Grouped by where the yield comes from. Every category has a switch; money markets and CDPs, which pay about 0%, are off by default.
Peaked at 83,000 BTC in early 2025 and lost a third since: airdrop-driven staking tokens emptied, and in August 2026 about 8,000 BTC of LBTC moved to options.
September 2026: farming and pools 8,700 BTC, carry 9,300, options 8,000, basis 5,600.
Farming and pools fell from 40,600 BTC to 8,700: strategy vaults from 28,900 (May 2025) to 2,200, points campaigns from 12,400 to 4,300. About 10,000 BTC of the November 2025 drop is DeSyn leaving DefiLlama.
1,843 BTC a year ago. Almost all the growth came after May 2026, and almost all of it is Kraken.
122,100 BTC in March 2025, about 89,800 now; 5 products have no history.
One BTC in Kraken's vault, where the dollars are borrowed and parked, and the same trade on Aave v4.
| You deposit | 1 BTC | $81,178 |
| Posted as collateral (83%) | 0.83 BTC | $67,378 |
| Dollars borrowed (64% of the collateral) | $43,122 | |
| Earned on the parked dollars (5.86%) | +$2,527 | |
| of which rewards (2.18%; funder not disclosed) | $940 | |
| Interest on the loan (3.25%) | −$1,401 | |
| Profit before the fee | $1,126 | |
| Operator's fee (25%) | −$282 | |
| What you keep | 0.0104 BTC | 1.04% a year |
| Without the rewards | 0.0017 BTC | 0.17% a year |
It matches the 1.01 to 1.03% the vault paid over the last 30 days.
| Market | Chain | Collateral | Borrowed | LTV | Rate |
|---|
Carry = (parking rate − loan rate) × LTV × share posted; the fee applies only when the carry is positive. Health factor = CF ÷ LTV; liquidation below 1, after a BTC fall of 1 − LTV ÷ CF. At the line Aave v4 repays enough debt to bring the position back to the 1.24 target and pays a 10% bonus: the last box is the share of the posted BTC that is sold and the share lost to the bonus.
Ranked by BTC deposited on 20 September 2026. Each product from deposit to exit: the flow of funds with links to the contracts and markets, who holds the keys, who pays the yield, the charts and the events.
Up to 1,758 BTC from an estimated 50 to 66 clients. Most of the 5% came, by our estimate, from Core's own hedge; when Core stopped paying, Maple wound down.
Grew in three steps as caps were raised, the last two after Curve raised its credit line, to a peak of 2,412 BTC on 19 December. In the February price gap an exit would have paid up to 21% below book value. Now 45% below the peak.
37,572 users, brought in by a button in the Kraken app. Deposits came when BTC fell, whatever the vault paid.
ether.fi Liquid BTC ran alongside from January 2025 (192 to 849 BTC). It repaid most of its $19.1M loan on 9 and 10 April 2025, after the trade turned negative, and the last $3M in June; with no debt and its BTC in points programmes, June and July 2025 count as farming. It borrowed again from August 2025.
| 1 · Kraken Bitcoin Vault | 2 · Yield Basis | 3 · Bitget bgBTC Earn | 4 · Midas mHyperBTC | 5 · ether.fi Liquid BTC |
|---|
Maple BTC Yield, the 2025 leader, is under Closed.
| Who | Role | Can change | Delay |
|---|---|---|---|
| One Kraken key | Kraken: the app, the built-in wallets and the kBTC wrapper | Mint, burn and freeze kBTC, blacklist addresses, upgrade the token | none |
| A 4-of-6 Safe, through a timelock | Admin of the Veda vault | Strategy permissions, fees, the payout address | 1 hour |
| Two strategist keys | Sentora: picks the loans and the parking. Its investors include Ripple and a fund chaired by Kraken's co-CEO | Anything on the allowed list | none |
| Two keys and a 2-of-4 multisig | Withdrawals | Cancel or stop withdrawals | none |
| 1-of-1 owners of RLUSD Main and PayPal USD Main | Sentora's vaults: at least 99.999% of the lending on both kBTC markets, which take 55.5% and 37.9% of the vaults' money | Fees; gate deposits and withdrawals at RLUSD Main; caps | none; caps 3 days |
| A Hastra hot wallet | Issuer of PRIME | Freeze PRIME, run cash redemptions | none |
| The markets' oracle | Prices kBTC as BTC | A kBTC depeg would not trigger liquidations | n/a |
| Morpho | Venue | Nothing: the markets are immutable | n/a |
Sentora met the SEC's crypto task force on 10 August 2026 about curator liability. SEC memo
Live, small and closed, with what each really paid.
