The Future NVIDIA is Opening — AI Revolution and the Outlook for Next-Generation Infrastructure

NVIDIA, as a symbol of creativity and technological innovation in the semiconductor industry, continues to pursue new challenges and development. At a keynote held on May 19, 2025, a grand vision of next-generation computing was laid out — from the company's history to future AI infrastructure, data centers, enterprise systems, and the robotics domain.

Opening with the memorable chant "Heat. Heat." and woven through with music and visuals, the presentation highlighted the importance of compute infrastructure and networking in the AI era, and a vision for the emergence of what it calls the "AI factory" as a new industrial structure. NVIDIA founder and CEO Jensen Huang spoke in detail about GPU acceleration via CUDA, the historic birth of the DGX1, and the latest Blackwell technology — including the Grace Blackwell system at its core and the introduction of MVLink Fusion. Behind all of this lies partnerships with countries including the United States and Taiwan, and the presentation contained many insights about the vision being realized by building a global ecosystem.

This article comprehensively covers the content of the keynote as a whole, explaining in detail the latest technologies in each domain, their future outlook, and the overall picture of the AI infrastructure NVIDIA is building. Aimed at modern businesspeople, it details the essence of technological innovation and the insights for corporate strategy — focusing on the future roadmap that commands attention from everyone.

The AI Infrastructure Revolution and the Fusion of Supercomputing — NVIDIA's Challenge and Historical Background Enterprise AI Solutions and the Outlook for Next-Generation Data Infrastructure The Future Robotics and Digital Twins Create — From AI Simulation to Physical Robots Conclusion

The AI Infrastructure Revolution and the Fusion of Supercomputing — NVIDIA's Challenge and Historical Background

The keynote showed, against the backdrop of a decades-long partnership history with Taiwan, how the company is building next-generation computing systems and leading the AI revolution. The close ties with Taiwan spanning more than thirty years are the origin of NVIDIA's growth and technological innovation, and large-scale global expansion centered on that region is expected going forward.

What is noteworthy here is that NVIDIA's history itself has been a continuous "computer revolution." Starting out as a chip company, the introduction of CUDA in 2006 broadly popularized GPU-based computing, expanding influence across all fields from science and technology to entertainment to deep learning.

Furthermore, in 2016, CUDA technology — NVIDIA's early GPU innovation — was spread worldwide, and the company developed new computing systems beginning with the DGX1. The DGX1 shattered the conventional concept of the data center, realizing the paradigm shift of "the data center itself becoming a single computer" — and by being donated to the non-profit OpenAI, it became the catalyst for the AI revolution. This new system, incomprehensible to anyone at the time of its announcement over a decade ago, is now widely recognized as an important technological foundation in fields ranging from generative AI to quantum computing.

One of the important points raised in the keynote was the recognition of two types of network traffic: fast east-west and north-south traffic. North-south traffic handles storage and external connectivity, while east-west traffic is essential for computers to work together simultaneously to solve problems. NVIDIA therefore adopted Mellanox's technology, optimized for east-west traffic, and by acquiring the company approximately five years ago, succeeded in reorganizing the data center into a single massive computing unit. In this way, the evolution of supercomputing contributes not only to improving raw computational power but also to improving the integrated performance of the entire system.

Also part of the technological innovation is the many library groups NVIDIA develops. The introduction of CUDA showed the world the possibilities of GPU-accelerated computation, and by accumulating rich library groups on top of it — CUDNN, Megatron, TensorRT, and more recently announced Dynamo QDF and CU Equivariance — the company has realized high-speed computation across all fields. This has had a major impact in areas including science and technology, weather forecasting, medical image analysis, quantum computing, and optimization of linear search. These innovative tools form the foundation of the new product lineup and the basis on which users and partner companies can freely pursue creative endeavors on NVIDIA's platform.

The most important thing to note here is that NVIDIA as a whole is building "AI infrastructure." Key points summarized below:

Overhauling the conventional CPU-centric computation model and realizing a new era of supercomputing through GPU-centric accelerated computing
Accelerating computation across various fields including deep learning using dedicated libraries, starting with CUDA
Establishing the potential of new systems and the foundation for future AI infrastructure through the announcement of the DGX1
Successfully building data centers capable of supporting enormous parallel processing through optimization of east-west traffic

Furthermore, data centers are now redefined to function as a whole — not like individual PCs or servers, but as a single computer. As a result, a single data center has evolved to the point where it has the potential to generate business opportunities on the order of trillions of dollars as a single "AI factory." Here, the concept of the "AI factory" is depicted as social infrastructure comparable to the electric factories of the 20th century (infrastructure built by GE and Westinghouse), suggesting that AI-based production and services will expand worldwide going forward.

