On the 19th (local time), the day before the opening of Asia's largest IT and computing exhibition, Computex 2025, in front of the Taipei Music Center in Taiwan, more than 5,000 audience members lined up at the venue to hear the keynote address by Jensen Huang, CEO of Nvidia. The keynote was scheduled to start at 11 a.m., but hundreds of attendees began waiting for seats two hours before the event.
Huang took the stage wearing his trademark leather jacket, accompanied by grand background music. He noted, "My parents are also among the audience here at the Taipei Music Center," and stated, "In the age of artificial intelligence (AI), I will reveal Nvidia's market strategy to provide not just technology, but complete AI infrastructure." In the front row packed with attendees at the Taipei Music Center sat more than ten prominent figures from Taiwan's business community, including the chairpersons of Foxconn and Asus, as well as executives from global partners, including Kwak Noh-jung, president of SK hynix, who watched the keynote.
Huang did not hide his affection for his home country of Taiwan. He remarked, "Taiwan, which I have been visiting for over 30 years, is the hometown of our valuable partner corporations. Together, we are opening up the AI ecosystem," and added, "As I am in Taiwan, I have prepared a surprise announcement. We will establish 'AI infrastructure' for a super-scale supercomputer ecosystem with Foxconn, TSMC, and the National Science and Technology Council (NSTC) of Taiwan."
Huang also announced plans to provide solutions for building customized AI infrastructure. He pointed to the semiconductors integrated into server partners like Asus, Gigabyte, Foxconn, and Pegatron, expressing, "I hope that corporations will build servers specialized for their respective applications," and said, "Nvidia will allow corporations to customize servers with a mixed design asset (IP) of our own solution NVLink based on how they want to integrate the central processing unit (CPU), graphics processing unit (GPU), and application-specific integrated circuits (ASIC)." NVLink is a solution developed by Nvidia that allows GPUs to directly exchange data without going through the CPU to increase data processing speed.
He unveiled a personal supercomputer called the 'DGX Station' aimed at accelerating AI personalization. He stated, "It will be a personal AI device capable of running a model with 1 trillion parameters," adding, "The era is approaching when excellent AI models can be operated individually."
Huang also announced the release of the RTX Server Pro, equipped with Nvidia GPUs, noting that it will support the AI model of China's DeepSeek. He mentioned, "We are unveiling the RTX Server Pro, an AI computing platform for corporations and industries," and stated, "The AI model 'R1' from DeepSeek can also be operated through this product." R1 is said to be the AI model DeepSeek unveiled last January, which recorded performance similar to OpenAI's ChatGPT-4.
He disclosed plans for the widespread introduction of Nvidia's digital twin solutions within Taiwan's manufacturing ecosystem. Huang stated, "Not only semiconductor manufacturing corporations like TSMC and Foxconn, but also corporations that produce servers incorporating these technologies will utilize Nvidia's digital twin in the process of building production facilities."
Huang announced plans to open a new office in Taiwan. He mentioned, "It will be the largest thing developed by Nvidia so far," and remarked, "We will build a new office in Beitou, Taipei, capable of accommodating our growing business. We have already begun discussions with the local market." He added that this new headquarters construction project will be named "Nvidia Constellation."