Stake-Directed Emissions
Stake-directed emissions are newly issued tokens whose destination is set continuously by where staked capital sits, rather than by a vesting calendar or a governance committee. The total issued per block stays fixed, but the split between recipients is recomputed in real time as capital moves.
TradFi parallel: Like a flow-weighted index that reweights on every tick rather than on a rebalancing date. The size of the fund is fixed by its mandate; who inside it gets the money is decided by where the money is already going.
Key Takeaways
- 01Stake-directed emissions set the destination of new supply by where staked capital sits, recomputed continuously, rather than by a schedule or a committee
- 02Under Bittensor's dTAO model, introduced in February 2025, each subnet runs an AMM between TAO and its Alpha token: staking is a swap that moves the Alpha price, and emission share follows a moving average of that price rather than the size of the stake
- 03Allocation is per block, roughly every twelve seconds, but the share is filtered: the price average has to move, a miner-burn penalty discounts it, and the v440 emission gate collapses the share of subnets ranked below roughly 32nd
- 04The total is capped rather than calendared. TAO is hard capped at 21 million and the December 2025 halving cut daily issuance from 7,200 to 3,600, and because halvings fire at total-issuance thresholds while recycling removes TAO from issuance, the next date is not predictable
- 05The claimant count is a governance variable: the Robin tau upgrade on 3 May 2026 doubled active subnet capacity from 128 to 256, adding 128 new Alpha economies competing for the same emissions
- 06Capital position is not usefulness. The SN28 case saw a subnet raise its emission share by concentrating stake and aligning validators rather than by producing better AI output, and Taoflow, the November 2025 attempt to replace price with net stake flow, was reverted to a price-based split in 2026
How It Works
Most tokenomics tooling rests on one assumption: emissions follow a schedule. Vesting cliffs have dates, linear schedules have rates, and the calendar tells you when supply arrives. Bittensor's dTAO model, introduced in February 2025, breaks that assumption. Each subnet runs its own automated market maker between TAO and its native Alpha token. Staking into a subnet is a swap rather than a deposit: TAO goes in, Alpha comes out, and the trade moves that subnet's Alpha price. Emission share then follows a moving average of the Alpha price, not the size of the stake and not the spot price, so a single inflow has to persist before it counts. There are no vesting calendars for subnet tokens at all: emissions are allocated continuously, in real time, by where capital is flowing.
The mechanics are continuous rather than periodic, though they are not frictionless. Every block, roughly every twelve seconds, the protocol releases new TAO and new Alpha and distributes them to participants. Buying into a subnet lifts its Alpha price and, once the moving average catches up, its share of block emissions, so miners and validators inside it earn more, while selling out works the other way. Two filters sit on top of that share: a miner-burn penalty discounts it, and the v440 emission gate collapses the share of subnets ranked below roughly 32nd. Tokenomist's framing of the design is that every staker is casting a live, capital-weighted vote on which subnets deserve resources, and that no committee decides which subnets get funded, your stake does. Root, netuid 0, is not a neutral abstention from that contest. Root TAO carries a governance-set tao_weight of 0.18 against Alpha stake, root dividends are recycled in full unless the summed moving-average prices of eligible subnets exceed 1.0, and the Root Reborn upgrade in v441, July 2026, turned root from a passive dividend system into a competitive allocation layer where dividends are redeployed across validator-curated baskets of subnets.
What is market-set is the destination, not the total. TAO has a 21 million hard cap, and the Opentensor Foundation states there was no pre-mine and no VC allocation, a claim sourced to the Foundation alone and contested by researchers who point to roughly 5.38 million TAO mined before subnets launched publicly. The first halving, in December 2025, cut daily issuance from 7,200 to 3,600 TAO, shrinking the pool that subnets compete for. Halvings fire when total issuance crosses fixed thresholds, 10.5 million, then 15.75 million and so on, rather than on a calendar or at a set block height, and because recycled TAO is removed from issuance the date of the next one cannot be fixed in advance. The number of claimants is a separate governance variable: on 3 May 2026 the Opentensor Foundation executed the "Robin tau" upgrade, doubling active subnet capacity from 128 to 256 and opening 128 new Alpha token economies at once, each with its own liquidity pool competing for emission weight. Forecasting under this design means tracking which subnets are capturing emission share and which are losing it, since the schedule only tells you the size of the prize.
The failure mode is that capital position is not the same as usefulness. Early versions of Bittensor showed that emissions can be influenced by capital and validator behaviour rather than by real usage: in the SN28 case, a subnet attracted more emissions by concentrating stake and aligning validators, raising its share of rewards not because of strong AI output but because of how capital was positioned inside the network. Bittensor tried replacing the price signal outright. Taoflow, live from November 2025, set emission share from a 30-day moving average of net stake flow and clipped negative-flow subnets to zero. The experiment was reverted in 2026: release v431 in July 2026 states that emission share is determined solely by each subnet's moving-average price, v440 confirms that emission remains price-based, and the current emissions documentation does not mention flow at all. Vote-escrow designs reach a similar outcome by a different route: Katana Network's vKAT model, launched in March 2026, uses vote-escrow to direct emissions toward high-revenue DeFi pools, adding an explicit lock and vote where dTAO reads capital position directly.
