Bittensor Subnets Hit $550M Valuation as Covenant-72B Marks Decentralized AI Milestone
The post Bittensor Subnets Hit $550M Valuation as Covenant-72B Marks Decentralized AI Milestone appeared on BitcoinEthereumNews.com.
TLDR: Bittensor’s top 10 subnets reached a combined $550M valuation, reflecting strong ecosystem growth. Templar SN3 finished Covenant-72B using 72B parameters and 1.1T tokens without any central cluster. TAO token demand rises directly with subnet activity, as purchasing subnet tokens requires TAO first. Grayscale’s ETF filing and Jensen Huang’s comments signal rising institutional interest in TAO’s future. Bittensor subnets have collectively reached a valuation of $550 million, drawing fresh attention to the TAO ecosystem. Templar SN3 completed Covenant-72B, the largest decentralized large language model pre-training in history. The run used 72 billion parameters and 1.1 trillion tokens, with no centralized cluster involved. This milestone has strengthened investor interest in both TAO and the growing subnet economy beneath it. Covenant-72B Sets a New Standard for Decentralized AI Training The Templar SN3 subnet trained Covenant-72B on 72 billion parameters and 1.1 trillion tokens. No centralized computing cluster was used throughout the entire training process. The model’s performance is competitive with Meta’s LLaMA-2-70B in published benchmarks. This places decentralized pre-training on par with established open-source AI infrastructure. Crypto analyst @ElCryptoDoc called the achievement “Bittensor’s DeepSeek moment” in a widely circulated post. NVIDIA CEO Jensen Huang also commented on the development, adding further visibility. The comparison to DeepSeek reflects growing confidence in cost-efficient, distributed training methods. Industry observers have described the run as concrete proof point for decentralized AI. Nobody is talking about the real money in the $TAO ecosystem. The real alpha? The subnets underneath it. Let me break it down: Templar SN3 ( @tplr_ai ) • Just completed Covenant-72B > the largest decentralized LLM pre-training run in history• 72B parameters. 1.1T tokens.… pic.twitter.com/8bIxGujNKQ — CryptoDoc (@ElCryptoDoc) March 22, 2026 Beyond Templar, Targon SN4 stands out as the highest-revenue subnet in the ecosystem. The subnet, operated by Manifold Labs, recently raised a $10.5…
Filed under: News - @ March 22, 2026 6:24 pm