Charlie Quant LabThe Risk-Free Squeeze
One number runs this page: what liquid stablecoin lending pays on-chain, minus the risk-free rate. When that gap goes negative, dollars have a reason to leave DeFi. When it stretches, borrowers are paying up for leverage. This page shows where the gap sits tonight, how unusual that is against the last year, and what each regime has moved with.
Updated 08 September 2026 · 20:38 UTCcapital flees on-chain0bp
parity with risk-free+260bp
leverage rebuilds
One year of the squeeze
Daily readings of the spread. Above the green dashed line, borrowers are paying a leverage premium. Below the red one, on-chain dollars pay less than a Treasury-backed rate and the carry case for DeFi is gone.
Who is paying this yield
The basket is tonight's largest single-asset USDC, USDT and DAI money markets. Only the base lending rate counts — token incentives are excluded because they are a marketing budget, not borrower demand.
| Money market | Asset | Base yield | Deposits |
|---|---|---|---|
| aave-v3Ethereum | USDT | 3.56% | $209M |
| fluid-lendingEthereum | USDC | 5.10% | $147M |
| aave-v3Ethereum | USDC | 3.61% | $146M |
| fluid-lendingEthereum | USDT | 4.70% | $131M |
| aave-v3Ethereum | USDT | 3.56% | $69M |
| justlend-v1Tron | USDT | 2.38% | $69M |
| fluid-lendingArbitrum | USDC | 3.98% | $62M |
| aave-v3Ethereum | USDC | 3.61% | $59M |
What each regime moved with
Every day in the recorded window was sorted by its spread, then matched with Bitcoin's 30-day realized volatility on the same date. This is a coincidence test, not a cause — it answers "what did the market look like on days like this", not "what will happen next".
| Spread regime | Days observed | Average BTC 30d realized vol | Share of days that were vol spikes* |
|---|---|---|---|
| <0bp | 276 | 43% | 24% |
| 0–200bp | 52 | 40% | 4% |
| >200bp | 7 | 40% | 0% |
*A vol spike is a day whose 30-day realized volatility sits in the top fifth of all observed days in the window. Windows differ per bucket because the spread series and the price series overlap only where both were recorded.
The spread and total DeFi value
If the squeeze is really a leverage gauge, it should swell when value pours into DeFi and shrink when it leaves. Both series below are rescaled to their own range so the shapes can be compared; the correlations are computed on the raw values.
Tonight's read
On-chain dollar yield sits 17bp above the risk-free rate — positive, but not the kind of premium that signals a leverage rebuild. This is the middle regime: carry exists, conviction doesn't. In the recorded window, days in this regime averaged 40% 30-day realized volatility — not the highest average Bitcoin volatility of the three spread buckets — with 4% of days in the top volatility quintile, not the largest share of volatility spikes (see the regime table).
Method, scope and what would make this wrong
What is measured. The on-chain rate is the TVL-weighted base lending APY of tonight's 8 largest single-asset USDC/USDT/DAI money markets (minimum $25M deposits each), as recorded by DeFiLlama's yields index at export time. Reward-token incentives are excluded. The history is an equal-weighted daily read of up to six of those pools, joined to the daily Secured Overnight Financing Rate published by the Federal Reserve Bank of New York via FRED (series SOFR, latest observation 2026-09-04).
Honest edges. Basket membership is tonight's — pools that were small or absent earlier still shape the back-history through their own recorded yields. Base APY is what lenders earned; it is not a guaranteed forward rate and can reprice within hours. SOFR is a secured wholesale rate, not a retail T-bill yield, so the spread is a market-plumbing comparison, not an offer comparison. The volatility table shows co-occurrence over one window (2025-09-08 → 2026-09-08); it contains one market cycle, not the average of all cycles.
Missing stays missing. Days with no recorded pool history or no SOFR print are skipped, never zeroed. If a source fails at export time, the affected figure is shown as not available and listed below rather than estimated.