Interactive Tokenomics Breakdown

How Value Flows (and Doesn't) Through Bittensor

A mechanistic map of $TAO supply, emission, staking, subnet economics, and the gap between narrative and revenue. All numbers sourced from on-chain data and public disclosures as of March 2026.

21M Hard Cap, Bitcoin-Mirrored Halvings

TAO deliberately mirrors Bitcoin's supply curve: 21M max supply, halving-based emission reductions, fair launch (no ICO/pre-mine). But with critical differences — halvings trigger on issuance thresholds, not block heights, and a recycling mechanism delays them.

Max Supply
21,000,000
Hard cap, BTC-mirrored
Circulating
~10.77M
51.3% of max supply
Post-Halving Emission
0.5 τ/block
~3,600 TAO/day
Current Inflation
~12-13%
vs Bitcoin ~0.85%
Projected Supply Curve with Halvings
Hover over the chart to inspect values. Green markers indicate halving events. Dashed line shows Bitcoin's curve for comparison.
Recycling Mechanism

Transaction fees and neuron registration costs are deducted from TotalIssuance and returned to the emission pool. This effectively slows the clock to the next halving — the system "unreleases" tokens, extending the subsidy period. Unlike Bitcoin where fees go to miners on top of block rewards, TAO fees replace future emissions.

Permanent Burns

Subnet registration costs are permanently burned — removed from max supply entirely. Each new subnet slot costs TAO that is destroyed forever. This creates genuine deflationary pressure on the cap, but at current registration volumes the burn rate is small relative to emissions (~dozens of TAO vs ~3,600/day emitted).

Halving Mechanics — Not Like Bitcoin

Bitcoin halvings occur at fixed block heights (~210K blocks). TAO halvings trigger when cumulative issuance reaches 50% of remaining supply. The recycling mechanism means halvings arrive later than a simple schedule would predict. First halving occurred December 15, 2025. Second halving estimated ~2029 depending on recycling rates.

Where New TAO Goes: The Three-Way Split

Every 12 seconds, 0.5 TAO is created and distributed across every active subnet. Within each subnet, rewards split between miners, validators, and the subnet owner. The Taoflow model determines how much each subnet gets.

Emission Split Calculator

Adjust parameters to see how TAO flows change. Default values reflect post-halving network state.

Block Reward 0.50 τ
Validator Take Rate 18%
Subnet Count 128
Taoflow: Cross-Subnet Allocation

Since November 2025, emission allocation across subnets is driven by net TAO inflows (staking minus unstaking), smoothed by an EMA with ~30-day half-life. Subnets with positive net staking inflows earn emissions; subnets with negative flows earn zero.

This replaced a price-based model that was gamed within weeks — Subnet 28 attracted meme-coin speculation, captured disproportionate emissions, and had to be killed by the Foundation using root stake privileges. Taoflow is harder to game but still rewards capital inflows rather than AI quality directly.

Following the Money: From Block Creation to End User

The full lifecycle of value in Bittensor: token creation → subnet allocation → intra-subnet distribution → staking/unstaking dynamics → external revenue (or lack thereof). Click any node for details.

The Subsidy Problem

At current prices (~$340), daily emissions of ~3,600 TAO = ~$1.22M/day in new supply (~$446M/year) entering the market as selling pressure. This entire amount must be absorbed by new capital inflows — either from buyers on exchanges or from stakers locking tokens — or the price declines. Actual external revenue from subnet services ($3–15M/year) covers approximately 0.7–3.4% of this emission cost.

76% Staked, Thin Float, Volatile Unstaking

Most TAO is locked in staking — creating scarcity that amplifies price movements in both directions. Understanding where stake flows is essential to understanding the token dynamics.

Total Staked
~8.19M τ
76% of circulating supply
Free Float
~2.58M τ
24% — highly illiquid
Root Staking APY
~4-5%
Denominated in TAO
Subnet Alpha APY
14-18%+
In alpha tokens, not TAO
Staking Ratio vs Free Float
76% STAKED
24% FREE

With only ~2.58M TAO freely circulating, even modest selling pressure can move the price significantly. Conversely, new buy demand faces thin order books. This is mechanically similar to a low-float IPO — amplified volatility in both directions.

Root Staking

Delegate TAO to validators on the root network (Subnet 0). Rewards paid in TAO at ~4-5% APY. Lower risk — no alpha token exposure. Validators set take rates (default 18%), remainder flows to delegators. Simple, lower yield, denominated in TAO.

Subnet Alpha Staking

Swap TAO for a subnet's alpha token via the on-chain AMM. Higher nominal yields (14-80%+) but denominated in alpha tokens, not TAO. If the alpha/TAO exchange rate falls, you lose in TAO terms even with high APY. Constant-product AMMs mean large exits crater the alpha price — concentrated liquidity risk.

dTAO AMM Mechanics

Each subnet's alpha token is traded via a x·y = k constant-product AMM. When you "stake" into a subnet, you swap TAO for alpha — moving the exchange rate. When you unstake, you swap alpha back for TAO. Large positions face significant slippage. "Market caps" quoted for subnets are theoretical (AMM reserve × implied price) and may not reflect realizable liquidity for meaningful exits.

Unstaking Cascade Risk

With 76% of supply staked through thin AMM pools, a loss of confidence in any major subnet triggers a cascade: alpha holders rush to swap back to TAO → alpha price craters → triggers more unstaking (as yields collapse in TAO terms) → net outflows push subnet emissions to zero (Taoflow model) → no rewards → accelerated exit. This is the decentralized equivalent of a bank run, but on constant-product curves with no circuit breakers.

