Capital Layer
← all notes

2026-08-31 · Stablecoins & yield

Ethena USDe: a synthetic dollar that turned into a credit book

Data as of 31 Aug 2026.

Bottom line. The mechanism isn't broken — it's working right now, and that's exactly the problem. With funding around 10% annualised and 79% of days positive over the past year, sUSDe pays 4.2% on a 30-day average against 3.83% on 3-month T-bills. A 37 bp spread is what top-of-cycle looks like, not a temporary dip. In its current form, the yield doesn't compensate for the risk.

What it is

A synthetic dollar that, until 2026, earned the perp basis (long spot, short perp). It has since been rebuilt into a credit portfolio wrapped in a stablecoin.

Findings

1. The engine has been almost entirely swapped out

fig. 1 USDe: the engine is only 13.3% of backingamber — the basis people bought it for · red — credit · grey — cash and Treasuries

The basis trade — the reason anyone bought this — is 13.3% of backing. The rest is DeFi lending, institutional loans, liquid stables and RWAs.

Ethena dashboard, Backing Assets, 31 Aug 2026 17:00. Red is the part the product was built around.

The credit sleeve is 43.1% (DeFi lending plus institutional loans). In August it picked up a billion dollars in a separate Cayman vehicle alongside FalconX — roughly a quarter of total supply (CoinDesk, 19 Aug 2026). The risk has moved from funding rates into credit and duration.

fig. 2 What a sUSDe holder carriesbottom — what it rests on; red — credit and counterparties

2. Nearly half the backing is stuff you can buy directly

Liquid stables (31.1%) plus RWAs (12.3%) = 43.4%. A sUSDe holder carries smart-contract, exchange-counterparty and credit risk on a portfolio that is almost half repackaged cash and Treasuries. And those liquid stables yield 4.1% — 27 bp over the risk-free rate.

fig. 3 Funding ~10%, the holder gets 4.2%annual, % · grey — no protocol risk · amber — the engine's rate

3. The yield doesn't pay for the risk

sUSDe — Ethena dashboard, Market Data, 31 Aug 2026. T-bills — TradingEconomics, 31 Aug 2026.

Ethena publishes this spread on its own dashboard: through August it ranged from 0.00% to 1.80%, with some days at zero premium. In 84.6% of weeks the yield came in under 5%.

4. The protocol isn't betting on its own engine

Funding is running at 9.99% annualised across venues, 79% of days over the past year were positive, and Ethena holds only 0.78% of BTC open interest ($28.8bn). There's plenty of room to scale the trade — yet the basis is held at 13.3% and a third of the book is parked in stables at 4.1% (Ethena dashboard, Market Data, 31 Aug 2026). The team is looking at the same funding rate and doesn't trust it enough to lean in.

fig. 4 October 2025: one event, two printsleft — a venue with no redemption access · right — protocol level

5. One event, two prints — both true

October 2025: USDe printed $0.65 on Binance (−35%), while at the protocol level backing held and the price never went below $0.97. Both are correct: Binance had no direct mint/redeem access, so arb couldn't close the gap. The lesson isn't the number, it's the plumbing — peg resilience on any venue is set by redemption infrastructure, not by collateral quality.

What would change the call

WhenWhat to pullThesis survives if
30 Sep and 31 Oct 202630-day sUSDe yield and 3M T-bill rate — same day, two sourcesspread ≥ 200 bp on both dates
1 Oct 2026reserve fund in absolute terms (not %, which flatters while supply shrinks)reserve above $62.08m
15 Oct 2026basis share of backing; whether a report on the Cayman vehicle (borrowers, collateral coverage) existsbasis ≥ 30% and the report exists
15 Dec 2026equity-basis share and 30-day yieldequity basis ≥ 25% with yield ≥ 8% — the engine has genuinely restarted

For reference: a $62.08m reserve against $4.12bn supply is 1.51% — above the usual stress line for a delta-neutral stable (reserve under $50m with supply over $3bn). Proof of reserves dated 28 Aug 2026 came from three independent attestors.

Calls, kept separate from the conclusion

  • By 15 Dec 2026, equity basis will not reach 25% of backing with a 30-day yield of 8% or higher. Checkable on Ethena's Backing Assets and Market Data pages. If wrong, the engine really has restarted and this needs a fresh look.
  • By 31 Oct 2026, the sUSDe–3M T-bill spread will not exceed 200 bp on a 30-day average.

Unknowns

Obtainable:

  • What's inside the Cayman vehicle — borrowers, collateral coverage, tenors, rates. Without it, there's no way to say what backs a quarter of supply. Route: a governance-forum request or FalconX directly.
  • Institutional loan rates are published as a "4–7%" range rather than a figure — no precise number for 11.9% of the book.

Unknown to anyone:

  • How much capital equity perps can absorb before the premium compresses. The 14–17.5% rates exist precisely because the short side is empty. Nobody knows the capacity, Ethena included — it lists this as a risk itself.
  • How the structure behaves in a credit event. Every stress it has survived — October 2025, the Bybit hack in February 2025 — was about funding and exchange counterparties. There has been no credit event with the new backing mix. Sample size: zero.

Takeaway

At 37 bp of premium, there's nothing to be paid for. And this isn't a cyclical trough — conditions for the basis trade are already favourable, so this is the outcome under good conditions. The picture changes only if the T-bill spread holds ≥ 200 bp for two consecutive months, or equity and commodity basis reach ≥ 25% of the book at ≥ 8% yield.

Research for information only. Not investment advice.