Capital Layer
← all notes

2026-10-22 · DeFi & lending

USDG on Aave V3 Ethereum pays suppliers 23%: a coin problem or a utilisation spike in a leftover pool

Data as of 23.10.2026.

In short. USDG itself is fine: price $1.0000, market cap ≈$3.16bn (CoinGecko, 23.10.2026). The 23% is a local story about one reserve. Suppliers left Aave V3 for V4, borrowers stayed, pool utilisation climbed to 96.6%, and the steep part of the rate curve produced a spike. For a supplier the opportunity is mostly an illusion: the spike lasts hours, a new deposit kills it on its own, and exiting the pool is close to impossible exactly at that moment.

What it is

The USDG reserve (Global Dollar, issued by Paxos) in the Aave V3 Ethereum Core pool. USDG can only be borrowed here; it cannot be used as collateral in this pool (listing ARFC and Snapshot vote, 09.12.2025). Since April 2026 most USDG liquidity on Aave has moved to Aave V4 Core, so the V3 reserve has become a "leftover" pool on the old protocol version.

Keep the asset and the wrapper apart. The underlying is USDG. A supplier holds aUSDG, a claim on the V3 pool. How fast that claim turns back into USDG is set by the pool's free liquidity, not by the coin's price.

Three questions, straight answers

Is something wrong with the coin? No. USDG trades at $1.0000, 7-day low $0.9992, market cap ≈$3.16bn (CoinGecko, 23.10.2026). Issuer reserves are cash and short-dated US T-bills, with the cash held at DBS Bank. Supervision is by MAS in Singapore; in the EEA USDG is issued as e-money under MiCA (Aave ARFC, 04.12.2025). The market shows no stress either: about $3.2bn and +340% year on year (Crypto Briefing, late September 2026).

Is it just heavy local borrowing? Yes, but the cause sits on the supply side, not the borrow side: suppliers left, borrowers stayed. The V3 pool holds 3.58M USDG, of which 3.46M is borrowed — 96.63% utilisation (Aavescan, 23.10.2026). On 01.04.2026 deposits were still 55.6M. The money moved to Aave V4 Core, which now holds 89.3M USDG supplied and 66.0M borrowed (Aavescan V4, 23.10.2026).

Fundamental problem or opportunity? The problem is local, and for a supplier the opportunity is mostly an illusion, for three reasons covered below: the spike is short, entry kills the rate, and exits are blocked exactly when the rate is high. For borrowers it is the opposite: anyone borrowing USDG on V3 pays 8–9 times what V4 charges (3.28–3.70%).

FIG. 1 Past 80% utilisation the rate takes off rate, % APR · curve rebuilt from Aavescan figures, 23.10.2026 0% 10% 20% 30% kink 80% borrow 29.94% supply 23.15% at 80% supply ≈3.2% 0 40 80 100 pool utilisation, % · now 96.63% grey — borrower · red — supplier

Why 23%: the curve mechanics

Aave sets rates off a curve of pool utilisation (borrowed as a share of supplied). At the USDG listing (ARFC, 04.12.2025) the parameters were: base 0, kink at 80%, slope above the kink 50%, reserve factor 20%. On 06.03.2026 risk provider Chaos Labs proposed cutting the post-kink slope to 30%. Their reasoning: two large borrowers with "sticky" debt, a large supplier had left, and utilisation had already peaked at 88%.

Current figures fit a 30% post-kink slope and a ≈5% pre-kink slope: 5 + 30 × (96.63 − 80) / 20 = 29.95% borrow rate versus 29.94% on Aavescan. The supply rate is borrow rate × utilisation × (1 − reserve factor): 29.94 × 0.9663 × 0.8 = 23.15%, an exact match. The 5% pre-kink slope is inferred from these figures, not read from the contract.

The steep section is doing what it was designed to do: a high rate is meant to pull in new suppliers and push out borrowers until the pool is back below the kink.

A pool in wind-down

USDG supplied on V3: 55.6M (01.04.2026) → 21.9M (01.09) → 6.17M (01.10) → 3.58M (23.10), i.e. −94% since April. Over the same period V4 Core grew from 0.15M to 74.6M on 01.10 and 89.3M on 23.10 (Aavescan V3 and V4, 23.10.2026). The successor already holds ≈96% of USDG supplied on Aave.

There is no forced migration from V3 to V4: Aave Labs dropped it after DAO pushback (eco.com, The Block, March 2026). So V3 will wind down slowly, and utilisation will jump every time someone moves an amount that matters against a few million.

FIG. 2 Liquidity moved from V3 to V4 USDG supplied, M · Aavescan V3 and V4 Core, 23.10.2026 01.04.2026 V3 55.6 V4 0.15 01.10.2026 V3 6.17 V4 74.6 23.10.2026 V3 3.58 V4 89.3

Three reasons 23% is not a yield

1. The spike lasts hours. Daily supply-rate snapshots: 19.10 — 15.68%, 21.10 — 3.01%, 22.10 — 3.13%, 23.10 — 23.15%. Average V3 supply rate ex-incentives: 1 day 4.84%, 7 days 6.58%, 30 days 3.78%, 6 months 4.74% (Aavescan, 23.10.2026). The 30-day average sits below the 3-month US T-bill: 4.05% discount rate and 4.15% coupon-equivalent on 08.10.2026 (US Treasury). Only the 7-day figure beats the bill — that week had three spikes. These averages are built from snapshots and exclude incentives; the cleaner measure would be realised yield from the reserve's income index (share price), which was not pulled on-chain.

