| 6 | $10,284 | $284 |
| 9 | $10,428 | $428 |
| 12 | $10,575 | $575 |
The year produces about $575 — roughly $150 more than the same capital in a tokenized T-bill product at 2026 rates. That $150 is the compensation for a different risk stack: secured-financing exposure concentrated in crypto-market borrowers, with fixed-term facilities rather than direct bill credit. Whether the premium is adequate is exactly the judgment each lender makes; no table can do it for you. Note facilities are term products, so early exits may involve queues or secondary discounts even though the wrapper token transfers freely.
Who are the real borrowers — and what do they pay?
Despite the "real companies" framing, the largest active on-chain book finances crypto-market infrastructure rather than Main Street. The borrower mix:
| Borrower type | Example venue | Senior yield |
|---|---|---|
| Market makers / trading firms | Maple facilities | ~4.7–5.6% |
| AAA CLO strategies | Centrifuge JAAA | ~6.5% |
| Diversified corporate credit | ACRDX via Morpho | ~5–11% |
| US high-yield bonds | NYLIM HYB | High-yield index + |
| Trade invoices / receivables | Legacy Centrifuge pools, Huma | ~8–12% historical |
| Emerging-market fintechs | Smaller pools | ~10–15%, higher risk |
The sector's funding base also differs: where T-bill tokens compete with money-market funds, private credit competes with secured financing and credit funds. How it compares with tokenized government paper is covered in our tokenized treasuries vs DeFi lending analysis, and the head-to-head platform numbers sit in our Maple vs Centrifuge comparison.
What has actually gone wrong?
Three episodes deserve to be remembered. In 2022, Maple's undercollateralized crypto-corporate book collapsed with the industry it lent to; the platform survived by changing the model, not by pretending losses hadn't happened. In 2023, Centrifuge's Harbor Trade Credit series saw all remaining assets default and lenders with redemption requests waited months because the pool reserve was zero — the mechanics were publicly visible in real time, which is transparency, not protection. In June 2026, Goldfinch moved to wind down its Prime product after serious borrower losses, leaving its legacy pool down roughly 97% from peak.
The TradFi parallel is instructive: in Q1 2026, Blackstone's $82B BCRED gated $3.7 billion of redemption requests. Every private-credit vehicle, on-chain or off-, ultimately answers the same question — can lenders exit faster than the underlying loans can be sold? On-chain markets make the mismatch visible in minutes rather than quarterly statements.
How to read a pool before supplying capital
The dashboards exist, and the data is more granular than anything a private credit fund ships to its limited partners — but only if you know which fields answer the real questions. On rwa.xyz and the protocols' own pages, six items tell most of the story:
- Active assets versus cash reserve. A pool with a thin reserve repays redeemers only as loans mature; a healthy reserve signals ordinary redemption capacity.
- Weighted average maturity. Short, self-liquidating paper (trade claims, repo-style facilities) behaves differently from multi-year corporate loans.
- Borrower count and concentration. Sixty borrowers across diversified collateral is not the same risk as three borrowers, however blue-chip their names.
- Realized losses and recovery history — separate from the "99% repayment" marketing, which counts performing loans rather than lifetime pool results.
- Tranche structure and subordination size — how much capital stands between a senior position and the first dollar of loss.
- Redemption mechanics — instant, queued, or NAV-based, and whether a committed liquidity facility backs the queue.
If a pool's page can't answer these, that itself is data. The sector's transparency pitch only protects lenders willing to actually read it.
Risk-adjusted: how does private credit compare with Treasuries?
Strip the products to essentials and the trade is the same one every bond investor makes: roughly one to three extra percentage points of yield in exchange for credit exposure, longer effective duration of capital, and exit dependence on markets that are small. T-bill tokens give you sovereign credit with instant redemption in normal markets; private credit gives you a secured claim on operating businesses or market infrastructure and a multi-month horizon. A sensible portfolio treats them as layers — Treasury paper as the core, credit as a sized satellite — rather than as a winner-takes-all choice, because the one thing 2022 proved on-chain and off- is that correlation inside credit books arrives exactly when investors believed themselves diversified. Higher yield is payment for risk, not evidence the risk has disappeared.
Who should lend into on-chain private credit?
It fits stablecoin holders who want yield tied to actual borrowing demand rather than token emissions, and who accept credit risk, illiquidity windows and a multi-month or multi-year horizon. Start with the secured, senior end — Maple's syrupUSDC exists precisely as an entry product, detailed on our Maple review. Avoid junior tranches until you can read the underlying loan documents and interpret the pool-level loss data yourself, and compare the venues against the broader field ranked among the best lending protocols. If your alternative is a yield-bearing stablecoin, our Ethena vs Sky breakdown shows what those pay and how they're backed.
Frequently asked questions
What is on-chain private credit?
Lending arranged through blockchain protocols: a vetted borrower, a bankruptcy-remote legal entity holding the loan, and tokens representing lender claims. Supplying stablecoins earns interest from real loan repayments. The token is a wrapper, not a credit guarantee.
How big is tokenized private credit in 2026?
RWA.xyz counted about $7.0B distributed on-chain in July 2026; broader methodology shows $18.9B active and ~$33.7B originated, up ~180% YoY. The global market the FSB measured is $1.5–2T, so on-chain remains a rounding error of the whole.
What is the difference between Maple and Centrifuge?
Maple operates an overcollateralized lending book for crypto-market firms at roughly 4.7–5.6% via syrupUSDC. Centrifuge is infrastructure for asset managers — JTRSY, JAAA, Apollo credit, high-yield funds — plus historically real-economy pools. One is a lender; the other is rails.
What yields can you earn?
Senior money earns roughly 5–12% depending on collateral — Maple ~4.7–5.6%, JAAA ~6.5%. Historical junior strategies paid up to 15%. Double-digit headline yields typically mean first-loss or junior exposure.
Have there been defaults?
Yes: Maple's 2022 Babel and Orthogonal defaults led to restructuring; Centrifuge's 2023 Harbor Trade Credit default produced months-long redemption queues; Goldfinch began winding down Prime in June 2026. Transparency reveals defaults faster; it doesn't stop them.
Sources and further reading
- Financial Stability Board — Vulnerabilities in Private Credit (May 2026) — global market size and risk assessment.
- rwa.xyz — Private Credit dashboard — active loans, originations and pool data.
- CoinDesk — Private credit's breakout moment (Jan 2026).
- AltStreet — Goldfinch vs Maple — TVL, AUM, originations and default track record.
- Web3AI — Tokenized Private Credit Explained 2026 — platform-by-platform yields and loss history.
- Centrifuge — official site — live TVL and product documentation.
Model the yield, grade the credit
Project interest over your lockup window and compare a pool's risk profile against alternatives before you supply a dollar.
Related reading: Tokenized Treasuries vs DeFi Lending, Maple vs Centrifuge and the Best Lending Protocols.