TAO - Bittensor, Investor Metrics
Public
By Jump LlamaTAO, TVL, Subnets revenue, TAO volume and other metrics
- tao
- bittensor
- subnets
- metrics
- ai
76Views
🧠 Bittensor (TAO) — Decentralized AI Investor Dashboard
The decentralized AI network rewriting incentive economics for machine learning. Bittensor operates 128+ active subnets (4× growth from ~32 in early 2025), each competing for TAO emissions across AI verticals: compute, inference, data curation, protein folding, prediction markets, and AI search. The dTAO upgrade (Feb 2025) gave each subnet its own alpha token priced via bonding curves — making TAO the reserve currency of decentralized AI.
Institutional access is accelerating. Yuma (DCG-backed) launched the Bittensor Total Market Fund on June 27, 2026, providing single-vehicle exposure to TAO + subnet alpha tokens. [3] Grayscale increased TAO's weighting in its Decentralized AI Fund to 43% (Q2 2026 rebalance). Spot TAO ETF filings from Grayscale and Bitwise are pending SEC review, with decision window expected August 2026. [5]
TAO trades ~73% below its March 2024 ATH of $757.60 at a ~$1.97B market cap — institutional accumulation is building at a steep historical discount.
⛏️ Tokenomics — Bitcoin Hard Cap + Halving Schedule = Supply Squeeze
TAO mirrors Bitcoin's scarcity model with an AI utility layer. Hard cap: 21M TAO total. Daily emissions have been halved (reducing from 7,200 TAO/day), with an estimated 65–70%+ of circulating supply staked or delegated — leaving a thin liquid float for new demand to absorb. [1]
dTAO bonding curves compound the squeeze: Each of the 128+ subnets requires TAO staked into its own liquidity pool to purchase that subnet's alpha token. As institutional and retail capital rotates into high-performing subnets, TAO is locked as the reserve asset. Expanding to the theoretical 256-subnet maximum (each requiring ~700 TAO at registration) could pull an estimated ~$32M worth of TAO out of circulation with no timeline confirmed. [5, unverified]
Current supply context: ~7.6M TAO circulating out of 21M max — roughly 64% of supply is yet to be minted via future emissions, meaning long-term dilution risk is real but gated by the halving schedule. [2]
🔗 On-Chain Activity — TVL, Fees & Network Revenue
Bittensor's EVM layer hosts emerging DeFi primitives including TaoFi (CLMM DEX) and wTAO bridge infrastructure (ERC-20 representation of native TAO on Ethereum). EVM-layer TVL is early-stage — the core value accrual mechanism is TAO emissions flowing to subnets, not traditional DeFi TVL.
Chain fees = transaction fees generated by Bittensor network usage. Revenue = the portion accruing to validators and stakers. Growing fees signal real network utility — AI compute subnet usage, not just token trading. Watch fees/revenue trends as the leading indicator of compute demand expanding beyond the crypto-native user base.
TaonSquare (launched May 2026): Consumer-facing subnet output directory — the shift from miner-facing infrastructure toward user-facing AI products could drive the next leg of adoption and fee growth.
🏦 Institutional Momentum — Funds, ETF Filings & Grayscale
The institutional access layer is being built right now:
Yuma Total Market Fund (June 27, 2026): DCG-backed Yuma launched the first institutional product giving diversified exposure to TAO + subnet alpha tokens in a single vehicle — no need to pick individual subnets. Combined subnet value per Taostats: ~$300M (vs Yuma's $900M claim — a ~3× discrepancy to monitor). [3]
Grayscale Decentralized AI Fund: Raised TAO's weighting to 43% during Q2 2026 quarterly rebalance — the largest allocation in the fund. Weighting has since adjusted as NEAR Protocol gained ground (~44% NEAR vs ~20% TAO currently). [3]
Spot TAO ETF (Grayscale + Bitwise): SEC decision window expected August 2026. If approved, ETF inflows would create a structural demand channel at current price levels. [5, unverified — monitor SEC EDGAR filings directly]
Risk flag: Validator concentration and governance centralization remain open concerns flagged by independent analysts — large validators control a disproportionate share of emissions distribution. [2]
🤖 Subnet Economy — 128+ Competing AI Markets, Each with Its Own Token
The subnet economy is Bittensor's differentiated moat. Every subnet issues its own alpha token priced via a TAO-denominated bonding curve. Subnet selection = market signal. Emissions flow to subnets based on validator votes (root-network) and increasingly on market demand under dTAO mechanics.