| Product | Status | What it does | Real yield (BTC) | Size | Biggest risk |
|---|
Checked on 24 September: no large live carry product is missing. Morpho on 14 chains, Aave, Spark, Compound, Euler, Fluid, Kamino and Tydro were swept for BTC-backed dollar borrowers, and the 27 wallets with over $20M of debt were traced: all are single-owner books, none takes outside depositors on-chain. The largest is an unidentified institutional cluster with about 9,900 cbBTC against $364M of USDC on Aave and Morpho, funded through Coinbase and Paxos; then Galaxy Digital's desks, Abraxas Capital's funds, a Binance-funded whale cluster, and Nexo's own leg (1,031 cbBTC against $40M of USDS on Spark, the on-chain side of its CeFi earn program). About 120 BTC of small carry vaults (Upshift Sentora and Gamma, Concrete) sits in farming and pools, and Yearn's Katana vaults count as money markets. Beyond Ethereum, a second sweep on 24 September (Tydro on Ink, Aave v3 on Avalanche and Polygon, Frankencoin, Curve's crvUSD markets, Takara on Sei, Vesu on Starknet, Zest on Stacks) found one carry vault with outside depositors, Noon's WBTC vault on Vesu, added above; the largest position elsewhere, 913 kBTC against $45M of USDC on Tydro, is a Kraken-linked book: the kBTC comes from Kraken's Ink hot wallet and the dollars go back to a Kraken hot wallet. Not finished: Venus, Aave and Lista on BNB (Lista has about $116M of stablecoin debt against BTCB; the active borrowers found so far are single wallets, the largest $10.6M) and JustLend on Tron ($550M of BTC, no holder data). Real yield: share-price growth after fees.
Every wallet with over $20M of stablecoin debt against BTC on Aave, Spark, Morpho and Compound on 20 September, traced on 24 September through its funding and where the borrowed dollars went. Only Kraken's vault takes outside depositors; the rest are single-owner books. Addresses link to Blockscout.
| Wallet | Venue | BTC collateral | Debt | Who | Evidence |
|---|
| Vault | Last 30 days | Last 90 days | Last year | Advertised |
|---|
In 2025 and 2026 the losses came from the parked dollars and from keys, not from Bitcoin's price.
| Risk | Seen | What good looks like |
|---|
What the top five imply for a carry product on Babylon vaults.
| Case | Loan | With parking rewards | Without |
|---|---|---|---|
| Aave's USDC rate today | 4.4% | −0.05% | −0.45% |
| A supply programme on the hub | 3.0% | 0.55% | 0.21% |
| A subsidized loan | 1.0% | 1.40% | 1.06% |
| The hub runs short of dollars | 7.0% | −1.35% | −1.75% |
Parking: at least 70% in same-block venues at about 3.8%, up to 25% in rewarded vaults at about 5.6%, 5% in credit at about 5.1%.