The Subnet 28 meme-coin incident demonstrated this dynamic. Capital flowed in speculatively, captured emissions, then exited — destroying the subnet. The Foundation intervened using root stake privileges that it has since committed to relinquishing.

The 5.38M TAO Black Box

Bittensor claims "fair launch" — no ICO, no pre-mine, no VC allocation. But from network launch (January 2021) through subnet launch (October 2023), 5.38M TAO were mined with limited public documentation of who received them.

Token Distribution — What We Know
Pre-Subnet Mining (Jan 2021 – Oct 2023) 5.38M τ (51% of current supply)
Post-Subnet Emission (Oct 2023 – present) ~5.39M τ (49% of current supply)
Known holders of significant positions:
EntityRelationshipNotes
Polychain CapitalInvestorLead investment; amount undisclosed
Digital Currency Group (DCG)Investor / Grayscale parentGrayscale trust creator; clear conflict of interest
Dao5InvestorEarly backer
Opentensor FoundationCore teamControls root validators + PoA nodes
Early minersNetwork participantsUnknown concentration; no public registry
The "Fair Launch" Tension

A critic estimated that ~62.5% of TAO sits with insiders and early participants. The project categorizes itself as a fair launch because there was no token sale, but institutional investors (Polychain, DCG, Dao5) acquired significant positions through early mining or OTC deals. DCG's dual role as investor and Grayscale trust sponsor creates a structural promotion incentive. The lack of a transparent allocation table is unusual for a $3.4B network.

128 Subnets: What Actually Ships?

Each subnet defines its own AI task, evaluation criteria, and reward mechanism. They range from LLM inference (Chutes) to decentralized training (Templar) to sports prediction (Score). Most have no external revenue.

SN64 — Chutes
Serverless AI Inference
9.1T tokens processed · 400K+ users
Revenue: $1.3–2.4M/yr
Emission subsidy: ~$52M/yr
Subsidy ratio: 22–40:1
Has Revenue
SN3 — Templar
Decentralized LLM Training
Trained Covenant-72B (1.1T tokens, 70+ contributors)
Market cap: ~$135M
Revenue: $0
Proof-of-concept milestone
No Revenue
SN4 — Targon
LLM Inference Verification
Client: Dippy AI (8.6M users)
Raised $10.5M Series A
Product-market signals
Has Clients
SN62 — Ridges AI
Autonomous Coding Agents
Claims 80.3% SWE-Bench score
Competitive with frontier agents
Revenue: not disclosed
Unverified
SN51 — Lium
P2P GPU Rental
500+ H100 GPUs onboarded
Competes with Akash, io.net
Revenue: not disclosed
Early Stage
SN8 — PTN
Proprietary Trading Network
Trading signal generation
Inherently hard to validate
Performance claims unaudited
Unverifiable
The Long Tail Problem

Of 128 active subnets, approximately 2-3 have identifiable external revenue, 5-8 have demonstrated technical milestones, and the remaining 115+ operate primarily on emission capture with no publicly documented commercial traction. Many subnet alpha tokens have market caps in the tens of millions despite generating zero revenue — their value derives entirely from TAO emission allocation and speculative inflows.

$3.4B Market Cap vs. $3–15M Revenue

The single most important number for any investor: the ratio between what the network produces in real economic value and what the market says it's worth. Compare to centralized AI infrastructure companies and other crypto-AI tokens.

Network Market Cap
~$3.4B
External Revenue
$3–15M
Annualized estimate
Revenue Multiple
230–1,100x
vs 40-60x for hot AI cos
Annual Emission Cost
~$446M
At $340/TAO
The Pricing Vise
SELF-HOSTING CEILING Users can run open-source models themselves → caps max price BITTENSOR SUBNET PRICING Currently subsidized 22-40x by emissions Unsubsidized: ~$1.41/Mtok — MORE expensive than centralized HYPERSCALER FLOOR Together.ai $0.88/Mtok · DeepSeek $0.40-0.80/Mtok · Economies of scale

Bittensor subnets are caught between two forces: users can self-host open-source models (ceiling), and hyperscalers offer the same models cheaper at scale (floor). Current "85% cheaper than AWS" pricing is real for users but funded entirely by inflation. As halvings reduce the subsidy, pricing must either rise (becoming uncompetitive) or margins collapse.

Comparative Valuation
MetricBittensorAkashTogether.aiCoreWeave
Valuation ~$3.4B~$130M~$12B (private)~$35B (IPO)
Revenue (ann.) $3-15M~$5-10M~$200M+~$2B+
Rev Multiple 230-1,100x13-26x~60x~17x
GPU Utilization Unknown~80%High~90%+
Switching Costs Zero (OSS models, std APIs)LowMediumMedium-High
Thesis A: Narrative / Scarcity

TAO performs as a token because: Bitcoin-mirrored supply, halving scarcity, Grayscale ETF catalyst, AI sector beta, 76% staking ratio = thin float. Fundamentals secondary to tokenomics + narrative. This thesis has worked in March 2026.

Thesis B: Sustainable AI Network

Bittensor builds commercially competitive decentralized AI at scale. Evidence: deeply unfavorable. $3-15M revenue on $3.4B valuation. Emission subsidies mask uncompetitive pricing. Zero switching costs. Rewards driven by stake, not quality. Weight-copying unsolved.