Real World Examples
Bittensor dTAO: the Alpha price sets emission weight
View →Since February 2025 each Bittensor subnet has had its own Alpha token and its own AMM against TAO. Staking is a swap: TAO goes in, Alpha comes out, buying pressure drives the Alpha price up, and the protocol reads a moving average of that price as the subnet's claim on block emissions, discounted by a miner-burn penalty and filtered by the v440 emission gate below roughly the 32nd-ranked subnet. Tokenomist's summary is blunt: there are no vesting calendars for subnet tokens, emissions are allocated continuously, in real time, by where capital is flowing, and the market decides supply rather than a spreadsheet.
Root (netuid 0): not the neutral option it looks like
View →Staking TAO to root spreads exposure across subnets without selecting individual ones, which reads as the passive choice in a system where standing aside forfeits emission share. It is not passive. Root TAO carries a governance-set tao_weight of 0.18 against Alpha stake, and root dividends are recycled in full unless the summed moving-average prices of eligible subnets exceed 1.0. Root Reborn, shipped in v441 in July 2026, went further and turned root into a competitive allocation layer where dividends are redeployed across validator-curated baskets of subnets.
SN28: emission share bought with stake concentration
View →Tokenomist's account of the gaps in the design points at SN28, a subnet that attracted more emissions by concentrating stake and aligning validators, increasing its share of rewards. The driver was not strong AI output but how capital was positioned inside the network. Bittensor's response was Taoflow, live from November 2025, which replaced the price signal with a 30-day moving average of net stake flow and clipped negative-flow subnets to zero. It did not last: v431 and v440 in 2026 restored an emission split determined solely by each subnet's moving-average price.
Robin tau and the December 2025 halving: more claimants, smaller pool
View →Two governance events changed the arithmetic in opposite directions. The first TAO halving, in December 2025, cut daily issuance from 7,200 to 3,600 tokens, reducing the total each subnet competes for. Secondary sources disagree on which day it landed and no primary record fixes it, so the month is as precise as the date gets. Then on 3 May 2026 the Opentensor Foundation's Robin tau upgrade doubled active subnet capacity from 128 to 256. More claimants against a smaller pool means average emission share per subnet falls even if no capital moves at all.
Katana Network (KAT): the vote-escrow route to the same outcome
Katana, a DeFi-native Layer 2 built on Polygon's CDK stack and incubated by Polygon Labs and GSR, held its TGE on 18 March 2026 with a vote-escrow (vKAT) incentive model designed to direct emissions toward high-revenue DeFi pools. The design reaches stake-directed allocation through an explicit lock and vote rather than by reading pool inflows, which makes the signal deliberate and periodic instead of continuous.
Frequently Asked Questions
If emissions have no calendar, what do you track instead?
Emission share. The total issued per block is still known and scheduled, so the forecastable quantity moves from dates to distribution: which recipients are gaining share and which are losing it. Tokenomist's own framing of the Bittensor case is that tracking this kind of emission system requires different tooling than a standard unlock calendar, and that monitoring which subnets capture emission share and which lose it is where the supply intelligence sits. A calendar view answers the wrong question here.
Does stake-directed allocation change how much is issued in total?
No. The two clocks are separate. Bittensor's issuance is capped at 21 million TAO and paced by halvings that fire when total issuance crosses fixed thresholds rather than on a calendar, and the December 2025 halving cut daily issuance from 7,200 to 3,600 tokens regardless of how capital was positioned. Because recycled TAO leaves issuance, the next halving date cannot be fixed in advance. Staking decides only who receives the block emissions, not how many there are. Confusing the two leads to the mistake of treating a subnet's rising emission share as network-level inflation when it is redistribution.
Can emission share be gamed?
Yes, and it has been. Tokenomist documents the SN28 case, where a subnet raised its share of rewards by concentrating stake and aligning validators rather than by producing better output. Bittensor's answer was Taoflow, live from November 2025, which replaced the price signal with a 30-day moving average of net stake flow and clipped negative-flow subnets to zero. It was reverted in 2026: v431 and v440 restored a split determined solely by each subnet's moving-average price. Emission share is a measure of capital positioning first and of usefulness only by inference.
How is this different from vote-escrow gauge voting?
In the signal and its timing. Vote-escrow systems require holders to lock tokens for a term and then cast explicit votes that set emission weights for a period, as Katana's vKAT model does in directing emissions toward high-revenue DeFi pools. Bittensor's dTAO skips the vote: swapping TAO for a subnet's Alpha is itself the signal, read continuously through a moving average of the resulting price, with no lock term and no voting round. The result is similar, a market-set emission split, but one is a periodic election and the other is a live price.
Related Terms
Track on Tokenomist
Supply-side analysis for educational purposes. Not financial advice. Verify assumption and precision labels on the relevant token page.
Tokenomist.ai provides a complete solution for supply-side tokenomics data. Analyze future token emissions, track vesting schedules, and compare standardized tokenomics and allocation across projects to gain actionable insights