The Hidden Cost of Bloated Code: Small language models, Internet Strain, and the Erosion of Net Neutrality

The rapid expansion of artificial intelligence (AI) has introduced remarkable opportunities in automation, prediction, and problem-solving, yet it also carries a hidden infrastructural burden: the proliferation of bloated code. As AI development increasingly relies on, open-source libraries, iterative reuse, and AI-assisted code generation such as vibe coding, redundant functions, verbose structures, and legacy dependencies accumulate. While these practices accelerate innovation, they also generate excessive data traffic, storage demands, and server strain. This inefficiency is more than a technical inconvenience—it poses a systemic risk to the accessibility and equity of the global internet. This paper examines how inefficient AI coding practices intersect with energy consumption, environmental sustainability, and digital infrastructure. Evidence suggests that code bloat compounds global bandwidth congestion, increases carbon emissions, and intensifies disparities between private and public networks. Wealthy corporations increasingly insulate themselves with proprietary, optimized infrastructures, while public networks—especially in resource-limited contexts—suffer congestion and latency. This accelerates the rise of “private internets,” undermining net neutrality and weakening the internet’s role as a shared democratic space. The paper argues that code efficiency must be treated as digital hygiene: a collective responsibility akin to environmental sustainability. Policy frameworks, developer standards, and regulatory interventions are needed to promote optimization and prevent unchecked inefficiencies. Future research should explore methods to define and measure “code weight,” assess infrastructural impacts, and design governance strategies that protect open, equitable access to the internet in an era of rapidly scaling AI.