| Who pays | Cost | For what |
|---|---|---|
| Stablecoin issuers, on Sentora's vaults | 3.0% | Demand for their stablecoins |
| The same, per dollar of Kraken's vault | 1.26% | More than all the yield Kraken's users get |
| Bitget, on its own product | 1.7-2.6% | A yield number for bgBTC |
| Yield Basis, in its token | 3-4% | Growth (14 to 21% in late 2025) |
| Rule | Limit | Seen in the market |
|---|---|---|
| Loan to value | 50% target, 60% ceiling | Kraken peaked at 79% against an 86% line |
| Reaction to a breach | Automatic, within an hour | Kraken repaid 40 to 48 hours late |
| Vaults per position | At least two | With one vault, one liquidation takes everything |
| Debt repayable in the same block | 70% or more; 90% within a day | Kraken 48% |
| Slow or operator-redeemed assets | 5% of debt at most | Kraken 31% in PRIME |
| Our share of the hub | Half its dollars at most; utilization under 85% | Kraken's own borrowing caused most of its rate spikes |
| Keys | 3-of-5 named signers; a timelock longer than the 3-day exit | Kraken 1 hour; Bitget none |
| Fee | 10 to 15% of gains above the high-water mark | Kraken's 25% takes a quarter of a 1.4% yield |
| Disclosure | Weekly: yield, rewards and their payers, LTV, our hub share, the repayment ladder | None of the five publish all of it |
| Role | First choice | Watch out for |
|---|---|---|
| Dollar supply | A stablecoin issuer paying for hub supply (USDG suppliers on Aave v4 get 3.3 to 3.7% in rewards today); GHO through Aave's DAO | Rewards with no named payer or end date |
| Distribution | Desks that stake client BTC with Babylon: Kraken first, then Binance, OKX, Figment | Without a desk, another $20-50M DeFi product |
| Strategy and risk | Steakhouse; Gauntlet | Gauntlet runs Bitget's loop; Sentora lends to itself; Re7 after Acre |
| Parking the dollars | Aave v3 USDC, sUSDS, Morpho vaults with forced withdrawal | Credit takes 1 to 2 business days to exit |
Figures as of 20 September 2026 unless marked.
Counting. Each BTC is counted once, at the product the holder owns. On-chain reads of Ethereum, Ink, Base, Morph, Optimism, Monad, Rootstock, Core, Bitcoin, HyperEVM, Stacks and Mezo at 20 September 2026, 12:00 UTC; Morpho, Merkl, Upshift and Midas APIs; DefiLlama for protocols, pools and history; Binance prices; project documents and filings.
Left out. Wrappers that earn nothing on their own (WBTC, cbBTC, kBTC, bgBTC and others), bridges, yield-trading venues whose BTC is counted elsewhere (Pendle), dollar products partly backed by BTC (Ethena, Falcon) and pools of $1M or less.
Categories and switches. Each category has a switch that takes it out of every market number. Money markets and CDP collateral earn about 0% and are off by default: 160,306 BTC supplied to lending markets (Aave, Morpho, Spark, Zest and others, directly or through vaults that only lend) and 4,242 BTC posted for stablecoins, counting only BTC that no product on the map holds. Credit funds and uncollateralized lending pools that pay a yield are counted, as credit. Lending BTC itself pays about nothing: on 24 September the BTC pools paying more than 0.5% a year held $25M in all. A link carries the choice, for example ?staking=off&mm=on.
Money markets and CDPs. DefiLlama's collateral plus unlent supply, less staking tokens counted at their issuer and positions of counted products. About 37,900 BTC of it is the cbBTC behind Coinbase's BTC loans. Method: mm_notes.md.
Judgement calls. Avalon CeDeFi (1,283 BTC against a USDT loan) is not carry: it is a pool for four institutional accounts, its dollars went to Binance, and nothing has moved since November 2024. Midas mHyperBTC is carry: its strategy wallet posts cbBTC and borrows dollars. Lombard's BTCe is credit, not carry: LBTC posted as slashable cover for a market maker's loan. Yield Basis counts what depositors own, not the borrowed half of its pools. Veda's other vaults are counted net of Lombard's LBTCv (885 BTC) and of the LBTC held by ether.fi's eBTC vault (247 BTC at the snapshot, up to 3,000 BTC in early 2025), which DefiLlama values inside Veda as WBTC and which the Lombard Vaults and LBTC rows already hold; the history nets both month by month. DEX pools count only BTC not counted elsewhere: without the staking-token side of a pair (LBTC, eBTC, uniBTC, pumpBTC, stBTC), LP held by Stake DAO, Beefy or Troves, and the Kraken vault's own Uniswap position.
