1.1 The Centralized Status QuoThe current landscape of Artificial Intelligence (AI), particularly concerning Large Language Models (LLMs), is dominated by a centralized paradigm. Industry leaders such as OpenAI, Anthropic, and Google DeepMind deploy and manage their models on massive computational infrastructure within expansive data centers.1 This architecture, where a single, colossal model serves millions of users via cloud-based APIs, has enabled unprecedented scalability, high performance, and the capacity for continuous updates.1 However, this model is not without significant drawbacks. It requires substantial financial investment in infrastructure, is entirely reliant on constant internet connectivity, and introduces considerable privacy and data control concerns, as sensitive information must be processed in the cloud.1 The user also frequently pays for compute capacity they do not fully utilize, leading to an inefficient cost model based on subscriptions or per-token usage.1