Aaron Levie uses Jensen Huang's case for open-weight AI to explain why model access is becoming an economic and strategic issue. Open models let researchers and companies inspect, adapt and deploy systems on their own terms, while closed frontier providers can still compete through proprietary capability, hosted APIs and enterprise-grade service.
Levie argues that attempts to block Chinese model development are unlikely to stop progress because China has deep technical talent and industrial capacity. The greater risk is an uneven market in which American companies pay premium prices for closed systems while much of the world builds on inexpensive open alternatives. He therefore sees competitive US open-weight models as important infrastructure rather than a rejection of commercial AI.
For enterprise buyers, model choice is only one layer of the stack. Levie expects applied platforms to route work among multiple models according to quality, latency and cost, then combine those models with governed company data, permissions and workflows. Box's own experience with newer models supports that view: stronger reasoning can improve domain-specific knowledge work, but dependable deployment still requires evaluation, safety controls and integration with business systems.
The same dynamic changes software development. Coding agents allow teams to attempt both small quality-of-life improvements and projects that previously looked too expensive or slow, expanding the amount of software worth building rather than simply eliminating engineering roles. Levie's broader conclusion is that falling model costs can shrink margins at the foundation layer while increasing demand for the software and infrastructure that make AI useful inside organizations.
Watch on YouTube