Not verified. Who funds the RLUSD and PYUSD rewards; who controls the unlabelled keys; fees at Coinbase's fund and Lombard Bitcoin Earn; Gauntlet's pay from Bitget. Aave v4 and vault terms in the playbook are testnet values or from the governance proposal.
They pay a yield on BTC, but we cannot size them or do not count them. Sizes re-checked on 24 September: three turned out to be closed (Ledn's BTC account, Botanix stBTC, Core lstBTC), the exchange programs publish reserves but not Earn balances, and two Stacks products that pay in BTC are new and measurable.
| Product | Type | Yield, % | Size | Why it is not counted |
|---|
The map was checked pool by pool against DefiLlama's BTC yield pools; pools over $1M that were missing were added.
| Added | Category | BTC | Why |
|---|---|---|---|
| Uniswap v3, 9 BTC-pair pools | Farming and pools | 529 | Trading fees in BTC. Less the Kraken vault's own WBTC/kBTC position (268 BTC) and the staking-token half of the uniBTC and stBTC pools |
| Uniswap v4, 5 BTC-pair pools | Farming and pools | 438 | Less the LBTC half of the LBTC/cbBTC pool |
| Fluid DEX, 3 BTC-pair pools | Farming and pools | 252 | The pools also serve as loan collateral. Less the eBTC and LBTC halves |
| Curve, 4 BTC pools | Farming and pools | 166 | Less LP staked through Stake DAO or Beefy (counted there) and the LBTC and pumpBTC parts. Two more pools are fully staked through Convex and Stake DAO |
| GMX v2 BTC-only pool (GM BTC/USD, Arbitrum) | Farming and pools | 131 | LPs take the other side of perp traders for fees: 6.9% now, 5.2% over 30 days. It held about 500 BTC a year ago |
| Across WBTC pool | Farming and pools | 115 | Bridge liquidity for relayers: 0.6% plus 0.8% in ACX now, 1.3% over 30 days |
| Bancor v3, Chainflip, Orca, Ekubo, Hydration, Aerodrome, BeraPaw | Farming and pools | 299 | Single-sided WBTC (Bancor, paying almost nothing), native BTC (Chainflip), Solana and Starknet BTC pairs, the tBTC Omnipool, the cbBTC half of a cbBTC/LBTC pool and the BTC half of a WBTC/WETH pool |
| Wildcat: loans to Wintermute and two smaller borrowers | Credit (lending) | 161 | Uncollateralized BTC loans at 4 to 4.5%. Snapshot only: DefiLlama's history counts only undrawn BTC |
| Multipli xWBTC | Basis | 47 | WBTC routed to delta-neutral strategies (Multipli names Nomura, Fasanara and Edge Capital; not verified) |
| Added in all | +2,139 |
Seen but not added. Pools of $1M or less, such as Makina's Dialectic BTC. Pendle markets on mHyperBTC, whose BTC is already counted. Solv's base SolvBTC reserves ($523M at 0%), which DefiLlama lists under Solv Basis Trading: a wrapper. Concrete's Berachain vaults ($38M at 0%), which are not in Concrete's own TVL. A Bluefin pool on Sui that opened after the snapshot. Aggregator vaults whose use of BTC we could not check (Belt, Autofarm, Yield Yak and others) stay in farming and pools, not in lending.
Every one of the 88 rows that comes from a DefiLlama series equals DefiLlama's 20 September point to the BTC (the only differences are the double counts this study nets out). Read again on 24 September, the median row moved by 0.3%; the table shows the largest rows, the five carry products read on-chain, and the three rows that moved more than 10%. Sizes on the page stay at the 20 September snapshot.
| Row | 20 Sep, BTC | 24 Sep, BTC | Change | Source and note |
|---|