NVIDIA Chief Executive Officer Jensen Huang has made a strong case for embracing open-weight artificial intelligence models, arguing that the United States’ future leadership in AI will depend not on a handful of proprietary systems but on creating an open and competitive ecosystem that allows innovation to flourish across industries.
In a paper titled Open Weights and American AI Leadership, Huang draws parallels between today’s AI revolution and the rise of open-source software in the 1980s.
He argues that just as open-source software became the backbone of the internet and modern computing, open-weight AI models can democratise access to advanced artificial intelligence while strengthening America’s technological sovereignty.
According to Huang, the next phase of AI leadership will not be measured solely by who develops the most powerful frontier model. Instead, success will depend on whether advanced AI becomes widely available to startups, universities, public institutions and businesses of every size.
He says open-weight models—AI systems whose model weights can be downloaded, inspected, modified and deployed on an organisation’s own infrastructure reduce barriers to entry by eliminating the need for every company to build expensive frontier models from scratch.
This, he argues, will lower costs while allowing organisations to tailor AI solutions to their own needs.
Huang believes broader access to AI will accelerate adoption across sectors including manufacturing, healthcare, agriculture, education and small businesses.
Rather than concentrating AI capabilities in a few technology giants, open-weight models would enable thousands of companies to build specialised applications, creating stronger competition and distributing economic gains more widely.
Competition, he notes, extends beyond AI model developers to cloud providers, semiconductor manufacturers, software applications and digital services. Such rivalry encourages innovation, reduces costs and gives customers greater flexibility in choosing technologies that best suit their operations.
Another major advantage, Huang argues, is customer control. Organisations investing heavily in AI increasingly want assurance that they will not become dependent on a single technology vendor.
Open-weight models allow businesses to retain ownership of their data, customise AI systems and preserve institutional knowledge instead of surrendering strategic capabilities to external providers.
While acknowledging that open-weight AI introduces genuine security risks because models can be modified after release, Huang rejects calls for broad restrictions.
Instead, he argues that openness can actually improve cybersecurity by enabling researchers, developers and security experts to identify vulnerabilities, strengthen safeguards and respond more effectively to emerging threats.
He warns that concentrating advanced AI capabilities within a small number of closed systems creates single points of failure while reducing transparency and competition. By contrast, open models allow broader independent testing, benchmarking and evaluation, resulting in more robust and secure AI over time.
To sustain American leadership, Huang urges policymakers to expand computing resources for startups and researchers, invest in shared AI infrastructure such as datasets and evaluation frameworks, and avoid regulations that unnecessarily restrict open-weight development.
He also distinguishes legitimate AI development techniques, such as model distillation, from unlawful attempts to extract value from proprietary systems, arguing that targeted legal frameworks—not sweeping restrictions—are the appropriate response.
Huang concludes that AI presents an unprecedented opportunity for economic prosperity if governments choose policies that encourage openness, competition and innovation.
By supporting open-weight AI, he argues, the United States can expand access to advanced technology, strengthen its innovation ecosystem and ensure that the benefits of artificial intelligence are shared broadly across society rather than concentrated among a few dominant players.