TAO is the unit of account for every subnet market

Bittensor is not “an AI token” in the way most people mean it. It is a chain where markets decide which AI workloads get subsidized, block by block, via liquidity injection. Dynamic TAO (dTAO), as described in the dTAO mechanics, turned each subnet into its own micro-economy with its own token, and TAO sits underneath all of them as the common denominator.

If you’re new to the terminology, our tokenomics FAQ covers common definitions that show up in analyses like this.

TAO’s job in the product is concrete:

1) Staking collateral and influence. TAO can be staked with validators to earn dividends and influence consensus outcomes.

2) Base currency for subnet exposure. In dTAO, “staking into a subnet” functionally means swapping TAO into that subnet’s alpha token through the subnet pool. That makes TAO the routing asset across the entire subnet graph.

3) Paying chain costs. TAO is used for transaction fees on-chain. The nuance is what happens to those fees. They are recycled in a way that matters for float modeling.

CoinGecko’s page is still the cleanest single place to sanity-check supply surfaces (max supply, circulating supply, and how they define FDV).

Supply, issuance, and why “circulating supply” is a misleading headline

TAO’s supply story is simple on paper and messy in practice because Bittensor actively “recycles” tokens back out of circulation. For deeper dives into these dynamics across networks, see our research reports.

Max supply: TAO has a fixed maximum supply of 21,000,000.

Denomination: 1 TAO = 109 RAO (RAO is the smallest unit used internally).

Block cadence and issuance: New TAO is created each block (about every 12 seconds). The starting emission rate is 1 TAO per block (about 7,200 TAO/day).

Halving mechanics: The protocol describes a “Bitcoin-like” halving math explicitly, using a 10,500,000 issuance interval so total supply approaches 21,000,000, with the first halving reducing the block emission from 1 to 0.5. For a contrasting take on emission schedules, compare this with our Kaspa tokenomics review.

Current regime (important): The official emissions documentation (last edited January 25, 2026) states that the total block emission used in TAO injection calculations is currently 0.5 TAO per block.

Surface circulating supply: CoinGecko currently reports a circulating supply of 9,597,491 TAO and max supply of 21,000,000.

Here is the liquidity-structure gotcha. In Bittensor, “circulating” is not a clean proxy for “effective float.” Large amounts of TAO can be:

• Staked (still counted as circulating in many definitions, but not necessarily liquid).

• Sitting inside subnet pools as protocol-provided AMM reserves, which changes the shape of sell pressure.

• Recycled (deducted from TotalIssuance), which can leave “issued over time” and “currently circulating” telling different stories.

dTAO emissions: liquidity injection first, rewards second

dTAO’s core move was to stop treating subnets like budget lines in a governance spreadsheet and start treating them like tradable markets.

Subnet pools are constant product AMMs. Each subnet has its own pool for swaps between TAO and the subnet token (alpha).

No LP fee layer. Swaps in these pools do not charge fees because liquidity is provided by emissions rather than third-party LPs.

Two-stage emission system. The official docs frame emissions as:

1) Injection: every block, TAO and alpha are injected into subnet pool reserves based on the active distribution model.

2) Distribution: at the end of each tempo (about 360 blocks, about 72 minutes), rewards are distributed to participants via Yuma Consensus.

The allocation split for a subnet’s outgoing alpha emissions is explicitly documented.

The liquidity realist takeaway is that Bittensor is not merely “emitting rewards.” It is continuously building and reshaping market depth across many pools, then paying participants mainly in alpha, which must often be converted back into TAO to become broadly liquid. That conversion pressure is endogenous to the AMM design, not an afterthought.

Recycling, registration costs, and burn optics vs actual tradable float

Bittensor uses the word “burn” in user-facing interfaces in a way that can confuse supply analysis. The recycling glossary is clearer: a lot of what looks like burn is really recycling, and recycling directly impacts effective circulating supply and the timing expectations around halvings.

All transaction fees are recycled. When transaction fees are collected, they are deducted from TotalIssuance, recycling them back into the system for future emission.

Subnet creation cost is recycled (mostly). Subnet creation fees are recycled, except for one TAO used to initialize the subnet’s TAO liquidity pool.

Subnet “burn cost” is dynamic. The burn cost is a required amount of TAO to be recycled. It is dynamic, lowers gradually, and doubles every time a subnet is created.

New subnets are inactive for about a week. New subnets cannot be started for 7 × 7,200 blocks (roughly one week) after registration, and receive no emissions while inactive.

Neuron registration fees also have a recycling leg. When a user registers a hotkey on a subnet to participate as a miner or validator, they pay a TAO registration fee, and alpha tokens worth the swap value of that fee are taken from the subnet’s alpha liquidity pool and recycled.

True “burn” does exist, but it’s not the main supply story. Several contexts exist where subnet alpha tokens are burned, including creator-emissions burning under certain ownership/permission conditions.

Why this matters: a recycled TAO is not a permanently destroyed TAO. It is a TAO temporarily removed from circulation that can be re-emitted later. From a float-dynamics view, recycling is a supply sink today and a supply source later. That changes how you should think about “supply shocks” and how cleanly halving narratives translate into market impact.

Governance and parameter control: who can change the money knobs

Bittensor’s token design has real parameter risk because a meaningful portion of “tokenomics” is embedded in on-chain configuration. Understanding who can change what is part of understanding TAO.

Bicameral governance. The governance overview describes a transition away from a single privileged sudo key into a bicameral legislature where the Triumvirate creates proposals and the Senate approves them. Triumvirate members are described as Opentensor Foundation employees, while the Senate is formed from the top K delegate hotkeys.

Proposal execution conditions. A proposal executes only after it receives (50% + 1) Senate approvals and then a Triumvirate member closes it.