NVIDIA's next-generation product lineup also deserves attention for its ingenuity in overcoming physical limits. For example, the new "Grace Blackwell" system overcomes conventional limits by connecting multiple chips to realize large-scale integrated computing. Further, the company's proprietary CoSL process developed jointly with TSMC and the implementation of ultra-high-speed MVLink at more than 7.22 terabytes per second accumulate technological innovations on the hardware side. These technologies not only pursue computational speed but also contribute to improving overall system efficiency including networking and power management — and can be said to lead to renewal of the IT foundations of enterprises and society as a whole.

As described, NVIDIA continues to challenge the construction of entirely new computer systems, with the fusion of AI and supercomputing as its axis, in a history of technological innovation continuing from the past through the present and into the future.

Enterprise AI Solutions and the Outlook for Next-Generation Data Infrastructure

As demand for enterprise AI solutions grows alongside the rapid advancement of AI technology, NVIDIA is pushing beyond traditional x86-based systems to promote the construction of a new data center as AI infrastructure. The RTX Pro Enterprise and Omniverse servers announced this time begin to prove their practicality across many fields as optimal platforms for companies to leverage AI — realizing the fusion of conventional IT systems and AI technology.

First, conventional IT systems have been operating primarily around existing technologies such as hypervisors and virtual desktops, but handling the load of AI-based generation and inference has become difficult under conventional design philosophy. NVIDIA built a new architecture for this challenge, applying AI optimization not only to the computing layer but also to the storage and network layers. For example, the RTX Pro Enterprise maintains industry-standard x86 compatibility while incorporating a completely new network interface, achieving ultra-high throughput difficult to realize in conventional CPU-centric IT environments through adoption of the CX8 switch enabling data transmission of over 800 Gbps. Such high-performance systems enable far faster and more reliable responses in enterprise business processes than conventional systems, greatly advancing companies' digital transformation.

Furthermore, NVIDIA is also strengthening its work on digital workers, agent systems — "Agentic AI" — as AI solutions. Agentic AI possesses capabilities beyond mere pattern recognition — collecting information, making logical judgments, and operating autonomously in digital space — and holds great potential for improving enterprise operational efficiency and creating new business models. For example, AI substituting for work that previously relied on human hands in areas such as human resource management, customer support, and supply chain management is expected to yield enormous cost savings and acceleration of business processes.

Attention is also being paid to AI-based data platforms. The NVIDIA AI Data Platform was proposed as a new platform that overhauls the concepts of conventional SQL servers and file storage — enabling AI to understand meaning and instantly extract useful information from vast amounts of unstructured data. AI-based semantic search technology, rapid query processing, and sophisticated ranking algorithms are combined here, making it possible for companies to make business decisions at unprecedented speed and accuracy. In the modern big data era, these technologies provide a decisive competitive advantage while dramatically improving overall data utilization efficiency.

Another notable area in enterprise AI is the application of digital twin technology. Techniques for reproducing actual physical environments in virtual space and performing real-time optimization and simulation are being adopted across many fields — manufacturing, urban infrastructure, and semiconductor factory operations. In advanced electronics regions including Taiwan, many partner companies including TSMC, Foxconn, and Pegatron are using this digital twin technology to contribute to factory efficiency and production line optimization. Through this, companies can simulate the design and operation of physically expensive infrastructure in virtual environments, enabling early detection of problems and identification of optimal operational methods.

Furthermore, for expansion of enterprise AI services, coordination with cloud and edge computing is indispensable. NVIDIA is deepening collaboration with cloud service providers, partnering with CoreWeave and Oracle Cloud among others, to provide high-density, high-efficiency AI inference environments. These systems realize high-speed inference performance for AI models by generating large numbers of tokens per second, directly connecting to enterprise real-time decision-making support. And going forward, by combining with high-speed network technology such as CX8, ultra-large-scale enterprise AI systems hold the potential to completely overhaul conventional frameworks.

The Future Robotics and Digital Twins Create — From AI Simulation to Physical Robots

NVIDIA is not content with innovating enterprise foundations; it is also boldly challenging technological innovation in the robotics field. The keynote covered the theme of real-time simulation based on physical laws and application to physical robots, explaining in detail the mechanism by which AI enables robots to perceive their environment, make decisions, and actually take action. Here, the way in which the high costs of human-operated demonstrations and limited experimental environments in the real world that afflicted conventional robot development can be transformed by fusion with the latest AI simulation technology was depicted, with specific technical examples.

NVIDIA uses the simulation engine "Omniverse" and "Isaac Sim" to build a digital twin environment where robots can accurately grasp the physical laws of the real world — inertia, friction, causality, object permanence, etc. This makes it possible to confirm and optimize robot behavior in virtual space before implementing on physical machines. The augmentation of data that AI needs to learn, and high-speed, high-precision processing in simulation environments, contribute to improving robot behavioral stability and versatility. The keynote introduced application examples across a wide range, particularly autonomous vehicles, in-factory transport robots, and home agent robots — depicting a future image where robotics will be active in every industrial setting.

