Written by: Pine Analytics

Translated by: Saoirse, Foresight News

The current price of TAO is approximately $275, with a market cap of $2.6 billion and a fully diluted valuation of $5.8 billion. The project has received endorsements from Grayscale (an ETF listing application was submitted to the New York Stock Exchange in December 2025) and has also been publicly recognized by NVIDIA CEO Jensen Huang. Furthermore, the narrative surrounding token supply is highly attractive: the total cap is set at 21 million tokens, following a Bitcoin-like halving mechanism. After the first halving in December 2025, the daily issuance will decrease from 7,200 tokens to 3,600 tokens. Within a year, the number of subnets has increased from 32 to 128, and Templar's Covenant-72B training also proves that decentralized computing power can run large language models with competitive benchmarks.

This report does not deny the above facts. What we want to explore is whether the economic model of the network can generate real external revenue to support the current scale of valuation, and how it competes with centralized service providers and self-hosted computational power.

Bittensor (TAO) token issuance distribution ratio

How network value circulates

Bittensor has four types of participants:

• Subnet owners build specialized AI markets and receive 18% of the TAO issuance rewards from the subnet;
• Miners execute AI tasks (inference, training, data processing) and receive 41%, totaling approximately 1,476 TAO daily, with an annual value of about $148 million;
• Validators score the outputs produced by miners and receive 41%;
• Stakers put TAO into the subnet liquidity pool in exchange for subnet-exclusive tokens.

Under the Taoflow model, the reward share for a subnet is determined by the net inflow of staked TAO, with no rewards if the net inflow is negative. The top ten subnets control about 56% of the total network issuance.

TAO is the universal token across the entire network: miners register, validators stake, subnet tokens are purchased, and service payments must use TAO. Theoretically, subnet activities will generate structural demand for the underlying token.

Comparison of Bittensor subnet Chutes (SN64) with the inference costs of centralized service provider LLaMA 70B model

Current state of demand

Transparent supply vs opaque demand

The supply side of Bittensor is highly transparent: 3,600 TAO are allocated daily through a programmed process, halving rules are hardcoded, and stake rates (approximately 70%), allocation ratios, and flowing data are all recorded on-chain.

However, the demand side is entirely opaque. There is no unified dashboard to track external revenue by subnet, and the actual calls for AI services (inference, computation, training) happen off-chain and are not recorded on the blockchain. Investors can only infer demand through indirect indicators like stake flows, subnet token prices, and self-reported data from the project. This opacity is structural and not a temporary phenomenon. The blockchain records only token circulation, not API calls.

The following is the most complete picture of the demand side as of March 2026.

Chutes (SN64): Low prices rely entirely on subsidies

Chutes accounts for 14.4% of total network issuance, the highest among all subnets. Developed by Rayon Labs, it provides open-source model serverless inference services, priced 85% lower than AWS and 10%–50% lower than Together AI. Its usage data stands out in the ecosystem: over 400,000 users (over 100,000 API users), daily request volume exceeding 5 million, and a cumulative processing of 91 trillion tokens, with the average token generation rising from 6.6 billion to 101 billion over three days. It is also a leading inference service provider on OpenRouter, with some models performing better than centralized competitors.

However, this low price does not come from operational efficiency, but from subsidies.

Based on a 14.4% share, Chutes receives approximately 518 TAO daily, with an annual value of about $52 million. Yet its external annual revenue is only around $1.3 million to $2.4 million (the higher figure is self-reported by the team and not independently audited). The protocol's subsidy ratio for this subnet is approximately 22:1 to 40:1. For every $1 paid by users, the network must release $22 to $40 of TAO through inflation to subsidize.

If the subsidies are removed, based on its daily processing of about 101 billion tokens, the cost price would be approximately $1.41 per million tokens. The current prices in the centralized market are:

• Together.ai's LLaMA 3.3 70B Turbo is about $0.88 / million tokens;
• DeepSeek V3 is about $0.40–0.80;
• Smaller models can go as low as $0.18.

This means that without subsidies, Chutes' price would be 1.6–3.5 times more expensive than centralized solutions. The so-called 85% cost advantage is completely reversed; its low price is essentially subsidized by TAO holders through inflation rather than structural efficiency derived from decentralization.

When the next halving occurs (expected by the end of 2026 or 2027), either the price must double, or miners will exit the market, or the gap between subsidies and revenue will widen further.

