Bittensor TAO Rallies 90% as Covenant-72B Launch Drives Subnet Valuation to $1.5 Billion

Bittensor's native TAO token has rallied approximately 90% over the past month, climbing from $180 to above $332 as of late March 2026. This surge has propelled the valuation of the entire subnet ecosystem to a cumulative $1.5 billion, driven by reflexive staking mechanisms and high-volatility subnet tokens. The rally is underpinned by the successful training of Covenant-72B, a 72-billion-parameter large language model developed on the network's decentralized infrastructure.

Market participants have shifted capital from consolidating Bitcoin to decentralized machine learning initiatives, with the TAO/BTC trading pair jumping nearly 78% in March. Social discussion volume has reached its second-highest recorded level, yet sentiment remains balanced with a positive-to-negative mention ratio of 1.5:1. This data suggests the current price action is supported by fundamental network expansion rather than pure speculative exuberance.

The technical milestone of training a large-scale model without centralized supercomputing clusters has attracted significant attention from industry leaders. Nvidia CEO Jensen Huang and Chamath Palihapitiya endorsed the project on the All-In Podcast, likening the decentralized training process to a modern version of Folding@home.

What specific technical breakthrough validated Bittensor's decentralized AI infrastructure?

The catalyst for the market surge was the March 2026 release of Covenant-72B by the Templar team on Subnet 3 (SN3). This model was trained on 1.1 trillion tokens using over 70 independent peers contributing commodity hardware, achieving an MMLU score of 67.1. This performance places it in a competitive range with Meta's Llama 2 70B, demonstrating that large-scale AI models can be trained permissionlessly.

The technical architecture relies on two core components: SparseLoCo for communication efficiency and Gauntlet for incentive compatibility. SparseLoCo reduces data transfer by compressing gradients with a 146x compression ratio, enabling 94.5% computational utilization on standard internet connections. Gauntlet validates gradient quality by testing loss reduction, ensuring only productive nodes are rewarded within the decentralized network.

This success challenges the traditional political economy of AI by breaking the monopoly of centralized training data centers. However, analysts note that while the technical validation is real, the model is already outdated compared to 2026 state-of-the-art models. The value chain depends on continuous training output rather than direct model usage revenue, creating a unique economic dynamic for subnet participants.

How does the TAO ecosystem mechanism drive subnet token performance?

Bittensor operates on a dynamic mechanism where each subnet has its own automated market maker and token. Valuation is directly tied to the amount of TAO staked in subnet reserves, creating a reflexive cycle where TAO appreciation inflates subnet token prices. As TAO reserves increase, subnet token prices rise, attracting further staking and expanding the ecosystem's total value.

Top performers include τemplar (SN3), which surged 444% in 30 days, and OMEGA Labs (SN24), up 440%. These subnet tokens range from $1 million to $110 million in valuation, offering significantly higher volatility than the parent protocol. The ecosystem plans to expand subnet capacity from 128 to 256 later this year to accommodate growing demand.

Key subnets demonstrate diverse real-world utility, from decentralized compute to blockchain storage and pharmaceutical discovery. Chutes (SN64) is the first subnet to surpass $100 million in value, processing over 9.1 trillion tokens with 400,000+ users. Targon (SN4) projects $10.4 million in annual revenue and has raised $10.5 million in Series A funding.

What are the market risks and investor implications of this rally?

The rally is supported by on-chain metrics showing a 78% rise in the TAO/BTC pair and a surge in staked TAO across subnets to $620 million. March 2026 saw TAO's monthly trading volume reach $5.7 billion, the highest figure recorded in the first quarter. Institutional interest is also rising, with a potential Grayscale TAO ETF decision expected by late 2026.

While the momentum is strong, risks include potential post-rally corrections and network strain from increased activity. Market sentiment remains in the low 30s on the Fear & Greed Index, and retail participation is cautious despite rising prices. Funding rates are only mildly positive, suggesting controlled positioning with limited downside risk from forced liquidations.

Subnet tokens act as leveraged bets on the parent protocol's success, exposing investors to higher volatility. The decentralized AI narrative has become a primary driver as capital moves away from Bitcoin's range-bound action. Traders must weigh the potential for sustained upside against the technical reality that the current model is already outdated.