STRK Inflation Analysis · September 2026 · Supply growing, projected to keep growing
Starknet adds more STRK to the market than almost any large layer two, and it does it on a published calendar. The MrNasdog Pressure Framework books 718.7M STRK of sell pressure and zero buy pressure over the trailing 90 days — a net of +9.78% against a circulating base of 7,350.6M STRK — and projects +9.11% for the next 90 days. The mechanism is not the mint. Starknet minted only 37.7M STRK for stakers over the window, while a vesting cliff released 381.0M STRK and a project distribution wallet sent out another 300.0M STRK. There is no fee burn and no buyback to take any of it back.
The verdict, in one paragraph
The framework reads STRK at +9.78% over the trailing 90 days and +9.11% over the next 90. The inflation monitor reads +12.70% for the same window, a gap of 2.92percentage points, which is far outside the framework's half-point tolerance, so a data-conflict flag ships on the overview page. The deep walk explains most of the direction and none of the last slice: the token contract minted 37.7M STRK in twenty-nine identical steps, the published lock-up released 127M STRK on three dated cliffs, and one project wallet fell from 1,250.0M to 950.0M in two traced transfers. Together that is 718.7M of the 828.5Mthe monitor's float added. The remaining 109.8M sits in Starknet-side project wallets that no public holder index would return this session, so the framework keeps the number it can prove rather than the one that would close the arithmetic. The label for Starknet is structurally inflationary on a published calendar, with an empty buy side.
Sell pressure: where new STRK comes from
Protocol inflation is the smallest of the three live rows and the only one that creates STRK that did not exist before. Starknet pays validators from a minting curve that scales with the square root of the share of supply staked, capped by a mainnet coefficient of 4%. Read off the chain rather than off the curve, that produced 37.7M STRK in the window: twenty-nine discrete mints of exactly 1.3M STRK, one roughly every three days, each one forwarded across the bridge the same block it was created. Annualised that is about 1.5% of supply, which implies roughly 14% of all STRK is staked — a small, orderly number that would barely register on its own.
Vesting unlocks are the row that defines the page. Early contributors hold 20.04% of supply and investors another 18.17%, all of it under a transfer restriction that releases 127M STRK on the fifteenth of every month from April 2025 through Mar 15 2027. Three of those cliffs fell inside the window — Jul 15, Aug 15 and Sep 15 2026 — for 381.0M STRK, and three more fall in the next one. Six remain after that, 762M STRK in total, and none of it depends on a vote or a decision. The framework checked whether the calendar overstates what actually moves: the 168 readable lock contracts Starknet deployed fell 479.4M to 418.7M over the window, but that escrow holds well under a fifth of the entitlement the cliff pays, so the published calendar governs and the claims inside the escrow are a subset of it rather than an addition.
The third row is discretionary and it fired twice. A project distribution wallet went from 1,250.0M STRK to 950.0M STRK, sending 100M on Jul 28 2026 and 200M on Sep 10 2026. Both passed through the same forwarding address, which finished each window holding two STRK. The first tranche landed in a custody wallet; on Sep 16 2026 the second was split into 90M sent straight into the Starknet bridge and 110Minto that same custody wallet. Under Starknet's own rule — Foundation allotments are not circulating until they leave the originating wallet — all 300M crossed into the float. The fourth sell row is empty: there is no bankruptcy estate, no trustee and no court-ordered STRK distribution.
Buy pressure: where new STRK goes
Nowhere. All four buy rows read 0, and each of them for a structural reason rather than a quiet quarter. There is no programmatic buyback: no contract, no announced programme, no disclosed open-market purchase by StarkWare or the Starknet Foundation. There is no protocol fee burn either, and this is the one most readers get wrong — Starknet fees can only be paid in STRK since the September 2025 upgrade, which sounds like a burn and is not. The project's own fee documentation says plainly that Starknet does not implement burning and the sequencer receives every fee charged. Both burn surfaces were read at both ends of the window to confirm it: the unspendable addresses hold a fraction of a single STRK and did not move, and the number of STRK in existence rose rather than fell. Foundation buying is zero in the other direction — the Foundation was a net seller of 300M STRK this window. And nothing new was locked: leaving a Starknet validator costs a seven-day wait, which is an exit delay rather than a term lock, and staked STRK is counted inside the tradable float regardless.
