Originally published at mrnasdog.com/research/tao/inflation by MrNasdog.
Bittensor has no premine, no investor allocation and no vesting calendar of any kind — and it is still one of the more inflationary large-cap networks the Pressure Framework tracks, at +2.81% over the trailing 90 days and +2.77% projected forward. All of it comes from one mechanism: the block reward. Bittensor minted 323.2K TAO across the 646,405 blocks it actually produced, at the 0.5 TAO reward left by the December 2025 halving, and only 4.8K TAO came back out of issued supply. The ceiling is a hard 21,000,000 TAO, and only 53.9% of it has been mined.
The verdict, in one paragraph
Against a circulating base of 11,325,285 TAO — the chain's own issued supply, which is what Bittensor and its explorer both mean by circulating, since free and staked TAO are the only two places a TAO can be — the framework books 323.2K TAO of sell pressure and 4.8K TAO of buy pressure over the trailing 90 days, a net of +2.81%, and projects +2.77% for the next 90 days on the same unbroken reward. The inflation monitor reads +0.02% for the same window, a gap of 2.79 percentage points, which is far over the framework's 0.5pp tolerance and therefore ships with a monitor-gap warning on the overview page. That gap is not a disagreement about Bittensor's mechanism; it is a broken input. The monitor's public supply feed has been frozen near 9,597,500 TAO since Aug 2025, so it registers only 2,051 TAO of growth where the chain itself issued 318,425 TAO — and it divides by that same frozen figure. Decomposed, 3.2972pp of the gap is the frozen numerator and −0.5069pp is the base it divides by, and the two sum to the gap exactly. On the frozen public figure the reading would be +3.32% rather than +2.81%; the mechanism and the coin count are identical either way. The label for TAO is a fair-launched chain that is still mid-mint: nobody is dumping an allocation, because there is no allocation; the issuance itself is the pressure.
Sell pressure: where new TAO comes from
It comes from the block reward, and from nothing else. Bittensor mints TAO on every block and pays it into an open market for machine intelligence, the way Bitcoin mints and pays miners. Sell #1, protocol inflation, is 323.2K TAO. Two measurements produce that figure rather than one assumption. First the reward: Bittensor's halving is triggered by issued supply, not by a date, and the first one fired when issuance crossed 10,500,000 TAO on Dec 15 2025, cutting the reward from 1.0 to 0.5 TAO per block. Sampling thirty consecutive blocks at each end of the window puts the reward at exactly 0.5 TAO on both dates, so no halving fell inside it; the next one waits for issuance to reach 15,750,000 TAO, projected for December 2029. Bittensor's own emission-rate field is worth naming as a trap here: it still reads 1.0 TAO at both window ends, nine months after the halving, because it is a dead legacy field that dynamic TAO stopped writing. Reading it instead of measuring the chain would have doubled this row.
Second the block count. Bittensor targets a twelve-second block; over these 90 days it produced 646,405 blocks at a realised 12.03 s, so it ran 0.247% slow. Because the reward is paid per block and no protocol constant absorbs the drift, that measurement corrects the mint downward by 1,595 blocks, or 797.5 TAO — the nominal 648,000-block assumption would have over-stated issuance. On-chain issued supply moved from 11,006,860 TAO to 11,325,285 TAO, and a second, separate on-chain counter agrees with the first to one billionth of a TAO.
Everything else on the sell side is a proved zero rather than an assumed one. Sell #2, vesting unlocks, is 0: Bittensor launched in January 2021 with no premine, no sale and no team or investor allocation, and the chain carries no vesting records, no locked balances and no reserved balances at either end of the window. Sell #3, Foundation and unscheduled unlocks, is 0, because there is no foundation reserve and no treasury on this chain to release from. Sell #4, long-term locked or bankruptcy, is 0 as well — TAO has no estate, no trustee and no court-ordered distribution attached to it.
Buy pressure: where new TAO goes
Almost nowhere. Buy #2, protocol fee burn, is 4.8K TAO — one coin returned for roughly every sixty-eight minted. Two flows drive it, and both are recycling rather than destruction. Registration costs, which is what it costs to put a miner, a validator or an entire new subnet onto Bittensor, are taken out of issued supply. And a network upgrade that went live on Aug 12 2026 added ordinary transaction fees to the same path, which tripled the pace: 0.0044 TAO per block before that date against 0.0139 TAO per block after it. Because that is a dated change to a live mechanism, the framework re-bases only the forward buy leg onto the post-upgrade rate, giving 9.0K TAO for the next 90 days while the trailing column stays as measured. The word matters: recycled TAO returns to unissued supply and can be minted again, so the effect is to push the next halving further out rather than to shrink the total.
Buy #1, programmatic buyback, is 0, and it could not be otherwise without a governance vote creating the pot first — Bittensor has no treasury and no foundation reserve to spend, because fees are recycled into unissued supply instead of being paid to an entity. Buy #3, Foundation buy, is 0: the single account carrying the Foundation's registered on-chain name went from 348 TAO to 580 TAO across the window, which is operating change on a rounding-error balance, not accumulation. Buy #4, new long-term lock, is 0 — and this is where dynamic TAO deserves a careful look, because it is the mechanism most likely to be misread in either direction.
