Bittensor's subnet SN3 has successfully trained a 72-billion-parameter model, Covenant-72B, using decentralized computing, sparking significant market interest. The model, which matches the performance of Meta's LLaMA-2, was trained without centralized data centers, utilizing over 70 independent nodes. This achievement has drawn comparisons to the distributed computing project Folding@home and has been highlighted by industry figures such as NVIDIA CEO Jensen Huang and Anthropic co-founder Jack Clark. The market has responded sharply, with SN3's value surging over 440% in the past month, reaching a market cap of $130 million. The TAO token, associated with Bittensor, also doubled in value, peaking at $377. This development underscores the growing interest in decentralized AI training, which allows permissionless participation and challenges traditional centralized models. However, the achievement also raises questions about data security and the sustainability of such decentralized approaches in high-compliance industries.