Nvidia Executives: AI is the Core Driver of the Future, H200 is a Powerful Booster for Generative AI

[baoshi.rao]

Nvidia is a global leader in AI computing platforms and GPU manufacturing, continuously launching innovative AI products and solutions in recent years to provide strong support for AI applications across various industries.

Recently, at the GTC 2023 conference, NVIDIA released a graphics processing unit (GPU) specifically designed for training and deploying generative AI models—the H200. This superchip, based on the NVIDIA Hopper architecture, features advanced memory and processing capabilities, enabling it to handle massive datasets for generative AI and high-performance computing workloads.

This new product reveals NVIDIA's focus and strategic layout in the AI business, while also demonstrating the company's core competitive advantages in the AI domain.

Generative AI is an AI method that utilizes deep learning techniques to generate new data or content, such as images, videos, audio, and text. It has a wide range of applications, including virtual reality, digital entertainment, medical imaging, and drug design. However, generative AI also faces significant challenges, such as model complexity, data scale, and computational requirements. To address these challenges, NVIDIA has launched the H200 GPU, a GPU specifically designed for generative AI, with the following features:

The H200 is the first GPU equipped with HBM3e memory, providing 141GB of memory and 4.8TB/s memory bandwidth - nearly double the capacity of NVIDIA H100 Tensor Core GPUs and a 1.4x increase in memory bandwidth. This means the H200 can handle larger datasets and more complex models, improving the efficiency and quality of generative AI.

The H200, built on NVIDIA's Hopper architecture, employs multiple innovative technologies such as the Transformer Engine, MIMD execution units, and multi-layer cache systems to accelerate various generative AI tasks including image synthesis, video generation, speech synthesis, and text generation. The H200 also supports multiple precision formats (FP8, FP16, FP32, FP64) allowing flexible adjustments according to different application requirements, thereby improving both the performance and accuracy of generative AI.

H200 supports multiple programming models and frameworks, such as CUDA, TensorRT, PyTorch, and TensorFlow, enabling easy integration with NVIDIA's AI software ecosystem, including NVIDIA DGX, NVIDIA Omniverse, and NVIDIA Jarvis. It provides a one-stop solution for the development and deployment of generative AI.

The H200 GPU is NVIDIA's significant breakthrough in the field of generative AI. It will provide a powerful accelerated computing platform for generative AI research and applications, opening a new era for generative AI.

NVIDIA's founder and CEO Jensen Huang stated at GTC 2023 that AI is the core driving force of the future, which will transform our lives, work, and entertainment. He believes generative AI represents the new frontier of AI, creating infinite possibilities—such as generating virtual worlds, virtual characters, and virtual objects—to provide humans with more choices and experiences. He also emphasized that NVIDIA will continue to invest in AI research, development, and innovation, delivering cutting-edge technologies and solutions to advance AI.

At the GTC 2023 conference, NVIDIA's Chief Scientist Bill Dally shared the company's latest research achievements in the AI field. For instance, they trained a trillion-parameter language model using H200 GPUs, currently the world's largest language model capable of generating high-quality text content such as novels, news articles, and poetry. He also demonstrated how NVIDIA's AI models utilize generative AI technology to reconstruct 3D scenes from 2D images or generate 2D images from 3D scenes, achieving bidirectional conversion and editing between images and videos.

NVIDIA's new products and research indicate that the company has made significant progress in the AI sector. NVIDIA's AI technologies and solutions now cover a wide range of industries and fields, such as healthcare, education, entertainment, retail, manufacturing, and finance, providing robust support for various AI applications.

NVIDIA's new products and research have also brought positive impacts to its commercialization and productization. For example, they have enhanced NVIDIA's brand influence and market share, increased its revenue and profits, expanded its customer and partner base, and strengthened its innovation capabilities and competitive advantages.

NVIDIA has not only heavily invested in its own AI products and research but has also made extensive investments in the global AI ecosystem, supporting numerous AI startups and projects to foster AI development and innovation. According to NVIDIA's public disclosures, the company has invested in over 30 AI startups, covering fields closely related to AI such as autonomous driving, data processing, image recognition, voice interaction, and software. Some of these startups have become unicorns in the AI field, such as Cohere, Inflection AI, and Recursion Pharmaceuticals, while others have become key partners for NVIDIA, including Nuro, DeepMap, and Momenta.

NVIDIA's investment logic is based on three aspects: First, the company's vision aligns with NVIDIA's, which is to use AI technology to create more value for society; Second, the company requires NVIDIA's assistance, such as technical support, market promotion, capital injection, etc.; Third, the company itself must be an outstanding one, with a strong team, technology, products, and market. NVIDIA's investment goal is to help these startups achieve rapid growth and development, while also bringing strategic benefits to NVIDIA itself, such as expanding NVIDIA's influence and collaboration network, increasing market opportunities and revenue sources, and acquiring new technologies and knowledge.

NVIDIA has announced that it will continue to increase its investments in the AI field, seeking out more promising AI startups and projects to provide greater support and resources for AI development and innovation. NVIDIA's next strategic direction will focus on generative AI and the metaverse, representing the new frontiers and trends in AI, as well as NVIDIA's new vision. Generative AI can create limitless virtual content and experiences, while the metaverse is a shared digital world composed of these virtual elements, bridging the gap between human reality and virtuality to offer more choices and possibilities. NVIDIA believes that generative AI and the metaverse will be the core drivers of the future, transforming how we live, work, and entertain ourselves, while also presenting NVIDIA with more business opportunities and competitive advantages.

NVIDIA's AI strategy is a long-term strategic investment that embodies the company's vision and innovation, showcasing its strength and confidence. This strategy has not only generated substantial returns for NVIDIA but has also made significant contributions to the advancement and innovation of AI, delivering greater value and well-being to society and humanity. NVIDIA's AI strategy is a winner-takes-all game, positioning the company as a leader and victor in the AI field.