UNDER DEVELOPMENT
sat rates
DAILY FIXING
No 059 · 2026-10-17
08:20 UTC · methodology 2026.08.1

Are you actually paid for giving up your keys?

Two orthogonal risks are traded off constantly and almost never named. Lightning pays you to accept hot-key risk with no counterparty. Lending pays you to accept counterparty risk with cold keys. Plotted against a ladder of custody arrangements, the best attainable bitcoin yield at each rung answers the question directly.

EFFICIENT FRONTIER — BEST ATTAINABLE BTC YIELD BY CUSTODY TIER, 3M

The tier axis is ordinal. Five stations, tick marks only, no ruler between them — T4 is not twice T2.

Efficient frontier in custody-tier and yield space, 2026-10-17. INDICATIVE ONLY — 1 OF 4 STATIONS RESTS ON AN OBSERVATION% P.A. — BTC-DENOMINATED, 3M TENOR−22460IF CUSTODY RISK WERE PRICED — HYPOTHETICALIMPLIED 3M (CIP)STATED ANCHOR — ASSUMED, NOT MEASUREDBEST ATTAINABLE — THE FRONTIERSLOPE −0.49 PP / TIERINDICATIVE ONLY — 1 OF 4 STATIONS RESTS ON AN OBSERVATIONT1SELF-COLDT2SELF-HOTT3COLLAB.T4CUSTODIALT5REHYP.CUSTODY TIER — ORDINAL; SPACING CARRIES NO METRIC
Interpolated — inferred, never observed Filled — a transaction
TENOR · 3M STATIONS WITH A SOURCE · 4 OF 5 SLOPE · −0.49 PP/TIER

WATERMARKED EXPORT — PNG 2× · SVG · CARRIES THE DATA DATE 2026-10-17, NOT THE EXPORT DATE

The result

The fitted slope is −0.49 percentage points per tier, and the verdict this site publishes is INDICATIVE ONLY — 1 OF 4 STATIONS RESTS ON AN OBSERVATION.

That is deliberately weaker than the number looks. The slope is fitted across 4 stations: 1 rest on a market observation, 1 is a stated assumption and 2 are interpolated. Excluding the assumed anchor it is -0.894 pp/tier. A single station moving changes the direction, so this is reported with its composition rather than as a standalone finding. A frontier fitted mostly through an assumption and an interpolation can point downward today and upward tomorrow without any market having moved, so it is reported with its composition rather than as a finding. The shape below is worth looking at; it is not yet worth citing.

Read the dashed line as the counterfactual: what the ladder would look like if each step up in custody risk were paid for at a plausible ≈0.75 pp per tier. It is drawn from the safest station's level and labelled as hypothetical, because nothing observed supports it.

What is on each station

  • T1 — Self custody, cold. 0.00% p.a. — stated anchor. Bitcoin's own storage cost makes the true risk-free rate slightly negative. v1 anchors at zero and says so; this station is an assumption, not a measurement.
  • T2 — Self custody, hot. 0.66% p.a. — magma filled. Median of 114 cleared transactions at this term.
  • T3 — Collaborative custody. −0.87% p.a. — implied (cip). Interpolated between listed expiries and drawn as such. Venues: cme.
  • T4 — Custodial. −1.13% p.a. — implied (cip). Interpolated between listed expiries and drawn as such. Venues: deribit, okx, bybit.
  • T5 — Custodial with rehypothecation. No source. No source on this site currently discloses rehypothecation clearly enough to be placed here. That is a disclosure failure in the market, not a gap in the data collection. The station is left empty rather than filled by interpolation: interpolating across tiers would imply an exchange rate between hot-key risk and counterparty risk, and no such rate exists.

Why the axis has no ruler

The custody tier is ordinal, and spacing carries no metric. The five stations are equally spaced because they have to be drawn somewhere, not because the distance between them is equal. There is no unit in which the step from collaborative custody to full custody is the same size as the step from cold storage to a hot key. Everything on this page is a ranking, and it is drawn as one.