If Sky wins, SKY wins — at half-rate, and last in line. Sky is a real, profitable, transparent dollar franchise: ~$430M of gross revenue at a run-rate, five straight surplus quarters, ~$11.2B of USDS and DAI on-chain, and institutional money alongside it (~$1.23B with Janus Henderson, ~$713M in BlackRock’s BUIDL). Since 13 August 2026 it does something most governance tokens never do — Stage 2 routes 50% of monthly net surplus to SKY stakers, 27.5% buying SKY on the market (5% of it burned) and 22.5% paid in USDS. But the token sees the last line, not the first. Savers and Agents are paid before the surplus exists, so only about 15 cents of every revenue dollar reaches holders: ~$62M a year against a ~$1.5B market cap, a ~5.3% yield on staked SKY versus ~3.5% on sUSDS, which carries no equity risk. Two points for taking first loss — and first loss is literal here, because reserves of ~$82–92M sit against an ~$11B book and anything beyond them is paid by minting new SKY. The payout itself is a parameter, not a covenant: it went 75% → 7.5% → 50% in six months. This report takes both halves seriously: a well-run bank, and a junior, discretionary claim on it.
One mechanism, many owners
MakerDAO invented the decentralised dollar in December 2017, and the mechanism has not changed since: over-collateralised debt issues a stablecoin, and a governance token both votes and recapitalises. What changed is the collateral, the owners and the name. a16z bought ~6% for $15M in 2018; Paradigm and Dragonfly ~5.5% for $27.5M in 2019; both were fully out by mid-2023, at roughly 7.6–15×. Backing shifted from crypto vaults to USDC and T-bills, which is why the March 2023 USDC de-peg dragged DAI with it. In August–September 2024 Maker became Sky, USDS and SKY launched at 1 MKR = 24,000 SKY, and SKY touched $0.1005 in December. Today it trades near $0.065 — roughly where the buyback programme itself has been buying.
One pattern repeats across that history, and it is the reason this report insists on the word junior: when something breaks, the token holder pays. On Black Thursday in March 2020, zero-bid liquidations created bad debt and new MKR was minted and sold to cover it. In February 2025 an emergency vote loosened borrowing against the governance token within days, framed as defence against a takeover — the cost was trust in process. And in March 2026, with reserves stuck near $50M and a coming “oil shock” cited, buybacks were cut ~87% from around $300K a day. There is a charitable reading of that last one and it is a fair one: reserves before payouts is exactly what S&P asked for, and reserves did rise 60–80% within five months. There is also a sceptical reading: for a year buybacks were paid while the capital cushion stayed thin, and when stress arrived, holder cash was the first thing cut. Both are true at once, and together they define what SKY is — the protocol’s shock absorber, not a fixed claim.
From vaults to a capital network — and the part that is not disclosed
The balance sheet today looks nothing like 2017’s. Crypto vaults are the smallest leg; Sky Agents — Spark, Grove, Keel, Obex, Osero — borrow USDS and deploy it into T-bills, credit and DeFi, holding $6.84B of $12.32B of collateral as of June 2026. Sky has effectively become a wholesale lender to its own franchisees. The rest of the architecture is a small number of parts doing a lot of work: USDS and DAI are ~$11.2B on-chain and idle USDS is pure margin; sUSDS is the savings token at ~3.5%, ~$4.7B, and it is simultaneously the growth tool and ~80% of expenses; the Peg Stability Module swaps 1:1 with USDC, which holds the peg and imports Circle’s risk; the Smart Burn Engine buys SKY on the market daily with the stakers’ share, limited by surplus rather than by capacity. Staking SKY for ~4.2%, and borrowing USDS against it, is the token’s only real utility beyond voting.
Two things follow. First, the growth is being bought. USDS nearly doubled year on year while revenue grew 10.5%, and new USDS came in mostly through the savings rate — when the rate was cut in July, ~$1.2B left sUSDS in a month. So the KPI that actually matters is not supply but the share of USDS that does not earn the savings rate. Second, the disclosure is incomplete in the place it matters most: $2.58B of the Agent book sits with named institutions, and what the other ~$4.3B holds is not clearly disclosed. Nor is the reserve figure settled — SFF reported $50.90M in March and $82.40M in June, the August update shows $76.99M (a fall), and Sky’s 16 September post says $92.19M. We could not reconcile the gap, and we say so rather than pick the flattering number. Commercially, Sky is the #3 dollar at ~4% of a ~$303B market behind USDT’s $183.4B and USDC’s $74.2B, and its edge — the savings rate and DeFi integration — is precisely what banks’ deposit tokens and stablecoin-yield regulation are aiming at. The CLARITY Act draft, which would have banned deposit-like stablecoin yield, stalled 49–50 on 15 September: a reprieve, not a resolution.