FIG. 3 Headline rate versus realised averages USDG supply rate, Aave V3, % · Aavescan, 23.10.2026 now 23.15 1 day 4.84 7 days 6.58 30 days 3.78 6 months 4.74 dashed — 3-month US T-bill, 4.05–4.15% (08.10) averages — from snapshots, ex-incentives

2. Entry kills the rate. Bringing utilisation back to 80% takes only ≈0.75M USDG of new supply. Along the curve: a new 0.1M deposit — utilisation 94%, supply rate ≈19.6%; 0.3M — 89%, ≈13.4%; 0.75M — 80%, ≈3.2%. Pool deposits already swing between 3.53M and 4.32M over 16–22.10: someone is moving 0.5–0.8M in and out daily and is likely already harvesting these spikes.

3. The exit is blocked exactly when the rate is high. Only 3.58 − 3.46 ≈ 0.12M USDG is free. A larger position waits for loans to be repaid or new suppliers to arrive. The post-kink slope was cut from 50% to 30% precisely to keep borrowers in the pool — so after a spike the curve pushes them out more weakly, and a supplier waits longer to get out.

FIG. 4 A new deposit kills the spike by itself supply rate after the deposit, % · computed along the curve no deposit · 96.6% 23.15 +0.1M · 94% ≈19.6 +0.3M · 89% ≈13.4 +0.75M · 80% ≈3.2 free liquidity available to exit ≈0.12M of 3.58M — the rest is lent out

The coin and the issuer

7-day price range $0.9992–1.0000; all-time low $0.9076, undated (CoinGecko, 23.10.2026). In October 2025 USDG traded below 0.97 on some centralised exchanges while the Curve pool held near 1.003 (ARFC, LlamaRisk). Issuer weak spots flagged by risk providers: cash at a single bank (DBS), redemptions up to T+5 with KYC, upgradeable contracts with MPC keys and a 3-hour timelock on admin changes only (ARFC, 04.12.2025). Curve depth for a 3.6M sale was assessed very differently: 9.28% slippage per LlamaRisk versus ≈1% per Chaos Labs. That disagreement concerns selling USDG on the market, not the Aave pool.

Block admission: - USDG (Paxos) as an issuer — yes, with a limit: as a dollar within normal per-protocol limits. What revokes it: a daily close below 0.99 · a delayed reserve attestation · regulator action (MAS or FIN-FSA) · a freeze of the pool contract · redemptions slower than T+5. - USDG reserve on Aave V3 Ethereum — no for new deposits: pool in wind-down, ≈0.12M of free liquidity, realised rate below the T-bill. What would flip the "no": over 20M supplied for 30 days straight and utilisation below 85%. - USDG on Aave V4 Core was not analysed separately.

Why borrowers don't move to the cheaper V4 (3.28–3.70%) is unknown. One hypothesis: their collateral is locked in V3 positions, for example a PT-USDG E-mode loop (ARFC PT-USDG-28MAY2026 on V3 Core). Another: with a 30-day average borrow rate of 6.06%, migrating doesn't pay. Neither is verified.

Positioning

Niche: dollar stablecoins from regulated issuers. USDG at ≈$3.16bn (CoinGecko, 23.10.2026) is about 1% of a ≈$316–323bn market (DefiLlama, May–June 2026); leaders are USDT ≈59% and USDC ≈24% (June 2026). Within Aave the V3 reserve is now small: 3.58M versus 89.3M on V4 Core. Market: global; stablecoins at ≈$316bn on 12.06.2026 versus $308bn at end-2025 (DefiLlama), no fresher October figure found; USDG is growing faster than the market, +340% year on year. Market stage: the two objects sit at different stages. USDG as an asset — scaling (medium confidence): the coin is validated and outgrowing the market. The USDG reserve on Aave V3 — transition and exit (high confidence): its successor, V4 Core, already holds ≈96% of USDG supplied on Aave. Macro stage: macro regime supportive but fragile (reading of 19.07.2026, stale — no fresher one); macro is secondary here because the rate is driven by pool utilisation, not the Fed; the 3-month US T-bill was 4.05–4.15% (08.10.2026). Timing conclusion: late — for new deposits into V3: the pool is winding down, spikes are short, and someone moving 0.5–0.8M in and out daily is already harvesting them.

What would prove this wrong

  • "The opportunity is real" — if on 22.11.2026 the 30-day average V3 supply rate (Aavescan, ex-incentives) is 6% or higher. Then the reading "curve plus arbitrage flatten the spikes" is wrong. Now 3.78%.
  • "The pool revives" — if on 22.11.2026 USDG supplied on V3 exceeds 10M. Now 3.58M.
  • "The coin has a problem" — if before 22.11.2026 there is a single daily USDG close below 0.995 (CoinGecko). None — the hypothesis is dead for good.
  • V4 to V3 ratio — on 22.11.2026 V4 deposits are expected to be at least 10 times V3 (≈25 times on 23.10).

What we don't know

  • Who the borrowers are (in March Chaos Labs described two large ones) and who moves 0.5–0.8M in and out daily. On-chain Borrow and Supply events would show it.
  • The exact curve parameters from the contract: the 5% and 30% slopes are inferred from Aavescan figures, not read.
  • Realised yield from the share price over 30 and 90 days: Aavescan averages are snapshots.
  • Whether the DAO plans to freeze the USDG reserve on V3 or cut its caps. Nothing of the kind was found on the Aave forum (search on 23.10.2026) — only the listing and the 06.03.2026 curve change.

The general lesson. A high APR on a "leftover" reserve of an old protocol version is pool utilisation, not a coin problem. A supplier earns the spike rate for a few hours and carries the risk of being stuck for weeks. A borrower in such a pool, on the other hand, should check whether the same loan is cheaper on the newer version.

Research for information only. Not investment advice.