On May 19, 2025, NVIDIA CEO Jensen Huang once again captivated audiences at COMPUTEX in Taipei. This keynote wasn’t just another corporate presentation—it was a powerful vision for the future of technology, filled with exciting product announcements, strategic collaborations, and heartfelt gratitude toward Taiwan, his home country and a vital partner in NVIDIA’s success story.
From Chipmaker to AI Architect: NVIDIA’s Evolution
At the start of his speech, Jensen made it clear: NVIDIA is no longer just a GPU chipmaker—it has become an “architect of intelligent infrastructure.” AI, according to Huang, is now emerging as the third major infrastructure shift after electricity and the internet.
Central to his speech was the concept of “AI Factories.” These aren’t merely data centers; instead, they convert electricity into AI outputs measured in a new unit called “tokens.” Just as watts quantify electricity, Jensen sees tokens quantifying AI production. These factories will democratize AI, making it accessible for industries worldwide, and accelerating digital transformation across every sector.
Launching Groundbreaking Products: Blackwell Architecture and DGX Spark
COMPUTEX is traditionally a stage for exciting product debuts, and this year NVIDIA did not disappoint. Jensen officially unveiled the company’s next-gen GPU architecture: Blackwell. Blackwell promises unprecedented AI computing power and superior energy efficiency, paving the way for a whole new generation of NVIDIA’s AI-driven solutions.
One standout product demonstration was the “DGX Spark,” a desktop-sized supercomputer designed specifically for AI developers. Though small enough to sit comfortably on a desk, the DGX Spark packs immense computing power, ideal for use in homes, offices, or research labs, giving developers instant access to leading-edge AI capabilities.
Jensen also introduced the NVLink Fusion technology, enabling enterprises to build highly customizable AI infrastructures by flexibly combining GPUs. This promises significant improvements in performance, scalability, and efficiency, empowering companies to tailor their AI setups precisely to their needs.
Major Investment in Taiwan: A New AI Supercomputer and Headquarters
Jensen, who maintains deep personal and professional ties with Taiwan, announced an ambitious investment in collaboration with TSMC, Foxconn, and the Taiwanese government. Together, they’ll construct a massive AI supercomputer equipped with 10,000 Blackwell GPUs, an investment totaling hundreds of millions of dollars. This new supercomputer positions Taiwan as a central AI hub in Asia, further cementing its pivotal role in global technology infrastructure.
Additionally, NVIDIA will establish its new Taiwanese headquarters, named “Constellation,” located in the Shilin-Beitou Science Park. More than just a corporate office, Constellation will serve as a hub for AI R&D, talent cultivation, and educational outreach. Jensen emotionally emphasized, “Without Taiwan, NVIDIA could not achieve what it has today,” expressing his profound gratitude and respect for Taiwan’s contributions.
AI’s Three Stages of Evolution: Perception, Generation, and Physical AI
Jensen elaborated on AI’s ongoing evolution, identifying three key developmental stages:
- Perception AI: Systems capable of recognizing images, speech, and patterns, as seen in applications like voice assistants and facial recognition.
- Generative AI: Recently popularized by tools such as ChatGPT and Midjourney, these AIs generate content, including text, images, music, and code, dramatically enhancing productivity across creative and technical fields.
- Physical AI: This future stage integrates perception and generation with real-world physical interactions, exemplified by autonomous vehicles, robotics, and automated factories. Jensen highlighted NVIDIA’s Cosmos platform, a powerful simulation environment enabling AI models to train extensively in virtual worlds, significantly improving their performance in real-world scenarios.
Promoting Global Collaboration: Praise for China’s Open-Source AI Model
Reflecting a global perspective, Jensen openly praised the recent launch of China’s DeepSeek R1, an open-source large language model. He called this development a “gift to the global AI community,” emphasizing the importance of openness and collaboration in advancing AI technologies.
Despite acknowledging that recent U.S. semiconductor restrictions on China have somewhat impacted NVIDIA’s business, Jensen optimistically noted that these challenges have spurred Chinese innovation, leading to a healthier, more diverse global AI ecosystem.
Gratitude to Taiwan: An Emotional Connection
In an emotional conclusion, Jensen played a video displaying a map of Taiwan surrounded by logos of NVIDIA’s local partners, concluding with a simple yet powerful message: “Thank You, Taiwan.” He reinforced Taiwan’s critical role in NVIDIA’s global operations—from chip manufacturing and hardware innovation to talent recruitment—stating clearly, “Without Taiwan’s support, NVIDIA’s vision could never have been realized.”
This heartfelt statement garnered enthusiastic applause, highlighting Jensen’s authentic appreciation and the unique bond NVIDIA shares with Taiwan, extending far beyond business considerations.
Closing Remarks: The Future of AI is Ours to Build Together
Jensen Huang’s keynote was more than just a tech showcase; it was a call to action. He stressed that the future of AI cannot be built by a single company alone—it requires collaborative effort across global industries, communities, and talent pools.
“The AI-powered future demands everyone’s participation,” Jensen declared, underscoring his optimism and openness toward a future built on shared innovation and global partnership.
