NVIDIA founder and CEO Jensen Huang delivered a keynote to more than 4,000 attendees at COMPUTEX 2025, presenting a comprehensive vision for AI infrastructure transformation that he positioned as comparable to electricity and internet adoption. Huang declared that "AI is now infrastructure, and this infrastructure, just like the internet, just like electricity, needs factories," describing modern data centres as "AI factories" that "apply energy to it, and it produces something incredibly valuable, and these things are called tokens."
Huang announced a deepened partnership between NVIDIA and Foxconn Hon Hai Technology Group, working with the Taiwan government to build an AI factory supercomputer delivering NVIDIA Blackwell infrastructure to researchers, startups, and industries including TSMC. "Having a world-class AI infrastructure here in Taiwan is really important," Huang stated, emphasising Taiwan's role in the global technology ecosystem.
The company introduced NVLink Fusion architecture enabling hyperscalers to create semi-custom compute solutions with NVIDIA's NVLink interconnect, designed to break down traditional data centre bottlenecks and enable more flexible, optimised system designs for specific AI workloads. Huang described this as making NVIDIA's technology "flexible and open for anybody to integrate into."
NVIDIA announced multiple new enterprise products including DGX Spark personal AI supercomputers for developers, available in "a few weeks" with partners including ASUS, Dell, Gigabyte, Lenovo, and MSI. DGX Station systems provide up to 20 petaflops of performance from a wall socket with capacity to run 1 trillion parameter models, described by Huang as having your "own personal DGX supercomputer."
The company launched NVIDIA RTX PRO Servers for agentic AI as part of the NVIDIA Enterprise AI Factory validated design, now in volume production. These servers provide universal acceleration for AI, design, engineering, and business applications, establishing foundations for partners to build on-premises AI factories.
Huang highlighted industrial applications where companies are building $5 trillion worth of factories worldwide, with Taiwan manufacturers including TSMC, Foxconn, Wistron, Pegatron, Delta Electronics, Quanta, and GIGABYTE using NVIDIA Omniverse to build digital twins for semiconductor and electronics manufacturing optimisation.
The AI factory concept transforms traditional data centereapproaches by focusing on token production as valuable output, while new enterprise server offerings enable organisations to deploy on-premises AI capabilities with validated designs and partner ecosystem support.
NVIDIA's positioning of AI as essential infrastructure comparable to electricity creates market opportunities across multiple trillion-dollar sectors. The company's "one architecture" approach spanning cloud, enterprise, personal, and edge AI enables comprehensive ecosystem development while Taiwan partnerships strengthen supply chain relationships critical for semiconductor manufacturing and AI hardware production.