Scale: 10 subnets at launch → 64 (Apr 2026) → 128+ active today — a 4× expansion in ~12 months driven by the economics of capturing emission streams. A subnet capturing just 3% of total emissions earns meaningful daily income in TAO terms, incentivizing continuous new entry. [2]
Real compute revenue: $43M in AI compute revenue generated across 128 subnets in Q1 2026. [5, unverified — corroborate via Taostats] Use cases span text generation, image synthesis, protein folding, fraud detection, decentralized compute, and AI search.
Covenant-72B (early 2026): The largest collaborative globally distributed pre-training run by model scale and compute — pre-trained on ~1.1 trillion tokens — proved Bittensor's architecture can produce frontier-quality models via decentralized coordination. [5] This is the technical credibility event that underpins the institutional thesis.
📡 Subnet Network Snapshot — Aggregate Pool Metrics
128 active subnets compete for 3,600 TAO/day post-halving emissions (~$763K/day, ~$278M/yr at $212 TAO). With Robin τ (May 2026), capacity is expanding to 256 slots — 128 new markets for AI teams to capture emission streams. Each registration costs ~700 TAO (~$148K), acting as a natural quality gate. [3]
All figures below from SubnetAIQ and SubnetRadar — external sources, unverified. Live data at taostats.io/subnets and subnetradar.com.
| Metric | Value |
|---|---|
| Active Subnets | 128 |
| Max Subnet Capacity (Robin τ) | 256 |
| Total TAO Locked in Pools | 7,266,189 TAO |
| USD Value of All Pools | ~$1.54B (at $212/TAO) |
| Average Alpha Token Price | 0.0103 TAO per alpha |
| Median Alpha Token Price | 0.0060 TAO per alpha |
| Daily TAO Emissions | 3,600 TAO/day |
| Next Halving (est.) | ~1,340 days (3,600 → 1,800/day) |
| Subnet Alpha Total Mcap | ~$460M–$560M (est.) |
| Subnet Alpha as % of TAO Mcap | ~27% (CoinGecko, Mar 2026) [11] |
| Top Subnet Liquidity Share | Chutes SN64 — 2.6% of total pool |
| Q1 2026 AI Usage Revenue | ~$43M (Canonical Labs) [5] |
| Emissions-to-Revenue Ratio | 22:1–40:1 (inference subnets) [5] |
🏆 All Named Subnets — Ranked by Market Cap (External Sources, Unverified)
Sources: SubnetRadar live rankings [10], SubnetAIQ liquidity data [8], CoinGecko March 2026 [11], Canonical Labs [5]. Market caps at ~$212 TAO price unless noted. † = March 2026 peak estimate (TAO was ~$700+ at time).
