The Emission Gate

What shipped

Spec version 440 changes one rule, and it is the one that decides how much emission each subnet earns. Until this release every registered subnet earned emission in direct proportion to its moving price. The release's own account of the problem with that rule is that it is simple but too flat — a subnet nobody uses still collects a slice of every block, so the tail of the network dilutes the subnets doing real work, while the passive income of a parked slot props up the cost of registering a new one. [source]

Emission remains price-based, but prices now pass through a threshold that the release centres around the 32nd rank. Above the threshold, emission is essentially unchanged. Below it, emission collapses toward zero. Technically, the release is careful to say, no subnet is hard-cut — every price still earns something — but a parked slot pays substantially less than an active one. That threshold is the emission gate this release is named for. [source]

The source for this page is one document: the project's own note for the V440 upgrade, dated July 2026. It is both the best available description of the change and an interested party's account of it, so what follows reports what the note says rather than repeating its verdict. [source] [source]

The release argues the change is really about the cost of entry. Under flat price-proportional emission, even the worst slot on the network carried a passive yield, so slots traded at inflated values — around 1300 TAO, by the note's own figure — and teams who wanted to build had to borrow against that inert cost, most of them taking on debt to do it; the release's summary of the outcome is that the people the network most wants were the ones the pricing punished. Lowering the base return of an idle slot to approximately nothing is meant to make an idle slot worth approximately nothing, so that the price of entry falls toward the cost of the registration transaction. A team no longer buys a guaranteed slice of emission, the note says — just a starting position below a strict competition bar. [source] [source] [source]

What it changes for the ecosystem

Subnet owners are now sorted into two groups by a single number. To owners below the bar the release says emission does not go to zero, but it no longer pays to idle; the bar is public, and owners are told to query it with btcli subnets bar and track their own share against it. To owners above the bar it says there is nothing to change — their emission rises as the tail's subsidy is redistributed. The release also puts a number on the payoff for climbing: a subnet at rank 36 that grows its demand by ten percent gains roughly twenty-six percent more emission, which it presents as making growth near the cliff the best-paid move on the network. [source] [source] [source]

The bar itself is not a fixed price, and that is the part worth understanding before anything else. The release derives it from the shape of demand: sort subnets by price and walk down the list until the subnets above you carry a set fraction of all prices; the price at that crossing is the bar. Two consequences follow, and the note states both. If the tail grows real demand the bar rises and the standard gets harder for everyone; if demand concentrates, the bar falls. And because the bar is a property of the distribution rather than a count of slots, registering new subnets does not move it. At the shipped defaults the crossing lands on today's distribution at the average demand share of an active subnet, around rank 32 — so a subnet earns its full linear emission only if it attracts more than an average slice of demand. [source] [source] [source]

How much this moves, the release shows with its own figures, computed from a live mainnet snapshot of all 126 subnets with every subnet treated as emission-enabled. Below-bar emission — the 94 subnets beneath the threshold — falls from 38.4% of the total to 12.5%, while the top eight rise from 32.8% to 52.7%. The count of subnets hard-zeroed is zero on both sides of that change, which is the release's way of insisting the tail is throttled rather than switched off. [source] [source] [source] [source] [source] [source] [source]

The two figures that say the most about what kind of network this produces are the concentration measures. The rank at which cumulative emission reaches eighty percent moves from 64 to 23, and the release's effective-subnet count — one divided by the sum of every subnet's squared share, a standard way of asking how many subnets the payout behaves as though it has — falls from about 50 to about 22. On the release's own snapshot the network pays out as if it had under half as many subnets as before. [source] [source] [source] [source] [source]

For stakers the instruction is short and unhedged: alpha positions in deep-tail subnets earn materially less emission after this upgrade, the head earns more, and the release tells holders to reprice accordingly. [source]

Unverified The release's closing instructions are addressed to subnet owners, stakers, teams waiting to build, node operators and signers. Miners and validators are not mentioned anywhere on the page — and they are the people the 94 below-bar subnets exist to pay. On the release's own retention figures, a mid-tail subnet keeps 15–30% of what the old rule would have sent it and a deep-tail subnet keeps 1–5% [source] — a fall of roughly seventy to eighty-five percent in the mid-tail and ninety-five percent or more in the deep tail, for the subnets that make up most of the network by count. On this document's own arithmetic that is a larger change to those participants' income than anything the release does discuss, and the release does not discuss it.

