Jen-Hsun Huang's Computex Speech: The "Generative AI" Revolution Across All Industries in the Next Decade, Making Token Production Humanity's Primary Job.

Speaking at Computex 2025, Jensen Huang predicted that within a decade, the entire industry will be disrupted by generative and analog AI, and the key to productivity will be transformed into the generation of discussion tokens, which will also become the new face of human future work and life. (Synopsis: Jensen Huang: NVIDIA Taiwan headquarters is located in Beishike!) Shipai, Tianmu shouted "Huida housing area" ready to rise) (Background supplement: Huawei shouts to surpass Huida, Ascend 910D "Performance Dick H100" Fastest May Shipping Sample) In his keynote speech at Computex2025, Nvidia founder and CEO Jensen Huang (19) once again drew a blueprint for the global technology industry for an AI-driven future, and declared that Nvidia has begun to shift from a technology platform to a more important existence, the next step in global reform. He introduced the latest evolution of AI from perception and generation to agent and physical intelligence, and predicted that a new era of AI deep empowerment is coming. The whole industry was revolutionized by AI Huang Jenxun first mentioned in his speech that Nvidia is no longer a simple technology company, but a human infrastructure company, and in the future, AI will use the word "intelligence" as a necessity of human life, which may be like sunlight, air, water and other necessities of life, he also gave examples of "electricity" that appeared after the industrial revolution has become an important infrastructure, in the Internet era, information has become a human life need, and the next step is "intelligence". Huang predicts that within ten years, all industries will cover all kinds of AI perception and generative forms, completely changing the way everyone works, and he also demonstrated through Omninverse how eight partner manufacturers such as Wistron, GIGABYTE, and TSMC can use physical AI and simulation platforms to carry out "digital twins", that is, fully automated optimization of factories and increase factory productivity. All companies will soon start talking about how many tokens to produce, how to produce tokens, how much to produce per hour, and every factory will do that. Huang believes that all industries need AI data centers, but in fact, these data centers actually produce tokens, which is a very scarce thing, because every industry needs AI agents, and the focus of future company operations will start to discuss how many tokens are produced per hour, just like factory optimization And to achieve this, it depends on the operation of the NVIDIA Cuda graphics card that generates the "token", and it can be said that all services will rely on "" The generation of "tokens", and the hardware, software, and ecology responsible for "tokens" will become the next focus of all industries, not including all other value-added benefits of the AI industry, and the "AI factory" alone will be a multi-trillion-dollar scale of infrastructure. Technical Demonstration: CUDA, GeForce and Omniverse Jensen Huang emphasized that all the foundations are related to accelerated computing, AI and CUDA X libraries, and Nvidia's continuous cultivation of core technologies, based on the CUDA accelerated computing platform, Nvidia provides powerful computing power support for scientific research, industrial manufacturing and content creation through its huge CUDA X library ecosystem. For example, CuLitho technology is used to accelerate chip design. On the accelerated computing side, the GPU is far more powerful than the CPU, and the global CPU computing center can not be compared with the GPU, and in the face of ASIC and other related competition, Huang Jenxun said that the GPU can achieve speed beyond the ASIC level through horizontal series, and in addition to the expensive Nvlink, the new network chip has also become a competitively priced horizontal connection, which can realize the deployment of eight graphics cards at the individual family and small and medium-sized enterprises. On the consumer side, GeForce RTX GPUs continue to lead the innovation of gaming and creative experiences. Jensen Huang demonstrated how the latest RTX 50 series GPUs can achieve smooth real-time ray tracing at very high image quality through AI-driven DLSS (Deep Learning Ultra High Sampling) technology, proving that AI has become an indispensable part of graphics computing. Through the newly announced RTX PRO 6000 platform, eight graphics cards can be connected through network chips with a bandwidth of up to 800Gbps per second, and its cost benefits can be introduced to industrial AI. The Nvidia Omniverse platform shines in the digital twin space. Huang introduced its built-in advanced physics engine, Newton, which accurately simulates complex physical interactions, enabling companies to create realistic digital twins for factories, production lines, and even entire cities to optimize design, simulate operations, and accelerate the training of AI robots. New products and technological breakthroughs The focus of the presentation was undoubtedly Nvidia's new generation of Blackwell architecture and its flagship Grace Blackwell superchip (such as GB200/GB300). Huang pointed out that Blackwell has achieved unprecedented computing scale and efficiency through revolutionary MVLink high-speed interconnection technology and advanced liquid cooling system, and will become the core engine driving the next generation of AI factories. In response to different market needs, Jensen Huang also released a series of innovative products: DGX Spark and the new DGX Station: desktop and personal AI supercomputers for AI developers, researchers and professionals, putting cutting-edge AI computing power at your fingertips. Nvidia AI Data Platform and AIQ Query System: Provides high-performance AI-driven storage and semantic search solutions for enterprise unstructured data. Robotics Platform: The new Isaac Groot platform, Jetson Thor robot processor, and Nvidia Cosmos and Groot Dreams models for large-scale synthetic data generation are designed to accelerate the development and deployment of all types of robots. MVLink Fusion Open Platform: Allows partners to integrate their own ASICs or CPUs into Nvidia's MVLink ecosystem to jointly create a customized, high-performance AI infrastructure. Robots and digital twins: in-depth cooperation with Taiwan factories Finally, Huang Jenxun also showed the case of robot linguistic and eight Taiwanese manufacturers' "digital twins", robots can already recognize human language and make corresponding actions, which means that robots will become full-featured production tools in the future, and the slides claim that in the future, robot systems based on AI language will be able to replace the labor of thousands of factories. Foxconn, Wistron, Pegatron, Quanta, Wiwing, Delta, GIGABYTE, TSMC and other manufacturers have simulated various physical simulation scenarios such as factory pipelines, thermal control, and reducing production scrap rates at Omniverse. When each digital twin is connected to a real-world machine, real-time interactive dashboards will appear, and even monitoring can be realized on the production line, discovering various causes that may cause yield degradation, so all production will be optimized according to Omniverse and digital twins: GeForce brought AI into the world, AI is back, and now transform, AI will come back and change everything you've been through, and I'm excited to witness this with you...