Closed and open-source models 'both absolutely necessary,' Nvidia CEO says
Equinix CEO pledges to serve as neutral hub for distributed data and AI
Nvidia CEO Jensen Huang has advised companies adopting AI to put in place the same rigorous safeguards they would apply when bringing on a new employee.
Huang delivered the remarks via video link at the Equinix Horizon event hosted by digital infrastructure company Equinix on Wednesday (local time) at The Midway in San Francisco.
When Equinix CEO Adaire Fox-Martin asked how companies could bridge the gap between AI experimentation and full-scale deployment, Huang opened by saying, "Deploying AI inside a company is like hiring a new employee."
Just as a new hire is not given access to every critical file or internal tool from day one, he said, companies must define clear permissions and boundaries for AI systems from the outset.
Huang said he expects future AI to evolve into a "highly distributed" architecture — one that spans multiple data centers and cloud environments rather than residing in a single location. In such a fragmented network, he said, robust security controls are essential to prevent unauthorized data leakage.
He also weighed in on the growing rivalry between closed and open-source AI models, saying both are "absolutely necessary."
Huang said he personally uses Anthropic, OpenAI, Perplexity, Grok and Muse, adding that companies should draw on as many frontier models as possible while also deploying open-source models in areas where data sovereignty, security or regulatory requirements demand it.
He noted that until recently there had been no frontier-class open-source models, but that the open-source ecosystem now boasts a range of leading models — with more on the way.
Looking ahead, Huang said infrastructure capable of securely connecting distributed AI systems running on fragmented data will be indispensable.
At the same event, Nvidia and Equinix joined forces with open-source AI platform startup Together AI to unveil the Equinix Inference Exchange, an enterprise-grade distributed inference platform.
In her keynote, Fox-Martin said neutral infrastructure is needed to link dispersed data and AI — much like a stock exchange connects scattered investors — and pledged that Equinix would serve as that crossroads.
mokiya@heraldcorp.com