| # | Subnet | SN | Category | Token Price | Market Cap | Alpha Price (TAO) | TAO Locked |
|---|---|---|---|---|---|---|---|
| 1 | Chutes | 64 | Inference | $14.63 | $80.2M | 0.0654 | 188,353 |
| 2 | Targon | 4 | Inference | $11.24 | $61.6M | 0.0501 | 125,796 |
| 3 | lium.io | 51 | Compute | $10.51 | $54.6M | 0.0452 | 107,736 |
| 4 | Score | 44 | Detection | $8.40 | $44.1M | 0.0395 | 65,375 |
| 5 | Affine | 120 | Training | $12.10 | $42.7M | 0.0553 | 76,693 |
| 6 | Vanta (PTN) | 8 | Trading | $6.94 | $37.1M | 0.0309 | 81,702 |
| 7 | iota | 9 | Training | $6.77 | $36.3M | 0.0307 | 57,327 |
| 8 | Actual | 95 | Inference | $11.87 | $34.6M | — | — |
| 9 | Training | 3 | Training | $4.87 | $25.9M | 0.0220 | 67,800 |
| 10 | NOVA | 68 | DeSci | $4.92 | $25.0M | 0.0230 | 43,480 |
| 11 | Gradients | 56 | Training | — | — | 0.0192 | 54,642 |
| 12 | Ridges AI | 62 | — | — | ~$50.8M† | — | — |
| 13 | Hippius | 75 | Infra | — | ~$41.3M† | — | — |
| 14 | Apex | 1 | Agents | — | $10.2M | — | — |
| 15 | Vidaio | 85 | — | — | — | 0.0075 | — |
| 16 | Bitcast | 93 | — | — | — | 0.0131 | — |
| 17 | Astrid | 127 | — | — | — | 0.0038 | — |
| 18 | Leadpoet | 71 | — | — | — | 0.0051 | — |
| 19 | Harnyx | 67 | — | — | — | — | — |
| 20 | Compelle | 82 | — | — | — | — | — |
| 21 | Minos | 107 | — | — | — | — | — |
| 22 | Bitsota | 94 | — | — | — | — | — |
| 23 | Bitstarter | 91 | — | — | — | — | — |
| 24 | Liquidity | 77 | DeFi | — | — | — | — |
| 25 | ByteLeap | 128 | — | — | — | — | — |
| 26 | qbittensor | 63 | — | — | — | — | — |
~102 additional subnets (SN IDs 2, 5–7, 10–43, 45–50, 52–55, 57–61, 65–66, 69–70, 72–74, 76, 78–81, 83–84, 86–90, 92, 96–106, 108–119, 121–126) are active but not individually documented in available research. Full list at taostats.io/subnets.
💧 Top 10 Subnets by TAO Locked
Per SubnetAIQ — external, unverified. TAO price ~$212.
| # | Subnet | SN | TAO Locked | Pool Share | Alpha Price |
|---|---|---|---|---|---|
| 1 | Chutes | 64 | 188,353 τ | 2.6% | 0.0654 τ |
| 2 | Targon | 4 | 125,796 τ | 1.7% | 0.0501 τ |
| 3 | lium.io | 51 | 107,736 τ | 1.5% | 0.0452 τ |
| 4 | Vanta | 8 | 81,702 τ | 1.1% | 0.0309 τ |
| 5 | Affine | 120 | 76,693 τ | 1.1% | 0.0553 τ |
| 6 | Training | 3 | 67,800 τ | 0.9% | 0.0220 τ |
| 7 | Score | 44 | 65,375 τ | 0.9% | 0.0395 τ |
| 8 | iota | 9 | 57,327 τ | 0.8% | 0.0307 τ |
| 9 | Gradients | 56 | 54,642 τ | 0.8% | 0.0192 τ |
| 10 | NOVA | 68 | 43,480 τ | 0.6% | 0.0230 τ |
Top 10 TAO Locked Total: ~867,704 τ (~$184M) Network Total: 7,266,189 τ (~$1.54B) Top 10 Share of Total: ~11.9%
TAO locked ≠ market cap. High alpha price with moderate TAO locked (e.g. Affine SN120: 0.0553 τ, 76,693 locked) signals strong demand momentum relative to supply.
