Should AI development be primarily limited by compute availability?

Context

AI development requires enormous computational resources, raising questions about whether limiting chip availability should be the primary policy tool for controlling AI advancement. Governments and companies debate whether compute restrictions—through export controls, chip manufacturing limits, or licensing—effectively manage AI risks, or whether they simply shift development to other countries while stifling beneficial applications. This tension between safety oversight and innovation access has become central to AI governance as nations compete for technological leadership.