Yield, not scarcity — what a dollar of surplus costs
SKY has no dividend and no fee switch. It receives a governance-set share of surplus, and the order of operations is revenue, then costs, then reserves, then payout — the token is last in that line by construction. Stage 2 splits the monthly net surplus as 27.5% buying SKY for stakers, 5% of that bought supply burned, 22.5% paid to stakers in USDS, and up to 50% retained in the Surplus Buffer until reserves reach the $150M floor. On ~$125M of annual surplus against a ~$1.52B cap, that is ~$62M of cash to holders and ~$6M (~96M SKY, ~0.4% of supply) actually destroyed. Which is why the honest framing is yield rather than scarcity: the buyback is a distribution pipe, not a supply story. Against that, dilution has exactly two real sources — minting SKY to cover losses beyond reserves, and leakage to the Agents’ own tokens, SPK and GROVE, which capture upside outside SKY.
So what does a dollar of surplus cost? On a ~$1.52B market cap: 3.5× gross revenue, ~12× net surplus, ~24× the cash that actually reaches holders, and 17–19× book equity. Read backwards, the market is pricing roughly $65–75M a year to holders — a $130–150M surplus run-rate, which sits below the SFF’s own plan of $157.8M. There is no growth premium in that number and no visible risk discount either. Standard Chartered initiated coverage on 11 September calling Sky “DeFi’s federal bank” with a $0.325 end-2028 target, which needs five to seven times today’s surplus. On 18–25× holder cash, the bear case ($20–35M, if rates fall, savers leave and a credit loss forces SKY minting) prices $0.015–0.037; the base case ($75–110M, plan delivered and reserves full in 2027) $0.058–0.117; the bull ($190–300M, USDS above $25B and payout raised to 55–60%) $0.146–0.320. The realistic close is this: Sky survived a decade, built a ~$10B dollar with institutional partners, rebuilt its capital under pressure, and now pays half its surplus to token holders. If reserves fill and USDS grows without paying for it through the savings rate, SKY becomes something rare in crypto — a governance token that behaves like the equity of a well-run bank. Until then, size it as what it is: first-loss capital. The two numbers to watch each month are idle USDS as a share of supply, and reserves against the $150M floor.
Key findings
- Only ~15% of revenue reaches the token. The bridge is the whole story: ~$430M of gross revenue, minus the savings rate and the Agents’ revenue shares (≈−$270M), gives ~$160M of net protocol revenue; minus 20% security and maintenance plus opex (≈−$35M), ~$125M of net surplus. Stage 2 pays half of that last line — ~$62M — to stakers. Most of the $430M was already spent before holders are considered.
- A thin premium over the savings rate. Staked SKY yields ~5.3% (~4.1% measured on the full market cap) against ~3.5% on sUSDS, which carries no equity risk at all. Two points is what the market currently pays you for standing in front of the losses. At ~12× surplus and ~24× holder cash, today’s price is paying for the 2026 plan — not for growth.
- The “burn” is mostly redistribution, not scarcity. Since February 2025 roughly $125M of buybacks bought ~2.0B SKY — and only 2.86M were actually burned. Under Stage 2 the burn runs ~$6M a year, about 0.4% of supply. Supply is 23.46B SKY, ~74% of it staked, and the 2.03B sitting in the MKR converter for the 84,771 unconverted MKR are already inside that number, so they add no dilution. The real dilution risk comes from somewhere else entirely.
- SKY is first-loss capital, and the book it stands in front of is opaque in part. Reserves of ~$82–92M back an ~$11B book — under 1% equity, and S&P rates Sky B-. Losses beyond reserves are paid by minting SKY, which is exactly what happened on Black Thursday in March 2020. Sky also traded crypto-price risk for credit risk: Agents hold $6.84B of $12.32B of collateral, and while $2.58B of that sits with named institutions, what the other ~$4.3B holds is not clearly disclosed. For a first-loss holder that is the single most important unknown.
- The payout is a vote, not a right. It has had three regimes in eighteen months: 75% of surplus to programmatic buybacks from February 2025, cut ~90% to 7.5% on 14 March 2026 to rebuild a reserve stuck near $50M, then 50% under Stage 2 from 13 August. The Atlas lets the Core Council modify the Treasury Allocation — the rule that defines holder cash — and governance is thin: the founder holds ~9% (S&P, Aug 2025), one listed vehicle ~10%, and in a November 2024 brand vote four holders cast ~98% of the winning option. Holders have no legal counterparty.