Originally published at mrnasdog.com/research/chip/inflation
CHIP, the governance token of USD.AI, has a supply that cannot change: 10,000,000,000 CHIP were minted once on Arbitrum, the token has no mint function and no burn function, and both dead addresses held nothing at either end of the window. Every unit of CHIP pressure therefore comes from a wallet opening rather than from issuance. In the 90 days to Sep 7 2026 the 32 wallets USD.AI controls fell from 8,768,141,256 to 8,611,012,595 CHIP — a release of 157,128,661.04 CHIP — against 26,116,193.53 CHIP locked into the slashable staking module. That is +6.55% net on a counted float of 2,000,000,000 CHIP, against a supply-monitor reading of +0.59%: a gap of 5.96 percentage points the monitor cannot close, because the CHIP supply figure it divides has not moved since listing.
The verdict, in one paragraph
For the 90-day window ending Sep 7 2026, the Pressure Framework reads CHIP at +6.55% net: 157,128,661.04 CHIP reached the market from identified USD.AI programme wallets, against 26,116,193.53 CHIP removed into the staking module. The independent supply monitor reads the same window at +0.59%, so the gap is 5.96 percentage points and CHIP ships with a data-conflict flag. That flag is not a disagreement about arithmetic. The monitor divides a circulating-supply figure pinned at exactly 2,000,000,000 CHIP for all 137 readings since listing on Apr 22 2026 — a classifier that cannot move, so it cannot register a release even when one happens in plain sight on-chain. The forward column reads +2.21%, lower only because the framework projects dated releases rather than averaging a burst: three monthly firings and one tranche remainder are booked, and further treasury payments are not. The label for CHIP is fixed supply, discretionary float: nothing is created, and everything that matters is decided by whoever holds the treasury keys.
Sell pressure: where new CHIP comes from
Sell #1, protocol inflation, is zero, and it was proved rather than assumed. The CHIP token is an ERC-20 on Arbitrum behind an upgradeable proxy, and its total supply read exactly 10,000,000,000 CHIP at both ends of the window. A flat number means nothing on its own — a supply figure compiled into bytecode reads flat whether or not anything happened — so the write path was tested directly. The CHIP supply value lives in writable storage and the literal appears nowhere in either bytecode, which makes the flat reading a genuine measurement; enumerating the implementation's selectors then showed why it never moves, because there is no mint, no burn and no burn-from anywhere in the CHIP contract. One caveat is booked rather than buried: the proxy can still be pointed at new code by its administrator, so CHIP supply finality is a policy, not a protocol guarantee, and no row on this page carries a permanent tag.
Sell #2, vesting unlocks, is zero. USD.AI investor and core-contributor allocations account for 53.1% of CHIP supply between them and share one twelve-month cliff that opens on Mar 30 2027 — a third at once, the rest over 24 monthly steps. Nothing in that CHIP vesting schedule touches this window or the next one, and the thirteen dormant allocation wallets holding 4,666,425,947 CHIP between them did not move a single coin across the ninety days, which is what a real cliff looks like on-chain. The 2027 overhang also shrank before it ever became float: Permian Labs, the USD.AI developer, repurchased 338,806,273 CHIP out of investor allocations in May 2026 and 31,731,287 unclaimed airdrop CHIP before that.
Sell #3, foundation and unscheduled unlocks, carries the entire CHIP page at 157,128,661.04 CHIP, and it is measured as a group so that reallocations between USD.AI wallets cancel instead of double-counting. The treasury multisig did most of it: it held 9,001,809,916.80 CHIP the day after launch and 697,255,336.07 today, sat completely still for the first eight weeks of the window, then paid 99,999,990 CHIP out on Aug 4 2026 and another 12,353,250 across Sep 1 2026 and Sep 2 2026. Two monthly release wallets added 26,562,500 and 13,281,250 CHIP on fixed dates, and a distribution wallet a further 4,859,346. The reading survives two independent checks: the treasury's own transfer log closes on its balance change with nothing left over, and an exchange hot wallet sharing none of those addresses rose 156,692,547 CHIP over the identical ninety days — a separate surface agreeing to within 0.3%. Sell #4, long-term locked or bankruptcy supply, is zero: CHIP launched in April 2026 and has no estate, no trustee and no long-dated lock unwinding.
Buy pressure: where new CHIP goes
Three of the four CHIP buy rows are zero, and each zero was measured rather than inferred. Buy #1, programmatic buyback, is zero: USD.AI runs no open-market buyback. The May 2026 repurchases described above totalled 3.71% of CHIP supply, but they were private purchases of allocations that were still locked, so no CHIP was lifted off an order book, and both landed before this window opened. Buy #2, protocol fee burn, is zero on both surfaces the framework is required to read: neither dead address held a single CHIP at either end of the window, and total supply did not fall. Since the contract carries no burn function either, a burn would have to be written before it could happen. Buy #3, foundation buying, is zero — the USD.AI treasury only pays out, and its sole inflow all window was 297,769.77 CHIP arriving from a protocol contract as fees taken in kind, not CHIP bought on a market.