🏷️ Subnet Category Distribution (128 Subnets)
Per SubnetRadar research directory [9] — external, unverified.
| Category | # Subnets | Top Example |
|---|---|---|
| Training | 9 | Affine (SN120), iota (SN9) |
| Trading | 9 | Vanta/PTN (SN8) |
| Detection | 9 | Score (SN44) |
| Infra | 8 | Hippius (SN75) |
| Yuma subnets | 10 | (DCG team, 10 subnets) |
| Prediction | 7 | — |
| DeFi | 7 | Liquidity (SN77) |
| Agents | 6 | Apex (SN1) |
| Inference | 6 | Chutes (SN64), Targon (SN4) |
| Security | 5 | — |
| Hardware | 5 | — |
| Macrocosmos | 5 | Apex (SN1) |
| DeSci | 4 | NOVA (SN68) |
| Compute | 4 | lium.io (SN51) |
| Data | 4 | — |
| TAO Ventures | 4 | — |
| Latent Holdings | 3 | — |
| Robotics | 3 | — |
| Sports | 3 | — |
| Storage | 2 | — |
| Code | 2 | — |
| Quantum | 2 | — |
| Marketing | 2 | — |
| Manifold | 2 | Vanta (SN8) |
| 3D | 1 | — |
| Cryptography | 1 | — |
| Rayon Labs | 1 | Chutes (SN64) |
| Rizzo | 1 | Targon (SN4) |
Key insight: Inference + Compute + Training = ~19 subnets but capture >60% of top-10 market cap, confirming AI compute as the dominant value vertical.
🔦 Top 5 Subnet Deep Dives — Revenue, Moat & Risk
1. 🟢 Chutes (SN64) — Serverless AI Inference | $80.2M Mcap
What it does: Serverless GPU marketplace for AI inference. API-compatible with OpenAI, priced 85% below AWS. Ranked #1 inference provider on OpenRouter — the major LLM aggregator serving millions of developers. [5] This is the clearest proof of real external demand in the Bittensor ecosystem. Moat: OpenRouter integration creates distribution without direct user acquisition cost. Switching cost for developers grows with integration depth. Risk: TAO-subsidized pricing; if emissions drop below demand-side revenue, pricing pressure increases. Emission-to-revenue ratio: ~22:1 to 40:1. [5] Key metrics: $80.2M mcap, $14.63 token price, 188,353 TAO locked, 0.0654 τ/alpha
2. 🟡 Targon (SN4) — Confidential GPU Compute | $61.6M Mcap
What it does: Privacy-preserving GPU inference using TEEs (Trusted Execution Environments). Target market: enterprise AI workloads requiring data confidentiality. Claims 4M+ users. [5, unverified] Moat: Privacy layer is a genuine differentiator vs Chutes and AWS — enterprise compliance requirements create a real demand niche. Risk: TEE security assumptions; hardware dependency; slower to scale than pure-software inference. Key metrics: $61.6M mcap, $11.24 token price, 125,796 TAO locked, 0.0501 τ/alpha
3. 🟡 lium.io (SN51) — Compute Marketplace | $54.6M Mcap
What it does: Decentralized compute marketplace connecting GPU providers with AI workloads. Key metrics: $54.6M mcap, $10.51 token price, 107,736 TAO locked, 0.0452 τ/alpha. Third-largest pool by TAO locked.
4. 🟡 Score (SN44) — AI Detection | $44.1M Mcap
What it does: Detection subnet — fraud, anomaly detection, and classification tasks. Enterprise-facing use case with direct commercial demand potential. Key metrics: $44.1M mcap, $8.40 token price, 65,375 TAO locked, 0.0395 τ/alpha
5. 🟡 Affine (SN120) — Distributed Training | $42.7M Mcap
What it does: Collaborative model training. High alpha price (0.0553 τ) relative to TAO locked (76,693) suggests strong speculative premium on training compute demand. Note on SN3 (Training/prev. Templar): Executed Covenant-72B — the largest distributed pre-training run on Bittensor, 1.1T tokens, outperformed LLaMA-2-70B on MMLU. [5] Technical credibility anchor for the training vertical. Key metrics: $42.7M mcap, $12.10 token price, 76,693 TAO locked, 0.0553 τ/alpha