A new Bank of Canada staff paper examines Aave V3's revenue model, margin trading behaviour, and liquidation dynamics, concluding that decentralised lending works but carries systemic risks that regulators should watch closely.
A Central Bank Turns Its Attention to DeFi
The Bank of Canada has published one of the most detailed institutional analyses of decentralised finance lending to date. The staff analytical paper, authored by Jonathan Chiu of the Bank's Banking and Payments Department and Furkan Danisman of the University of Toronto, uses transaction-level data from Aave V3 to examine how the protocol generates revenue, how borrowers use leverage, and what happens when liquidations cascade through the system.
Aave V3 is the largest decentralised lending protocol by total value locked, holding approximately 34 billion US dollars in smart contracts as of mid-2025. That figure represents roughly a quarter of all value locked across decentralised finance and half of the DeFi lending sector specifically. The protocol operates across Ethereum, Polygon, Avalanche, Arbitrum, Optimism, Base, and BNB Chain, accepting collateral ranging from stablecoins like USDC and USDT to wrapped assets like WETH and WBTC.
The paper arrives at a moment when regulators worldwide are still working out how to classify and oversee DeFi protocols. Rather than making policy recommendations, the authors focus on empirical observations drawn from on-chain data spanning January 2023 to May 2025.
Revenue Is Concentrated in a Handful of Tokens
One of the paper's headline findings is that Aave's revenue is heavily concentrated. Although the protocol supports dozens of tokens, just three of them — WETH, USDT, and USDC — account for approximately 83 percent of total platform earnings from lending. This concentration persists despite tokens like weETH and wstETH having much larger supply shares, because those assets have very low utilisation rates and therefore generate minimal revenue.
The protocol's overall net interest margin sits at just 0.64 percent, compared with 2.48 percent for major US banks and 1.69 percent for major Canadian banks. The authors attribute this gap partly to Aave's overcollateralised model eliminating credit risk, and partly to the absence of the overhead costs that traditional banks must fund from their margins — salaries, branches, compliance teams, and customer support infrastructure.
Perhaps the most striking comparison is on bad debt. Aave V3 reported zero non-performing loans during the study period, versus 0.59 percent for US banks and 0.65 percent for Canadian banks. The protocol's automated liquidation mechanism resolves undercollateralised positions before they can become bad debt, shifting the risk entirely onto borrowers rather than depositors.
A Fifth of All Borrowing Is Leveraged Margin Trading
The paper's second major finding concerns how borrowers actually use the protocol. Despite the requirement that all loans be overcollateralised, the authors developed a methodology to identify recursive leverage — the practice of borrowing, swapping borrowed funds into more collateral, and borrowing again to amplify exposure to a particular asset.
Using this approach, they estimate that margin trading accounts for roughly 20 percent of all borrowed volume on Aave V3 and about 8 percent of total borrowing transactions. The average size of these margin trades is significantly larger than regular borrows, suggesting that this activity is driven by sophisticated, large-scale participants rather than typical retail users.
Margin traders represent just 2 percent of all active users on the platform, yet they account for a disproportionate share of activity. They borrow four times more frequently than non-margin traders, maintain positions roughly twice as long, and rely on flash loans at four times the rate of other users. They also carry lower health factors — the metric Aave uses to measure how close a position is to liquidation — making them substantially more vulnerable to forced liquidation during market downturns.
Liquidations Come in Waves
The third pillar of the paper examines liquidation dynamics. The authors document that liquidations on Aave V3 are highly clustered, occurring in concentrated bursts rather than being evenly distributed over time. These waves are strongly correlated with sharp declines in ETH prices.
Four major liquidation events occurred during the study period. The largest, in August 2024, saw 258 million US dollars liquidated across 1,193 affected users in a single wave. Just four collateral tokens — WETH, wstETH, WBTC, and weETH — accounted for approximately 90 percent of all liquidated value across the entire study period.
The paper estimates that the total cost to borrowers from liquidation events — including the liquidation penalty paid to third-party liquidators and the opportunity cost of missed price recoveries — can range from 10 to 30 percent of the liquidated value. Large borrowers are disproportionately affected during these peak events.
Insurance Coverage Is Thinning
One underreported finding in the paper concerns Aave's Safety Module, which functions as a decentralised insurance backstop. Users can stake AAVE tokens into the module, and in the event of a protocol shortfall, up to 30 percent of those staked tokens can be slashed to cover losses.
However, the effective insurance per dollar of deposits has been declining. At the launch of Aave V3 in early 2023, there were approximately 10 cents of Safety Module coverage for every dollar supplied to the protocol. By mid-2025, that figure had fallen to around 2 cents. The protocol's deposits have grown much faster than the capital committed to its insurance mechanism, meaning the backstop is becoming thinner relative to the risk it is supposed to cover.
What This Means for DeFi
The paper's conclusions are measured. DeFi lending with proper governance is operationally viable, the authors write, but it faces real constraints around capital efficiency, liquidation risk, and systemic fragility within the crypto ecosystem. The overcollateralisation requirement that eliminates credit risk also limits how much capital can actually be put to productive use. The automated liquidation mechanism that protects depositors can impose severe losses on borrowers during volatility spikes.
For Aave and similar protocols, the research highlights a tension at the core of the DeFi lending model: the features that make it trustless and transparent are the same features that make it rigid and occasionally punishing. The system works, but it works within tight boundaries set by code rather than human discretion.
The full paper, titled 'DeFi Lending: Returns, Leverage, and Liquidation Risk,' is available through the Bank of Canada's staff analytical papers series with the reference SAP 2026-13.
