NVIDIA founder and CEO Jensen Huang addressed a crowd of over 6,500 industry leaders, press, entrepreneurs, and technology enthusiasts ahead of the COMPUTEX technology conference in Taipei. Sporting his signature black leather jacket, Huang declared that accelerated computing and AI are ushering in a new revolution in the computer industry, with NVIDIA at the forefront.

Huang highlighted the vital role of Taiwan as home to NVIDIA’s most trusted partners, emphasizing that everything NVIDIA does originates from this collaborative ecosystem. Together, they’ve built the foundation of the world’s AI infrastructure. His speech focused on three core themes: the nature and impact of generative AI, a roadmap for future development, and a glimpse into what lies ahead.

Accelerated Computing: The Sustainable and Inevitable Path Forward

Huang reminded the audience that the computer industry, now nearly 60 years old, has experienced only a handful of truly transformative technological shifts. We are at the precipice of such a shift today. While CPU performance growth has plateaued, our demand for processing power continues to surge exponentially, creating a phenomenon he termed “computing inflation.”

The answer, he argued, lies in accelerated computing, a field NVIDIA has pioneered for the past two decades. NVIDIA’s CUDA platform empowers CPUs by accelerating specific workloads best suited for specialized processors. Huang envisions a future where every application and data center will be accelerated, offering a more sustainable and efficient computing landscape.

This vision of accelerated computing is not merely theoretical but a practical and economically viable solution. Huang presented compelling data showcasing the remarkable speed, energy efficiency, and cost savings achievable with this approach. Accelerated computing can deliver up to a 100x speedup while only increasing power consumption threefold and costs by a mere 50%, resulting in a 25x performance-per-watt advantage compared to CPUs alone.

The Triumph of CUDA and the Rise of Generative AI

However, achieving accelerated computing is a complex endeavor, demanding a complete overhaul of software and algorithmic restructuring. Over the past 20 years, NVIDIA has diligently crafted domain-specific libraries to address this challenge. These include cuDNN for deep learning, a dedicated AI physics library, Ariel for 5G radio, cuLitho for chip manufacturing, Parabricks for gene sequencing, cuOpt for combinatorial optimization, and cuQuantum for simulating quantum computers.

These powerful libraries have empowered over 5 million developers to leverage NVIDIA’s platform, unlocking the benefits of accelerated computing across a diverse range of industries, including healthcare, finance, computing, and automotive. Recently, Google reinforced this trend by integrating NVIDIA’s cuDF library into its cloud platform to accelerate the popular data science library Pandas.

The widespread adoption of CUDA and accelerated computing has fueled a “virtuous cycle” within the industry. As the user base expands, costs decline, inspiring more developers to innovate and further fueling demand for accelerated solutions.

AI Factories and Digital Twins: A New Industrial Revolution

Huang emphasized that AI’s capabilities have expanded beyond perception tasks like image processing and speech recognition. AI can now generate, evidenced by the remarkable success of ChatGPT. This paradigm shift was made possible by the dramatic reduction in computing costs, making it both practical and economically feasible to train large language models on the entire breadth of internet data.

Leveraging AI’s generative power, NVIDIA is constructing “AI factories” that produce AI “tokens,” just as Nikola Tesla’s AC generator produced electrons. These tokens represent anything of value, from words, images, graphs, sounds, and videos to even complex entities like chemicals or proteins.

Huang unveiled NVIDIA’s ambitious Earth-2 project, which aims to create a digital twin of our planet. This digital replica will enable us to predict the Earth’s future, aiding efforts to mitigate disasters and adapt to climate change. Weather tokens generated through Earth-2 will eventually allow for continuous weather prediction for every location on Earth.

NVIDIA ACE: Bringing AI Assistants to Life

Huang also introduced NVIDIA ACE, a groundbreaking digital human technology. ACE aims to create AI-powered digital characters capable of interacting with users in a human-like manner, revolutionizing fields like customer service, advertising, and gaming.

To simplify the complex software infrastructure required to power these digital humans, NVIDIA introduced a new software platform called NIM (NVIDIA Inference Microservices). NIM packages pre-trained AI models into optimized containers, ready for deployment on clouds, data centers, or workstations. This flexible approach empowers users to select and assemble expert AI models to create customized applications.

The Robotics Revolution: Physical AI Powered by Omniverse

Huang confidently declared that the next wave of AI will be “physical AI,” AI that can understand the laws of physics and seamlessly collaborate with humans in the real world.

To realize this vision, NVIDIA developed the Omniverse platform, a simulated environment where robots can learn and refine their skills while adhering to the laws of physics. “Omniverse is where robots learn how to be robots,” Huang stated. Within this virtual world, robots can master tasks like precise object grasping, obstacle avoidance, and complex manipulation.

Blackwell: The Foundation for the Future of AI

Huang unveiled NVIDIA’s latest GPU architecture, Blackwell, designed to deliver the performance and scalability essential for training and running increasingly large and complex AI models. Blackwell boasts massive chips that can be interconnected to form a colossal 72-GPU system capable of housing large language models with 10 trillion parameters, offering an unprecedented level of computational power.

NVIDIA’s One-Year Rhythm: A Commitment to Innovation

Huang reiterated NVIDIA’s unwavering commitment to pushing technological boundaries. The company operates on a one-year rhythm, relentlessly delivering new platforms and technologies annually. This steadfast dedication underscores NVIDIA’s determination to lead the AI revolution and empower industries with cutting-edge solutions.

Huang’s address served as a powerful testament to the transformative power of accelerated computing and AI. NVIDIA stands at the forefront of this revolution, driving the development of intelligent technologies that will shape our future.