Foundation and overhang
Four overhangs are tracked. The project distribution wallet still holds 950.0M STRK after its two releases, and we read its balance on the chain; its fourteen-month record shows eight firings of between 69.6M and 250M, and a refill of 1,500M STRK in April 2026, which is why the framework projects 250M from it over the next 90 days rather than zero. A separate project safe holds 1,001.3M STRK and did not move a single unit across the window, and we read its balance too. The 168 lock contracts hold a remaining 418.7M STRK, vesting on the same contributor calendar already booked above. And roughly 445.8M STRK of the non-circulating pool could not be attributed to any address this session, because no public Starknet holder index answered — that leg is declared opaque rather than guessed at, and it is re-derived by subtraction at every rebuild. The rule for all four is the same: if one of these balances falls between refreshes, the outflow enters the Foundation row at the next refresh.
How STRK compares to other layer-two tokens
Against the rollup tokens it is usually grouped with, STRK is not unusual in kind, only in size and timing. A layer two does not have a block subsidy to pay, so its token supply is governed almost entirely by a launch allocation unwinding on a calendar. Where Starknet differs is that it also runs a live staking mint — most rollup tokens have none at all — and that its calendar is still in its heaviest phase. 127M STRK a month is 1.27% of the genesis supply arriving every month, on the same day, with no step down until it stops entirely in March 2027.
The sharper comparison is to chains that burn. Ethereum destroys base fees, so a busy day reduces supply; exchange tokens run quarterly buybacks and burns that scale with revenue and can push their readings negative. Starknet has the usage half of that design — every transaction pays in STRK — and none of the removal half, because the fee lands with the sequencer instead of an unspendable address. That single design choice is worth more to this page than the mint, the cliff or the treasury: a fee burn is the only mechanism that would let activity offset the calendar, and Starknet does not have one. Compared with a capped proof-of-work chain, the contrast is starker still — STRK has no protocol cap at all. The 10,000,000,000 figure still published as a maximum is the genesis mint, and the chain read 10,166.4M STRK at the close of this window.
What to watch in the next 90 days
First, Oct 15 2026, Nov 15 2026 and Dec 15 2026: three more 127M STRK cliffs, identical in size and composition, and the single most predictable supply event on this page. Second, the project distribution wallet at 950.0M STRK — its last two firings came seven weeks apart, so another release inside the window is the base case rather than a surprise, and its balance is the number to check. Third, the staking rate: the mint held at exactly 1.3M STRK every three days all window, so any change in that rhythm means the staked share has moved and the forward 39.0M needs re-basing. Fourth, the untraced 445.8M STRK of non-circulating supply — if a Starknet-side holder index becomes readable, that leg stops being opaque and the monitor gap should close with it. Fifth, Mar 15 2027, which is outside this window but is the date the whole page turns on: the last cliff pays, and the largest recurring sell row on STRK simply ends.
Summary
The MrNasdog Pressure Framework reads STRK at +9.78% over the trailing 90 days and +9.11% projected forward: supply growing, projected to keep growing. The mechanism is a published unlock calendar rather than inflation — Starknet minted 37.7M STRK for stakers while the contributor and investor lock-up released 381.0M STRK and a project wallet released 300.0M STRK more, against a buy side of exactly zero. The key risk is that none of it is discretionary and none of it is offset: there is no fee burn, no buyback, no protocol cap, and the cliff runs on the calendar whatever the price does. The genuine relief is dated — Mar 15 2027 ends the 127Mmonthly release for good, which removes more than half of this page's sell pressure on a schedule nobody has to vote for.
MrNasdog Pressure Framework analysis of STRK, Metric 1 — Inflation. Data + explanation only. Not financial advice. Updated Sep 20 2026.