Open-weight artificial intelligence (AI) models are gaining traction as businesses look for cheaper and more flexible alternatives to proprietary AI systems, but Bank of America says their lower upfront costs do not necessarily translate into lower demand for computing infrastructure. In a recent analysis, BofA analyst Vivek Arya argued that open-weight models can actually expand AI usage because running and deploying these models still requires significant computing power. Open-weight models accounted for about 71% of tokens on OpenRouter in 2026 year-to-date, up from roughly 27% in 2025, while cloud capital spending is expected to nearly double to about $1 trillion this year.
For NVIDIA (NVDA), that dynamic could be particularly important. Rather than reducing demand for AI chips, BofA believes open-weight models could shift spending away from model providers and toward the semiconductor and infrastructure companies supplying the compute needed to run them.