Under dynamic TAO, none of the minted TAO is handed to a holder. All of it is injected into subnet trading reserves as protocol-owned liquidity, and participants are paid in that subnet's own alpha token instead. That could easily mean the headline mint overstates real sell pressure by several times. On Bittensor it does not. Enumerating all 129 subnet reserves at both window ends, the reserves kept only 11,080 TAO of the 318,425 TAO net issued — 3.5% — while 307,345 TAO, or 96.5%, drained back out into ordinary wallets as holders swapped alpha for TAO. The timing is sharp: before Aug 3 2026 the reserves absorbed 83.1% of new issuance and free balances grew by only 602 TAO a day; after it, reserves fell outright and free balances grew 7,227 TAO a day. That is a composition change, not a supply change — the chain's own definition of circulating supply already counts reserve and staked TAO as tradable float — so it is not booked in any row. But it is the single most important thing about TAO this quarter, because it says the mint is arriving in sellable hands almost in full.
Foundation and overhang
TAO's overhang is unusual: there is essentially no team-controlled one. Every account on Bittensor carrying a registered on-chain identity was enumerated at both window ends — 975 of them — and the only one bearing the Foundation's own name holds 580 TAO, refreshed from the chain at every rebuild. The largest registered institutional holder is a custodial trust, which belongs to its shareholders rather than to the project and is excluded on that basis. What the enumeration could still miss is a project wallet that never registered an identity, since registration is optional. That residual closes a different way, by arithmetic rather than by a wallet map: every issued TAO is either free in wallets (4,005,256 TAO) or inside subnet reserves (7,320,030 TAO), the two sum to issued supply of 11,325,285 TAO with no third bucket, and the remaining 9,674,715 TAO under the cap is un-minted future emission rather than anybody's balance. There is no allocation for an unregistered wallet to be holding.
The two pots genuinely worth watching are protocol-owned, not entity-owned: the 7,320,030 TAO in subnet reserves, and the 30,403 TAO held against subnet registrations. Both are read from the chain at every rebuild. The trigger sentence applies to both: if either balance falls between refreshes by more than the mint accounts for, that outflow enters Sell #3 at the next refresh — and as the Aug 3 2026 drain shows, the reserve pot can move fast.
How TAO compares to other capped, halving-model chains
TAO is deliberately Bitcoin-shaped: a hard 21,000,000 ceiling, a fair launch with no premine, and a reward that halves. The difference that matters for this reading is how far through the schedule each chain is. Bitcoin has mined roughly 95% of its cap and sits four halvings deep, so its issuance reads a fraction of a percent a quarter. Bittensor has mined 53.9% and has had exactly one halving, so the same shape of schedule produces +2.81%. A hard cap tells you where issuance ends, not how fast it is running now, and a young capped chain is closer in supply behaviour to an uncapped continuous-emission network than its own marketing suggests.
Against uncapped Cosmos-style L1s, where a staking-linked emission of several percent a year is normal and the ceiling is a policy rather than a number in the code, TAO is stricter on the promise and roughly comparable on the current rate. Against a capped chain that has switched issuance fully off, TAO is far weaker — those chains read a flat zero on this row, while TAO reads its largest number there. And against exchange tokens that run quarterly buybacks and burns, the contrast is structural rather than one of degree: those chains can print a negative inflation reading because a demand-linked removal scales with usage. Bittensor has no buyback and no permanent burn at all. Its removal path recycles into unissued supply, so even a very busy quarter cannot make the reading negative — the best it can do is delay the next halving. For recycling to matter to this page, Bittensor would need to return TAO on the order of 320K a quarter, about 2.8% of circulating supply, against the 4.8K it actually returned.
What to watch in the next 90 days
First, the reward itself, which is the only thing driving the number: it stays at 0.5 TAO per block until issued supply reaches 15,750,000 TAO, projected for December 2029, so absent a governance change the next-90-day reading lands near +2.77% again. Second, the recycling rate, which stepped up threefold at the Aug 12 2026 upgrade and is read from the chain at every rebuild — it is lumpy, driven by subnet and neuron registrations, so a busy registration quarter is the one thing that could move the buy row meaningfully. Third, the realised block interval: Bittensor ran 0.247% slow this window, and because the reward is paid per block, a chain that speeds up mints more. Fourth, the subnet reserve balance at 7,320,030 TAO, which drained at 7,227 TAO a day after Aug 3 2026 — that drain is what converts the mint into sellable coin, and its pace is the difference between a reading on paper and pressure in the order book. Fifth, Bittensor governance, where the Triumvirate proposes and the Senate of top delegates votes: the upgrades that landed inside this window redistributed emission across subnets and changed how staker yield is routed without touching issuance, but the same vote type could change the reward or the halving rule.
Summary
The MrNasdog Pressure Framework reads TAO at +2.81% over the trailing 90 days and +2.77% projected forward: supply growing, projected to keep growing. The mechanism is issuance, plain and unhidden — 323.2K TAO minted at 0.5 TAO a block across 646,405 measured blocks, against 4.8K TAO recycled back to unissued supply, on a chain with no premine, no vesting, no treasury and no buyback anywhere in its design. The key risk is that this is the honest kind of inflation and therefore the kind nobody can switch off: it runs every twelve seconds regardless of price, and the only removal path returns coins to the mint rather than destroying them. The ceiling is the one genuine comfort — 21,000,000 TAO is hard and on-chain — but at 53.9% mined, that ceiling is still a long way from binding.
MrNasdog Pressure Framework analysis of TAO, Metric 1 — Inflation. Data + explanation only. Not financial advice. Updated Sep 10 2026.
Top comments (0)