Furthermore, as a new paradigm for robot development, the concept of "Agentic AI" emerges. Agentic AI is depicted not merely as a system that generates output in response to input data, but as a system that autonomously formulates and executes strategies for problem-solving through multiple thought processes. In other words, robots possessing reasoning capabilities, understanding of causality, and the ability to use tools necessary for operating in the physical world — going beyond conventional deep learning and pattern recognition — are being realized. Related to this, the next-generation physical simulation engine "Newton," developed in collaboration with DeepMind and Disney Research, provides highly accurate simulation results surpassing conventional physical simulations, bringing major advances in learning efficiency and robot behavioral optimization.

Additionally, NVIDIA's Isaac platform for groups of robots is an integrated environment consistently supporting robot learning, simulation, and deployment to physical machines. Dedicated processors like the newly released "Jetson Thor," when mounted on robots, realize sensor data processing and real-time inference, supporting high-level autonomy in actual work environments. Such systems will expand not only to automation of industrial sites but also to logistics, medicine, and home service robots — and the future market scale is thought to reach tens of trillions of yen.

Furthermore, NVIDIA is advancing efforts to reproduce entire factories and urban infrastructure in virtual environments through digital twin technology. Digital twins constructed by major companies throughout Taiwan simulate processes from design to implementation and operation, enabling early detection of problems and optimization suggestions. For example, world-leading electronics manufacturers including TSMC, Foxconn, and Pegatron use digital twins to optimize large-scale plants and manufacturing processes, contributing to real-world reconstruction on the scale of trillion-dollar investments. As a result, conventional manufacturing processes are being significantly shortened, yielding results that directly translate to improvements in productivity and quality.

The robotics and digital twin technology revolution continues to occupy an important position in NVIDIA's overall picture. Companies will be able to perform real-time optimization of robot behavior and coordination through digital representations of physical systems — resulting in improved efficiency and safety at production sites. The smart factories and autonomous robotics systems of the future thus have the potential to become the core of digital transformation across entire enterprises, going far beyond mere automation. Moreover, such systems also offer effective solutions to the challenge of conventional labor shortages, with digital workers via Agentic AI expected to greatly contribute to improving operational efficiency for enterprises.

As described, robotics and digital twins are being established as a mechanism that promotes large-scale innovation of existing industries by seamlessly integrating the entire process from AI simulation to actual robot operation. NVIDIA's initiatives affect wide-ranging application fields not as a bridge between technology and the real world, but beyond mere improvements in hardware. A future in which robots operate autonomously in complex environments and the knowledge learned in simulations can be applied directly to physical machines is already close to reality — and much attention will be focused on future market trends.

Conclusion

NVIDIA's keynote this time clearly depicts the path from the traditional CPU-centric computation model to GPU-led accelerated computing and a new era of supercomputing — marking the opening of a new chapter in the AI revolution and showing the overall picture of next-generation infrastructure construction. The keynote explained in detail many topics: CUDA's launch, the donation of the DGX1, optimization of east-west traffic, the innovations of Grace Blackwell and MVLink Fusion, the construction of new IT foundations through the enterprise-oriented DGX series and RTX Pro systems, and the future outlook fusing robotics and digital twin technology.

Through these cutting-edge technologies, companies and research institutions will be able to construct large-scale data centers beyond conventional hardware limits, operate digital workers leveraging Agentic AI, and advance robotics domains represented by autonomous driving and smart factories. The AI factories and new enterprise foundations proposed by NVIDIA hold unique value in future IT environments, simultaneously containing enormous potential to contribute to digital transformation worldwide.

The many technological innovations introduced up to this point are realized through cooperation with various companies and partners, forming a part of the company's global ecosystem. This keynote was not merely an introduction of new products; it was a deeply insightful presentation showing how the entire computing industry will evolve going forward, and how it will impact businesses and the social infrastructure. Through new infrastructure, companies will gain the means to dramatically improve efficiency and productivity — while undoubtedly feeling this as an investment with sufficient value toward the future. The future picture NVIDIA shows is not merely technological innovation but a message of transformation for all industries — and it can be said that the time has come for us to ride this wave of evolution.

Reference: https://www.youtube.com/watch?v=TLzna9__DnI

TIMEWELL's AI Implementation Support

TIMEWELL is a professional team supporting business transformation in the AI agent era.

Services Offered

ZEROCK: A high-security AI agent running on domestic servers
TIMEWELL Base: An AI-native event management platform
WARP: An AI skills development program

In 2026, AI is shifting from "something you use" to "something you work with." Shall we think through your company's AI strategy together?

Book a free consultation →

The Future NVIDIA is Opening — AI Revolution and the Outlook for Next-Generation Infrastructure

The Future NVIDIA is Opening — AI Revolution and the Outlook for Next-Generation Infrastructure