Some may compare this to the early internet subsidizing customer acquisition, but Uber, DoorDash, and AWS established switching costs during their subsidy periods: proprietary platforms, driver networks, and corporate ecosystems. However, the Bittensor subnets have no barriers: the models are open source, and the interfaces are standardized, allowing users to switch service providers at zero cost. Once subsidies withdraw, there will be no locking mechanism to retain users.

Rayon Labs also operates SN56 and SN19, together controlling about 23.7% of the total network issuance, with no disclosed external revenue. A single team controls nearly a quarter of the network's incentive distribution.

Targon, Templar, and other subnets

Targon (SN4) is the highest income subnet operated by Manifold Labs, offering confidential GPU computing services to enterprises, with an estimated annual revenue of about $10.4 million, corresponding to a valuation of $4.8 million, with a price-to-sales ratio of about 4.6 times, making it the most solid valuation in the ecosystem. However, $10.4 million is merely forecasted data cited in various reports, not audited numbers.

Templar (SN3) completed the Covenant-72B training and has a market cap of $98 million, but external revenue is zero. Training APIs and enterprise sales are still underway, and no paid products have been launched yet.

The remaining 120+ subnets either have no public revenue or are still in early product stages, primarily relying on token issuance subsidies to survive.

Overall overview

The total identifiable annual revenue on the demand side across the network is only about $3 million to $15 million. The annual subsidy for just Chutes (around $52 million) exceeds the upper limit of the entire network's external revenue.

With a market cap of $2.6 billion, its income multiple is about 175–200 times; based on a fully diluted valuation of $5.8 billion, it approaches 400 times. In recent years, centralized AI computing enterprises have only achieved financing valuations of 15–25 times forward revenues, and high-growth SaaS rarely maintains multiples above 50 times over the long term. The valuation multiple of Bittensor is 4–10 times that of aggressive industry targets.

The significant gap between valuation and demand fundamentals indicates that the market prices TAO almost entirely based on supply-side scarcity (halving, staking lockup), institutional catalysts (Grayscale ETF expectations, sureness of listing), and sentiment in the AI sector, rather than real economic output. These are indeed price-driving factors, but they are entirely different from the logic of "Bittensor creating sustainable value as an AI service network."

Comparison of AI capital expenditures by super-scale cloud vendors and Bittensor (TAO) annual subsidy scale

Pricing dilemma: squeezed from both ends

Subnets face pressure from both sides:

• Above: Self-hosted ceiling

All models on the platform are open source with weights public; a single H100 running a 70B model has a combined daily cost of only $40–50, and tools like vLLM and Ollama make local deployment extremely simple. NVIDIA's next-generation chips will significantly reduce inference costs. Institutions with sufficient usage will find self-construction cheaper.

• Below: Pressure from cloud giants

Microsoft, Google, Amazon, and Meta are estimated to exceed $200 billion in AI capital expenditures by 2025, having hardware priority quotas, dedicated data centers, corporate client relationships, and the ability to subsidize AI with cash flows from other businesses. Bittensor's total incentive budget for the year (about $360 million) doesn't even match Microsoft's AI infrastructure investment for a week. Professional service providers are also using VC subsidies to compete on low prices with open-source models.

Subnet pricing is compressed into a very narrow range while also bearing costs unique to decentralization: token friction, validator node expenses, subnet owner shares, network latency, etc.

Moat issues

Even if a subnet makes valuable services, the underlying models and methods are naturally public: Covenant-72B uses the Apache protocol, and technical papers are publicly published. Any competitor can replicate directly without participating in the TAO ecosystem.

Traditional moats (proprietary technology, network effects, switching costs, brand) do not hold:

• Technology is open-source;
• Network effects belong to TAO, not individual subnets;
• Model weights are consistent, and the cost of user switching is zero.

The community believes that the incentive mechanism is the moat, but this relies on continued large token issuance, while each halving will continuously shrink the incentive budget.

What is TAO really trading?

With a market cap of $2.6 billion, the price of TAO does not reflect demand fundamentals, and annual revenues of $3 million–$15 million cannot support it under any traditional framework. What the market is trading is: Bitcoin-like scarcity, expectations for Grayscale's ETF, sector rotation in AI, and the long-term optionality value of decentralized AI. These are all reasonable speculative factors but stem entirely from the supply side and market sentiment.

If you hold TAO based on scarcity and narrative, you may still profit even if demand is weak; but if you believe Bittensor will become a truly scalable AI service network, there is currently no evidence, and it faces significant structural barriers. Investors should clearly distinguish their investment logic.