Lending markets, meaning wrapped BTC deposited with a lending protocol as collateral or as liquidity for borrowers, now hold 146k of the 321k BTC of tracked balances, bridge escrow included, up from 79k of 264k two years ago. They gained 67k BTC while the total gained 57k, so the other groups together lost a net 10k. The largest loser was other contracts, contracts with no protocol label, down 16k. Idle wallet balances, the externally owned accounts with no protocol label, rose 10k to 113k but slipped from 39% to 35% of the total.
The rise was not steady. On 18 and 19 April 2026 lending balances fell 21k BTC while idle balances rose 17k. By September lending was back to 146k.
Where wrapped BTC sits, over time
Deposits in lending markets are not all collateral
The lending-markets group cannot separate collateral from liquidity supplied to borrowers. The collateral article measures the collateral side for Aave v3, Morpho Blue and Spark and finds 120k BTC pledged. The other 26k of the 146k is deposited without being pledged, held at a lending protocol outside those three, or held in labelled lending contracts that the collateral measure excludes, such as a protocol's treasury. Read 146k as deposits in labelled lending markets and 120k as the collateral measured at those three protocols.
One route in is Coinbase's loan product, launched in January 2025, which converts a customer's Bitcoin to cbBTC and posts it to a Morpho market on Base. Morpho Blue on Base holds 38k BTC of collateral, but any borrower can use those markets, so the balances do not show how much arrived through Coinbase.
Two days in April
Lending balances peaked at 147k BTC on 18 February 2026 and fell to 120k by 6 May. Most of that fall came in the two April days. The total barely moved in those two days. Over the next 17 days idle balances gave back 7k and the total fell 12k. cbBTC was net burned by 11k that month, its only large net-burn month (issuer flows).
The sequence is consistent with deposits withdrawn to wallets and part of them then redeemed. The data holds group totals, not transfers, and the burn is a monthly figure for one wrapper, so the lending outflow, the wallet inflow and the burn could belong to different holders. By June lending balances were back above 140k. On 13 September they stood at 146k, within 1k of the February peak, while idle wallet balances stayed 10k above their February level.
What "idle" includes, and how firm the labels are
"Idle" is a classification, not a statement about the holder. It covers every externally owned account (an address controlled by a private key rather than contract code) that has no protocol label. A custodian's hot wallet, an exchange's cold storage or a market maker's inventory counts as idle until someone labels it, so the group mixes private holdings with institutional balances that are not yet identified.
Label quality differs by group. Lending markets rest on the firmest labels, curated by hand or taken from each protocol's own contract registry. The custody, vaults and exchanges group is weaker, because part of its labels come from external aggregators. The 27k BTC in other contracts carries no label at all. If some of those contracts are lending protocols, the lending share is understated, and a later label would move the split without any token moving.
All labels are today's, applied to every past day. That is exact for Aave v3 and Morpho Blue contracts, which have had one role since deployment. A contract that changed role is counted in its current group for the days before the change.
Methodology
Full methodologyAuthor
Andrey Shivalin
Researches tokenized Bitcoin market structure: issuance and redemption, custody, collateral use, and the failure modes that connect them.
Working on tokenized BTC?
Commissioned research, measurement design and review on wrapped-BTC market structure.