Unverified How that reaches a miner's or a validator's balance needs care, because the units are not the same. Lesson 09 describes the pool destined for participants as capped at up to one alpha per block at the subnet's current halving rate, while the TAO side of the split governs what is injected into that subnet's own trading pool. If that still holds after this release, the first-order effect on a below-bar miner is not that less alpha is minted for it, but that far less TAO stands behind the alpha it is paid in. That is the same statement the release makes to stakers when it tells them deep-tail alpha positions earn materially less and to reprice accordingly [source] — and miners and validators are paid in that same alpha. Nothing here has been checked against chain state, and the note does not trace the path itself.

Teams waiting to build are the constituency the release is aimed at. Existing slots lose the passive carry that made them expensive, the note expects registration cost to fall toward the transaction fee, and it says plainly that a slot buys a starting position and not an income stream. It also names the follow-on it is not doing yet: because a slot below the bar earns little, slots no longer need to be scarce, which the release presents as clearing the path to raise the subnet cap later — with raising the cap explicitly deferred to a later release. [source] [source]

Node operators and signers have the least to do: wait for the on-chain spec_version to move to 440 and update to the matching release, then sign the proposed upgrade through the release train's own command. [source] [source]

Unverified That instruction lands on lesson 15 without overturning its argument. That lesson's finding — that a subnet's slice of a day's emission is set by how much people want to hold its token, not by any verdict on the work — survives this release, because what the split reads about a subnet is still a market price and still nothing about its work. What changes is the exchange rate: the same difference in price now buys a far larger difference in income, so the concentration that lesson measured should be read as a floor rather than a ceiling.

Unverified That lesson's printed formula, though, no longer holds as written, and this is the page that should say so. Lesson 15's technical section states that a subnet's cross-subnet share runs through one formula with exactly two inputs, its own price and its own burn penalty. After this release the resulting share is multiplied by the gate, which brings in two further quantities: the bar, derived from every subnet's price at once, and the exponent that sets how sharp the cliff is — both governance-set rather than properties of the subnet at all. What becomes of the burn penalty is not something this note answers; it never mentions it. Anyone revising that lesson should start at that sentence.

Unverified The gate also sits ahead of everything lesson 09 describes rather than inside it. That lesson picks up once a subnet's overall share has been decided and follows the money into balances — the alpha minted for participants, the slice reserved for root network stakers, the division between whoever does the work and whoever judges it — and none of that machinery is touched here. The same is true of the owner's cut that lesson 04 traced: the percentage is unchanged, and only the quantity it is a percentage of moves. Neither reading has been checked against chain state.

What to watch

The two numbers that set the gate are not fixed by the code. EmissionBarQuantile and EmissionGateExponent are both root-sudo settable and rate-limited, so every figure quoted on this page is a default rather than a constant. The bar itself is recomputed once per tempo from the same de-manipulated moving prices that already drive emission, which the release says is what stops a single block of wash trading from moving it, and what keeps emission near the bar steady rather than flapping under a sharp cliff. [source]

Unverified The open question is what a steeper price-to-income curve rewards. Making a subnet's income a sharply convex function of its market price means anything that moves that price near the bar now moves income far more than before, and the release does not discuss what constrains that beyond the per-tempo recompute and the de-manipulated price series it already relies on. Whether a bar derived from the demand distribution is harder to game than a flat share of it is not answerable from this document.

Unverified The cost-of-entry claim is a prediction, not a measurement. The release states the expected direction — passive carry disappears, so slot prices should fall toward the registration fee — and nothing on this page tests it. The test would be a period of registrations priced after the change, and the subnet cap increase that would put real pressure on that price has been deferred, so the prediction stays untested for now.

// pallets/subtensor/src/coinbase/subnet_emissions.rs
let theta = Self::q_mass_bar(&shares, EmissionBarQuantile::<T>::get()); // q
let h = EmissionGateExponent::<T>::get(); // h

// gate(s) = s^h / (s^h + theta^h); emission weight = s * gate(s)
let weights: Vec<U64F64> = shares
 .iter()
 .map(|s| {
 let sh = s.saturating_pow(h);
 s.saturating_mul(sh.safe_div(sh.saturating_add(theta.saturating_pow(h))))
 })
 .collect();