COMPUTEX 2025: Jensen Huang Lights Up the Stage with AI, Supercomputers, and Taiwan Love”
- ashley19241
- Jun 11
- 4 min read
May 19, 2025 | 21:04 (CST)Compiled by Lin Yi-Hsuan, Economic Daily News New Media Center
NVIDIA CEO Jensen Huang kicked off the 2025 COMPUTEX in Taipei on May 19 with an energetic keynote, opening with a cheerful “大家好” (“Hello, everyone”) in Mandarin. Huang announced NVIDIA's transformation from a technology company into a “global AI infrastructure company,” unveiling a series of cutting-edge products and technologies. In a major reveal, he also announced that NVIDIA's new overseas headquarters will be established in Taipei's Beitou-Shilin Science Park (Beishi Ke). Below is a summary of the key highlights from his speech:
Quick Highlights:
1. NVIDIA's Transformation into a Global AI Infrastructure Company
Huang traced NVIDIA’s evolution—from a chipmaker to a systems company, then a data center company, and now, an AI infrastructure company.“We've realized that NVIDIA is no longer just a tech company—it has become a foundational infrastructure company,” Huang said.
He recounted how in 2016, NVIDIA introduced CUDA and launched its first DGX-1 supercomputer, which he later donated to OpenAI. That marked the beginning of the AI revolution, though, at the time, few understood its significance.
“Today, we are building AI factories—just as critical to modern life as electricity and the internet. In ten years, you'll look back and realize that AI has become essential to every region, industry, country, and company,” Huang stated.
2. "AI for Taiwan"
First AI Supercomputer for Taiwan Huang announced a collaboration with Foxconn, TSMC, and Taiwan's National Science and Technology Council to build AI infrastructure in Taiwan.
“We're not only building supercomputers for the world, we're building AI for Taiwan,” Huang emphasized. NVIDIA will provide its Blackwell architecture and NVLink interconnect technologies, Foxconn will lead the construction, and TSMC will contribute process and application technologies. This marks Taiwan's first large-scale AI supercomputer, forming the backbone of its AI ecosystem.
3. Introduction of NVLink Fusion
Huang unveiled NVLink, likened to NVIDIA’s “spinal cord,”enabling ultra-fast GPU-to-GPU and CPU-to-GPU communications at 7.2 TB/s.
The new NVLink Fusion is a semi-custom system architecture that allows customers’ CPUs and AI accelerators to interconnect with NVIDIA GPUs. Partners include Qualcomm, Al chip, MediaTek, and Marvell.
4. GB300 to Launch in Q3
The Grace Blackwell NVL72 is currently in full production. Huang revealed that the upgraded GB300 system will debut in Q3 2025. It connects 72 Blackwell Ultra GPUs and 36 Grace CPUs based on the Arm Neoverse architecture. The GB300 will deliver 1.5× the inference performance of the GB200—earning the nickname "the thinking machine."
5. “Thank You, Taiwan” — Blackwell Tribute Video Highlights Taiwanese Partners
During his keynote, Jensen Huang repeatedly expressed his gratitude to Taiwan and presented a video showcasing the production process of the Blackwell chip. From the etching of the very first transistor on a silicon wafer to tightening the final screw on the Blackwell server rack, every step reflected the contributions of NVIDIA's partners. Huang praised Blackwell as a technological marvel, with the video concluding on a heartfelt note: “Thank you, Taiwan.”
The video also featured a list of all the Taiwanese companies involved in building the AI chip supply chain, including TSMC, SPIL, KYEC, Foxconn, Delta Electronics, Wistron, Inventec, Auras Technology, Asia Vital Components, and Cooler Master.
6. The Era of Physical AI
Huang described AI's evolution from Perception AI (recognition), to Generative AI, and now into Agentic AI, which can reason and suggest solutions.
The next frontier, he explained, is Physical AI—AI that understands physical laws and interacts with the real world. For instance, if a ball rolls under a car, it doesn’t vanish; it may emerge from the other side. This concept, called object permanence, is critical for AI in autonomous vehicles and robotics.
With Physical AI, robots can navigate factories, manipulate objects, or perform tasks like reaching for a bottle—capabilities that today's AI is beginning to master.
7. NVIDIA’s Overseas HQ to Be Built in Beitou-Shilin (Beishi Ke)
In a surprise announcement at the end of his keynote, Huang revealed that NVIDIA will establish a new overseas headquarters in Taipei's Beitou-Shilin Science Park, named “NVIDIA Constellation.”
8. Supplier Wall: Shout-out to Taiwanese Partners
As in his previous keynotes, Jensen Huang once again showcased a comprehensive list of supply chain partners on the backdrop, highlighting numerous Taiwanese companies. These included: AAEON, Accton, Acer, ADATA, ADLINK, Advantech, Apacer (Innodisk), Aspeed, ASRock, ASUS, AVC, Auras, AVerMedia, Browave, BizLink, Cathay Financial Holdings, Chenbro, Clevo, Compal, CTBC Financial, Delta Electronics, Edom Technology, E.SUN Financial, EverFocus, Flytech (FEC), Foxconn, FSP Group, GIGABYTE, Inventec, Jabil, GMM, King Slide, KYEC, Lanner, Leadtek, Lite-On Technology, JMicron, MediaTek, MiTAC, MSI, Neousys, Nexcom, Pegatron, Phison, Quanta, Realtek, Nan Juen, Solomon, Tyntek, TSMC, UMC, Unimicron, Wistron, Wiwynn, and Tron Future Tech.
NVIDIA CEO Jensen Huang displayed a list of supply chain partners on the backdrop during his keynote speech. Photo by Chien Yung-Hsiang / United Daily News
Source of Information: Economic Daily News
Comments