Electrical energy. The Web. Now it’s time for an additional main expertise, AI, to brush the globe.
NVIDIA founder and CEO Jensen Huang took the stage at a packed Taipei Music Middle Monday to kick off COMPUTEX 2025, charming the viewers of greater than 4,000 with a imaginative and prescient for a expertise revolution that can sweep each nation, each trade and each firm.
“AI is now infrastructure, and this infrastructure, identical to the web, identical to electrical energy, wants factories,” Huang mentioned. “These factories are basically what we construct at the moment.”
“They’re not information facilities of the previous,” Huang added. “These AI information facilities, if you’ll, are improperly described. They’re, in reality, AI factories. You apply power to it, and it produces one thing extremely precious, and this stuff are referred to as tokens.”
NVIDIA CUDA-X In all places: After displaying a towering wall of companion logos, Huang described how firms are utilizing NVIDIA’s CUDA-X platform for a dizzying array of purposes, how NVIDIA and its companions are constructing 6G utilizing AI, and revealed NVIDIA’s newest work to speed up quantum supercomputing.

“The bigger the set up base, the extra builders need to create libraries, the extra libraries, the extra wonderful issues are carried out,” Huang mentioned, describing CUDA-X’s rising recognition and energy. “Higher purposes, extra advantages to customers.”
Extra’s coming, Huang mentioned, describing the rising energy of AI to purpose and understand. That leads us to agentic AI — AI in a position to perceive, assume and act. Past that’s bodily AI — AI that understands the world. The part after that, he mentioned, is basic robotics.
All of this has created demand for rather more computing energy. To satisfy these wants, Huang detailed the most recent NVIDIA improvements from Grace Blackwell NVL72 techniques to superior networking expertise, and detailed big new AI installations from CoreWeave, Oracle, Microsoft, xAI and others throughout the globe.
“These are gigantic manufacturing facility investments, and the explanation why folks construct factories is as a result of you already know, you already know the reply,” Huang mentioned with a smile. “The extra you purchase, the extra you make.”
Constructing AI for Taiwan: All of it begins in Taiwan, Huang mentioned, highlighting the important thing position Taiwan performs within the international expertise ecosystem. However Taiwan isn’t simply constructing AI for the world; NVIDIA helps construct AI for Taiwan. Huang introduced that NVIDIA and Foxconn Hon Hai Know-how Group are deepening their longstanding partnership and are working with the Taiwan authorities to construct an AI manufacturing facility supercomputer that can ship state-of-the-art NVIDIA Blackwell infrastructure to researchers, startups and industries – together with TSMC.
“Having a world-class AI infrastructure right here in Taiwan is de facto necessary,” Huang mentioned.
NVIDIA NVLink Fusion: And transferring to assist its companions scale up their techniques nevertheless they select, Huang introduced NVLink Fusion, a brand new structure that allows hyperscalers to create semi-custom compute options with NVIDIA’s NVLink interconnect.
This expertise goals to interrupt down conventional information middle bottlenecks, enabling a brand new degree of AI scale and extra versatile, optimized system designs tailor-made to particular AI workloads.
“This unbelievable physique of labor now turns into versatile and open for anyone to combine into,” Huang mentioned.
Blackwell In all places: And the engine now powering this complete AI ecosystem is NVIDIA Blackwell, with Huang displaying a slide explaining how NVIDIA affords “one structure,” from cloud AI to enterprise AI, from private AI to edge AI.
- DGX Spark: Now in full manufacturing, this private AI supercomputer for builders will probably be accessible in a “few weeks.” DGX Spark companions embrace ASUS, Dell, Gigabyte, Lenovo and MSI.
- DGX Station: DGX Station is a strong system with as much as 20 petaflops of efficiency powered from a wall socket. Huang mentioned it has the capability to run a 1 trillion parameter mannequin, which is like having your “personal private DGX supercomputer.”
- NVIDIA RTX PRO Servers: Huang additionally introduced a brand new line of enterprise servers for agentic AI. NVIDIA RTX PRO Servers, a part of a brand new NVIDIA Enterprise AI Manufacturing facility validated design, at the moment are in quantity manufacturing. Delivering common acceleration for AI, design, engineering and enterprise, RTX PRO Servers present a basis for NVIDIA companions to construct and function on-premises AI factories.
- NVIDIA AI Information Platform: The compute platform is totally different, so the storage platform for contemporary AI is totally different. To that finish, Huang showcased the most recent NVIDIA companions constructing clever storage infrastructure with NVIDIA RTX 6000 PRO Blackwell Server Version GPUs and the NVIDIA AI Information Platform reference design.
Bodily AI: Brokers are “basically digital robots,” Huang mentioned, in a position to “understand, perceive and plan.” To hurry up the event of bodily robots, the trade wants to coach robots in a simulated atmosphere. Huang mentioned that NVIDIA partnered with DeepMind and Disney to construct Newton, the world’s most superior physics coaching engine for robotics.
- Huang launched new instruments to hurry the event of humanoid robots: The Isaac GR00T-Desires blueprint will assist generate artificial coaching information. And the Isaac GR00T N1.5 Humanoid Robotic Basis Mannequin will energy robotic intelligence.
- Industrial Bodily AI: Huang mentioned that firms are within the strategy of constructing $5 trillion value of factories worldwide. Optimizing the design of these factories is crucial to boosting their output. Taiwan’s main producers — TSMC, Foxconn, Wistron, Pegatron, Delta Electronics, Quanta, GIGABYTE and others — are harnessing NVIDIA Omniverse to construct digital twins to drive the following wave of commercial bodily AI for semiconductor and electronics manufacturing.
NVIDIA Constellation: Lastly, constructing anticipation, Huang launched a dramatic video displaying NVIDIA’s Santa Clara workplace launching into area and touchdown in Taiwan. The large reveal: NVIDIA Constellation, a model new Taiwan workplace for NVIDIA’s rising Taiwan workforce.
In closing, Huang emphasised that the work Taiwanese firms are doing has modified the world. He thanked NVIDIA’s ecosystem companions and described the trade’s alternative as “extraordinary” and “as soon as in a lifetime.”
“We’re in reality creating an entire new trade to assist AI factories, AI brokers, and robotics, with one structure,” Huang mentioned.