Rate limits are explicit policy. For example, the delegate take rate limit is 216,000 blocks (about 30 days).

Subnet-level economic parameters exist. Subnet hyperparameters include parameters like MinBurn (minimum range for dynamic burn cost) and MinDifficulty (minimum range for PoW registration difficulty), alongside flags controlling whether registrations are allowed.

Subnet slot scarcity is enforced. Subnet deregistration includes a hard subnet limit of 128 occupied slots, an immunity period of 4 months (864,000 blocks), and notes the feature deployed on September 17, 2025.

Governance here is not abstract. It governs emissions distribution logic, registration policy, and the operational constraints that decide which participants can earn and which can’t. That is tokenomics.

Liquidity structure view: TAO’s “real float” is what’s left after three sinks

TAO’s market behavior is better explained by float geometry than by fully diluted valuation narratives. CoinGecko itself calls FDV “theoretical” and flags that supply increases can move price. That’s not a footnote for TAO. It is the point.

In practice, tradable TAO is what remains after three sinks:

1) Root staking and validator plumbing. Root is the only subnet where TAO is staked directly without being exchanged for another token, and staking on root entitles stakers to subnet alpha dividends (claimable or swapped back via “root claim”).

2) Subnet “staking” that is actually a position in alpha. Staking TAO to a subnet is swapping TAO into the subnet token, and unstaking is swapping back out.

3) Protocol-owned liquidity sitting in pools. Because emissions inject reserves into AMMs, the ecosystem accumulates depth across many TAO/alpha pools over time. That TAO is “circulating” in some dashboards, but it is not sitting on exchange books ready to sell. It is in liquidity reserves, and its price impact is shaped by pool curves.

Then there is the reflexive part. TAO emissions to stakers are sourced by swapping a portion of alpha emissions to TAO through the subnet’s liquidity pool. That means some of the system’s yield distribution is literally mediated by AMM market impact.

Even the “simple” emission number matters for liquidity. If total block emission is 0.5 TAO per block and blocks are about every 12 seconds, that implies roughly 3,600 TAO/day of new TAO entering the system before recycling effects and distribution paths are considered.

This is why TAO can look “scarce” on liquid venues even when the protocol is emitting meaningfully. Emissions are not a straight pipe into spot sell pressure. They route through subnet pools, tempo-based distribution, validator takes, and user-level decisions about whether to hold alpha exposure or rotate back into TAO.

Risks: TAO is a liquidity-structure trade, not a simple scarcity trade

TAO’s design is coherent, but it is not forgiving. It rewards actors who understand where liquidity actually lives and how incentive routing creates second-order effects.

Dominant risk: subnet bank-runs driven by flow-based emissions.

The emissions regime states that, as of November 2025, TAO emissions across subnets moved to a flow-based model where emissions are based on net TAO inflows from staking activity, and that subnets with sufficiently negative net flows receive zero emissions.

That single rule creates a classic liquidity spiral:

Trigger path: a subnet loses narrative or performance credibility, users unstake, net flow turns negative.

Mechanism path: negative net flow pushes emissions toward zero, which slows liquidity growth in that subnet’s pool. Lower liquidity increases slippage for exits and entries, which raises the cost of staying or rotating, which can accelerate further exits. This is amplified by constant-product AMM structure.

Who bears it: alpha holders first (they are holding the asset whose exit price deteriorates), then validators and miners tied to that subnet’s reward stream, then TAO holders indirectly if alpha-to-TAO conversions become more disorderly and correlated across multiple large subnets.

Why it’s dominant: it is endogenous. It does not require hacks, governance failure, or external regulation. It can happen in a “working as designed” regime purely from positioning and crowd behavior. TAO’s effective float can tighten, then suddenly loosen, not because of unlocks, but because a major subnet cohort decides to de-risk simultaneously. That is a liquidity event, not an emissions event.

Top 3 risks

  1. Subnet liquidity spiral (dominant). Trigger: sustained negative net staking flows in a major subnet. Mechanism: flow-based emissions cut to zero, liquidity growth stalls, AMM slippage rises, exits accelerate. Who bears it: alpha stakers, subnet miners/validators, then TAO spot via correlated alpha-to-TAO rotation. Measurable indicators: net flow metrics, emissions share trending toward zero, shrinking pool reserves, widening price impact on typical-size unstake swaps.
  2. Governance concentration and parameter volatility. Trigger: contentious parameter change (emissions logic, registration policy, hyperparameter shifts) during market stress. Mechanism: proposal power sits with a Triumvirate and approval with a stake-weighted Senate, so control can be concentrated and incentives can diverge from smaller holders. Who bears it: TAO holders and operators exposed to policy shifts (validators, subnet owners) through changed cashflow expectations. Measurable indicators: Senate seat concentration (stake share of top delegate hotkeys), cadence of governance proposals, and emergency-style parameter changes.
  3. Supply-model misreads driven by recycling. Trigger: market pricing assumes “issuance” maps cleanly to “circulating,” or assumes halving timing purely from circulating dashboards. Mechanism: transaction fees and several protocol costs are recycled by deducting them from TotalIssuance, which can decouple gross emission from observed circulating trajectories. Who bears it: traders and long-onlys positioned on simplistic halving and inflation narratives, plus subnet teams optimizing around misunderstood incentives. Measurable indicators: fee and registration activity, observed changes in TotalIssuance methodology, and divergence between “emitted per block” and “net circulating change.”

If you’re building on Bittensor or designing a subnet economy, this is one of the few ecosystems where token economy design is inseparable from market microstructure. In practice, teams often need tokenomics consulting help less with “emissions math” and more with slippage-aware incentive routing and adversarial liquidity planning.



This article is part of our Tokenomics Deep Dive series.