Buy #4, new long-term locks, is the one CHIP buy row with a number: the staking module grew from 23,985,431.22 to 50,101,624.75 CHIP, a rise of 26,116,193.53, with 22,604,602.54 of it in the final four weeks. It is booked as a genuine lock rather than a relabelled balance because USD.AI's own documentation states that unstaking is not immediate and that cooldown periods set by governance apply to every withdrawal from the module, and because staked CHIP is the first loss if the GPU loan book takes a shortfall. Stakers are paid nothing for carrying that risk — the documentation describes no rewards, no emissions and no points for sCHIP — which makes the growth a deliberate alignment choice rather than a yield chase, and makes it more likely to stay locked than a subsidised stake would be.
Foundation and overhang
The CHIP overhang is unusually large and unusually concentrated, so it is enumerated in full. The USD.AI project treasury holds 697,255,336.07 CHIP of unlocked reserve — roughly 35% of the counted float — with no published release schedule, and it is re-read on every refresh rather than taken on trust. The distribution wallet holds 41,922,286 CHIP and moves in both directions with no fixed dates. The airdrop claim contract holds 27,837,632.45 CHIP and has stopped draining. The two largest unlabelled holders carry 1,640,354,201 and 1,291,460,704 CHIP and are watched without being booked, because coins held by no identified group cannot be projected. Two lock contracts are tracked and excluded from the ledger: the omnichain lock adapter at 115,070,847.53 CHIP and the bridge lock at 11,249,885.61 hold CHIP represented on other chains, so those balances move CHIP between networks rather than into supply. Exchange custody wallets are excluded by rule. If any tracked balance falls between refreshes and the CHIP does not land inside another lock, the outflow enters Sell #3 at the next refresh.
How CHIP compares to other infrastructure-credit tokens
CHIP belongs to a class the framework sees often and scores badly: a fixed-supply governance token launched at a low float over a business that is genuinely working. The mechanical contrast with an emission-driven Layer 1 is total. An uncapped chain mints new coins every block and hopes a fee burn offsets them, so its supply curve is knowable in advance. CHIP mints nothing at all, which sounds better and measures worse, because the entire supply already exists and the only question is when its holders choose to move it. A published emission schedule is at least a promise; an unlocked treasury reserve with no schedule is a decision that can be taken on any morning, and on Aug 4 2026 it was.
Against tokens that route real revenue back into the token, the gap is starker still. USD.AI is a real credit business — it leads its lending category on deposits and collected $24.0M of fees over the trailing year against a market capitalisation near $116M — but none of that reaches CHIP, and USD.AI's own legal note is explicit that CHIP does not represent equity, ownership, or a claim on protocol assets. A token with a fee-funded burn converts business growth into supply reduction; a token with a fee-funded buyback converts it into a bid. CHIP converts it into neither, so a good quarter for the protocol changes the loan book and leaves the CHIP ledger where it was. The one structural defence CHIP has is the staking module — and at 26,116,193.53 CHIP it offset roughly a sixth of the release.
What to watch in the next 90 days
First, the two monthly release wallets, which are the only dated CHIP supply events on the calendar: one pays 8,854,166.62 CHIP around the 22nd and holds exactly four more firings, the other pays 4,427,083.38 around the 21st and also holds four, so three of each fall inside the window from Sep 21 2026 onward. Second, the 30,555,625 CHIP still parked from the Aug 4 2026 tranche, which has been passing downstream in roughly monthly steps and should clear inside the window. Third, the USD.AI treasury reserve at 697,255,336.07 CHIP: it is deliberately booked at zero forward because two firings of wildly different size are an onset rather than a schedule, so any third tranche is pure upside to the sell row. Fourth, the governance contract, which has emitted no event since deployment — the first executed CHIP vote could introduce a fee switch, a buyback or a burn. Fifth, and furthest out but largest, the Mar 30 2027 cliff, when a third of 53.1% of CHIP supply opens in a single day.
Summary
The MrNasdog Pressure Framework reads CHIP at +6.55% net over the trailing 90 days and +2.21% over the next 90, against a supply monitor reading +0.59% off a circulating figure that has not moved since listing. The structural mechanism is that CHIP creates nothing and destroys nothing: total supply is fixed at 10,000,000,000 CHIP, there is no mint and no burn in the contract, and the whole ledger is USD.AI programme wallets deciding to open — 157,128,661.04 CHIP in this window, against 26,116,193.53 locked into a staking module that pays nothing. The key risk is that most of the release is discretionary and unscheduled, so the forward number counts only what is dated and leaves 697,255,336.07 CHIP of unlocked reserve entirely unbooked. The ceiling is the cliff: 53.1% of CHIP supply is locked until Mar 30 2027, after which the treasury stops being the main source of new float.
MrNasdog Pressure Framework analysis of CHIP, Metric 1 — Inflation. Data + explanation only. Not financial advice. Updated Sep 7 2026.
Top comments (0)