RTX 5060, Blackwell Superchips & More: Here Are Everything Jensen Huang Said On The First Day Of Computex 2025

The 2025 Taipei International Computer Show (Computex 2025) opened in Taipei, China, on Monday, with NVIDIA CEO Jensen Huang's official keynote speech at the opening ceremony becoming a global focal point.

At 11 a.m. on Monday, local time, Huang took the stage on time and delivered a speech lasting over 100 minutes, unveiling a series of updates from NVIDIA in both hardware and software, while sharing his grand vision for the future of global AI.

For observers in the semiconductor industry, this was not just an unmissable event but also a key indicator for understanding NVIDIA's strategic direction in the coming cycle.

AI Will Be Everywhere

At the beginning of his speech, Huang once again expressed his great expectations for the future of AI: AI will become as essential to our lives as the internet and electricity.

"I know that right now, when we say there's an intelligence infrastructure, it makes no sense. But I promise you, in ten years' time, you will look back and realize that AI has now integrated into everything," Huang said during his speech. "We need AI everywhere. And every region, every industry, every country, every company, all need AI."

When envisioning the prospects of AI, he stated: "$300,000,000 chip industry to a data center opportunity that represents about a trillion dollars, to now an AI factory and AI infrastructure industry that will be measured in trillions of dollars."

He explained NVIDIA's role in this AI infrastructure: "This infrastructure, just like the Internet, just like electricity, needs factories. And these factories are essentially what we build today."

Huang also showcased Oracle's Stargate project during his speech.

In the latter part of his speech, Huang shared his vision for AI agents and humanoid robots. In his vision, the IT departments of global enterprises will become the HR departments for digital AI agents, and humanoid robots could one day become a trillion-dollar industry.

Latest Product Announcements

In the product introduction segment, Huang first unveiled the highly anticipated RTX 5060 GPU and also showcased a new MSI laptop equipped with the RTX 5060. He mentioned that the laptop would hit the market as early as May.

Huang also discussed accelerated computing based on the CUDA-X library, one of the "many sub-projects" under his leadership. He explained that it's not just about GPUs but also includes Earth-2 for weather analysis, Megatron Dynamo for deep learning, and more.

"Very especially, very especially, the algorithms, the libraries, what we call the CUDA x libraries. In fact, we're the only technology company in the world that talks about libraries nonstop. And the reason for that is that libraries are at the core of everything that we do. Libraries are what started it all."

The Grace Blackwell Ultra Superchip was also unveiled on stage. Huang announced that Grace Blackwell is now in full production and began shipping in February. NVIDIA will launch the GB300 hardware system in Q3 this year. According to him, the GB300 features an upgraded Blackwell chip with 1.5x improved inference performance, 1.5x more HBM memory, doubled networking bandwidth, and overall enhanced capabilities.

Additionally, the personal AI computer DGX Spark, which Huang previously mentioned at CES, will be available in the coming weeks. This product is designed for those who want their own supercomputers. He stated that DGX Spark is already in full production and will launch soon, primarily targeting AI researchers.

"One thing for sure, everybody can have one for Christmas."

Licensing NVLink Technology

Huang explained that NVIDIA's collaboration with TSMC has led to a new process called COOS-L, enabling the manufacturing of large-scale chips. Meanwhile, the company has developed the world's fastest switch, NVLink, which currently operates at 7.2 TB/s.

He noted that the entire internet transmits about 900 TB of data per second. However, the NVLink Spine, as NVIDIA's core backbone, can handle massive traffic through nine NVLink switches- a single NVLink Spine transmits more data than the entire internet.

Huang also introduced a new version of NVLink technology: NVLink Fusion.

NVIDIA plans to license this technology to other chip designers to help them build powerful custom AI systems by connecting multiple chips together, accelerating inter-chip data transfers needed for AI tool development and deployment.

Huang stated that NVLink Fusion will help build AI infrastructure rather than just a single, specific superchip. It will integrate all companies' custom CPUs and requirements, seamlessly fitting into the ecosystem.

NVIDIA developed NVLink years ago to exchange massive amounts of data between various chips, such as its GB200, which combines two Blackwell GPUs and a Grace processor.

NVIDIA to Open New Office in Taiwan

Huang also discussed some of NVIDIA's future plans during his speech.

First, NVIDIA will collaborate with Foxconn and TSMC to build an AI supercomputer for Taiwan, called 'AI for Taiwan'.

Additionally, at the end of his speech, Huang introduced NVIDIA's new office in Taipei: "NVIDIA Constellation." This new office will undoubtedly bring NVIDIA closer to key partners like TSMC and other chip manufacturers, drawing cheers from the audience.

Huang explained that the location for the office has already been selected, and NVIDIA has secured the land lease, with construction set to begin soon.

This year's Computex will feature 1,400 exhibitors, marking the first large-scale gathering of computer and chip executives in Asia since former U.S. President Donald Trump threatened comprehensive tariffs.