Bittensor’s ecosystem has lit up markets this month, with its subnet tokens collectively hitting about $1.47 billion in market value as TAO’s rally and high-profile endorsements turbocharged investor interest.
Quick snapshot
- TAO has jumped roughly 90% this month, following a March climb from about $180 to above $332.
- The subnet token category reached a combined market cap of $1.47 billion and $118 million in 24‑hour trading volume, per CoinGecko.
- TAO’s market cap sits near $3 billion, while individual subnet tokens range from roughly $1 million to $137 million.
Subnet tokens lead the charge
While TAO’s run drew attention, the biggest moves were in subnet tokens—the protocol’s specialized tokenized networks. Notable 30‑day gains include:
- Templar (Subnet 3): +444%
- OMEGA Labs: +440%
- Level 114: +280%
- BitQuant: +230%
- Larger subnets also rose: Chutes +54%, Targon +166%
What Bittensor is and why subnets matter
Bittensor is a decentralized AI network that pays contributors in TAO for providing compute, data and model updates. Instead of a single company owning training infrastructure and models, Bittensor spins up purpose-built subnets—each focused on a different AI task (language models, compute, cybersecurity, etc.). There are currently 128 active subnets; each issues its own token whose value is directly tied to TAO staked into that subnet’s reserves.
A key mechanic: since the launch of dynamic TAO in February 2025, each subnet runs its own automated market maker. The more TAO staked into a subnet, the more valuable that subnet’s token becomes. That creates a reflexive, leverage-like relationship: TAO appreciation inflates subnet reserves, which boosts subnet token prices and can attract more staking—amplifying moves in both directions.
Technical credibility: Covenant‑72B
The market moves weren’t just hype. Subnet 3 (Templar) produced Covenant‑72B, a large language model trained permissionlessly across Bittensor by more than 70 contributors using commodity internet hardware. Trained on 1.1 trillion tokens, Covenant‑72B scored 67.1 on MMLU (Massive Multitask Language Understanding), placing it in competitive range with Meta’s Llama 2 70B, according to a March 2026 arXiv paper. That proof of concept helped validate Bittensor’s decentralized training approach.
High-profile endorsements
On March 20, Nvidia CEO Jensen Huang and investor Chamath Palihapitiya discussed Bittensor on the All‑In Podcast, framing decentralized AI training as complementary to proprietary models. Huang’s endorsement—given his outsized influence in AI and markets—resonated beyond crypto circles and likely amplified investor interest.
Institutional signals and what's next
- The protocol plans to grow from 128 to 256 active subnets later this year, paving the way for more token launches.
- A potential regulatory decision to convert the Grayscale TAO Trust into a spot ETF could open institutional access by late 2026.
- Digital Currency Group’s Yuma is already participating across 14 subnets, signaling some institutional or infrastructure-focused interest.
Bottom line
Bittensor’s subnet rally combines technical milestones (Covenant‑72B), tokenomics that create leveraged upside for subnet tokens, and a timely endorsement from influential figures. Whether gains hold will hinge on continued production of competitive, decentralized AI models rather than a single standout result or transient market enthusiasm.
Read more AI-generated news on: undefined/news
