Original Title: The Bear Case for Bittensor (TAO)

Original Source: Pine Analytics

Original Translation: Saoirse, Foresight News

TAO is currently priced around $275, with a market cap of $2.6 billion and a fully diluted valuation of $5.8 billion. The project has backing from Grayscale (which filed for a U.S. ETF listing in December 2025) and has been publicly endorsed by NVIDIA CEO Jensen Huang. At the same time, the token supply narrative is very attractive: a maximum total of 21 million tokens, with Bitcoin-style halving mechanisms. After the first halving in December 2025, daily issuance drops from 7,200 to 3,600 tokens. Within a year, the number of subnets increased from 32 to 128, and Templar’s Covenant-72B training demonstrated that decentralized compute power can run large language models with competitive benchmarks.

This report does not deny these facts. What we want to explore is: Can the network’s economic model generate real external revenue supporting its current valuation, and how competitive is it when competing with centralized service providers and self-hosted computing power?

Distribution of Bittensor (TAO) Token Issuance

How Value Flows in the Network

Bittensor has four types of participants:

· Subnet owners build specialized AI markets and receive 18% of TAO issuance rewards;

· Miners perform AI tasks (inference, training, data processing) and receive 41%, totaling about 1,476 TAO daily, with an annual value of approximately $148 million;

· Validators score miner outputs and receive 41%;

· Stakers put TAO into subnet liquidity pools in exchange for subnet-specific tokens.

Under the Taoflow model, a subnet’s reward share is determined by net TAO staking inflow; if net inflow is negative, no rewards are given. The top ten subnets control about 56% of the total issuance.

TAO is a universal token across the entire network: used for miner registration, validator staking, subnet token purchases, and service payments. In theory, subnet activities create structural demand for the underlying token.

Comparative Analysis of Bittensor Subnet Chutes (SN64) and Centralized Service Providers’ LLaMA 70B Model Inference Costs

Demand Side Status

Transparent Supply vs. Opaque Demand

Bittensor’s supply side is highly transparent: 3,600 TAO are distributed daily according to a programmed halving rule, with staking rate (~70%), distribution ratios, and flow data all on-chain.

But demand is completely opaque. There is no unified dashboard tracking external income by subnet; actual AI service calls (inference, computation, training) happen off-chain and are not recorded on the blockchain. Investors can only infer demand indirectly through staking flows, subnet token prices, and self-reported data from project teams. This opacity is structural, not temporary. Blockchains record token flows but not API calls.

Below is the most comprehensive demand-side picture as of March 2026.

Chutes (SN64): Heavily Subsidized Low Prices

Chutes accounts for 14.4% of total issuance, the highest among all subnets. Developed by Rayon Labs, it offers open-source model serverless inference services, pricing 85% lower than AWS and 10–50% lower than Together AI. Its usage data dominates the ecosystem: over 400,000 users (including over 100,000 API users), more than 5 million requests daily, processing a total of 9.1 trillion tokens, with token generation rising from 6.6 billion to 101 billion over three days. It is also a leading inference provider on OpenRouter, with some models outperforming centralized competitors.

But this low price is not due to operational efficiency, only subsidies.

Based on a 14.4% share, Chutes earns about 518 TAO daily, with an annual value of roughly $52 million. Its external annual revenue is only about $1.3–2.4 million (the higher figure is self-reported, not independently audited). The subsidy ratio for this subnet is approximately 22:1 to 40:1. For every dollar paid by users, the network must release 22–40 dollars’ worth of TAO via inflation to subsidize.

Removing subsidies, based on its daily processing volume of about 101 billion tokens, the cost price is approximately $1.41 per million tokens. In contrast, the current centralized market prices are:

· Together.ai’s LLaMA 3.3 70B Turbo at about $0.88 per million tokens;

· DeepSeek V3 at about $0.40–0.80;

· Smaller models as low as $0.18.

This means that, without subsidies, Chutes 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 paid for by TAO holders through inflation, not structural efficiency from decentralization.

When the next halving occurs (expected late 2026 or 2027), prices will either double, miners will exit, or the subsidy-revenue gap will further widen.

Some compare this to early internet subsidies for customer acquisition, but Uber, DoorDash, and AWS built switching costs during their subsidy periods: proprietary platforms, driver networks, enterprise ecosystems. Bittensor subnets have no barriers: open-source models, standardized interfaces, users can switch providers at zero cost. Once subsidies fade, no locking mechanisms can retain users.

Rayon Labs also operates SN56 and SN19, controlling about 23.7% of total issuance, with undisclosed external income. A single team nearly controls a quarter of the network’s incentive distribution.

Targon, Templar, and Other Subnets

Targon (SN4), operated by Manifold Labs, is the highest revenue subnet, providing confidential GPU compute services for enterprises, with an estimated annual revenue of about $10.4 million, valuing it at $48 million, with a P/S ratio of about 4.6. This is the most solid valuation within the ecosystem. However, the $10.4 million figure is a forecast cited by multiple reports, not an audited number.

Templar (SN3) completed Covenant-72B training, with a market cap of $98 million, but zero external income. API training and enterprise sales are still in progress, and paid products have not yet launched.

More than 120 other subnets either have no public income or are still in early product stages, mainly surviving on token issuance subsidies.

Overall Overview

Total confirmed external demand-side annual revenue across the network is only about $3–15 million. Just one subnet, Chutes, with an annual subsidy of about $52 million, exceeds the entire network’s external income ceiling.

With a $2.6 billion market cap, the revenue multiple is roughly 175–200x; with a fully diluted valuation of $5.8 billion, nearly 400x. In comparison, centralized AI compute companies’ recent valuations are only 15–25x forward revenue, and high-growth SaaS rarely maintains over 50x long-term. Bittensor’s valuation multiple is 4–10 times that of aggressive industry targets.

The huge gap between valuation and demand fundamentals indicates that the market’s pricing of TAO is almost entirely based on supply-side scarcity (halving, staking lock-up), institutional catalysts (Grayscale ETF, exchange listing expectations), and AI sector sentiment, rather than real economic output. These are indeed price drivers, but they are completely disconnected from the logic of “Bittensor as a sustainable value-creating AI service network.”

Comparing Large-Scale Cloud AI Capital Expenditures with Bittensor (TAO) Annual Subsidy Scale

Pricing Dilemma: Being squeezed from both ends

Subnets face dual pressures:

· Top: Self-hosted cap

All models on the platform are open-source, weights are public, and running a 70B model on a single H100 costs only about $40–50 daily. Tools like vLLM and Ollama make local deployment very easy. NVIDIA’s new generation chips will further reduce inference costs. For high-volume users, building their own deployment is cheaper.

· Bottom: Cloud giants’ squeeze

Microsoft, Google, Amazon, and Meta will spend over $200 billion on AI infrastructure in 2025, with priority hardware quotas, dedicated data centers, enterprise relationships, and the ability to subsidize AI with other business cash flows. Bittensor’s entire annual incentive budget (~$360 million) is less than a week’s worth of Microsoft’s AI infrastructure investment. Professional service providers also use VC subsidies to compete on open-source models at low prices.

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

Moat Issues

Even if a subnet offers valuable services, the underlying models and methods are inherently open: Covenant-72B uses the Apache license, and its technical paper is publicly available. Any competitor can directly replicate without participating in the TAO ecosystem.

Traditional moats (proprietary tech, network effects, switching costs, branding) do not hold:

· Technology is open-source;

· Network effect belongs to TAO, not any single subnet;

· Model weights are identical, and user switching costs are zero.

Community believes that the incentive mechanism itself is a moat, but this relies on continuous large token issuance, which halves periodically, shrinking the incentive budget.

What Is TAO Actually Trading?

At a $2.6 billion market cap, TAO’s price does not reflect demand fundamentals; annual revenue of $3–15 million cannot support such valuation under any traditional framework. The market is trading on: Bitcoin-like scarcity, Grayscale ETF expectations, sector rotation, and long-term options value of decentralized AI. These are all rational speculative factors, but they are entirely supply-side and sentiment-driven.

If you hold TAO based on scarcity and narrative, you might profit even if demand weakens; but if you believe Bittensor will become a truly large-scale AI service network, there is no evidence and significant structural barriers. Investors should clearly distinguish their investment logic.

Original article link

Click to learn about Rhythm BlockBeats’ job openings

Join Rhythm BlockBeats’ official community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Twitter Official Account: https://twitter.com/BlockBeatsAsia