Google Cloud and NVIDIA Strengthen Collaboration to Drive Progress in AI Computing, Software, and Services

NVIDIA’s Cutting-Edge AI Technology, Now Accessible Worldwide Through Google Cloud

A groundbreaking collaboration between Google Cloud and NVIDIA has birthed a new era of AI capabilities for businesses. The two tech giants have unveiled state-of-the-art AI infrastructure and software, empowering customers to harness the immense potential of generative AI models and expedite data science tasks.

During an engaging fireside chat at the Google Cloud Next event, CEO of Google Cloud, Thomas Kurian, and NVIDIA’s founder and CEO, Jensen Huang, elaborated on how their partnership is revolutionizing the landscape of machine learning. This transformative collaboration is democratizing AI supercomputing for some of the world’s largest AI stakeholders. By offering seamless integration of AI supercomputers through Google Cloud, fueled by NVIDIA’s cutting-edge technologies, the partnership is reshaping the possibilities of innovation. Notably, this advancement builds on the successful utilization of NVIDIA technologies by Google’s DeepMind and research teams over the past two years.

Jensen Huang emphasized, “The fusion of accelerated computing and generative AI marks a pivotal moment that accelerates innovation like never before. Our extended collaboration with Google Cloud is set to empower developers with infrastructure, software, and services that amplify energy efficiency and drive down costs.”

Kurian echoed the sentiment, stating, “Google Cloud has consistently pioneered AI innovations to expedite progress for our clients. Many of Google’s cornerstone products rely on NVIDIA GPUs, and numerous customers are actively seeking NVIDIA’s accelerated computing to propel the efficient evolution of Large Language Models (LLMs) for advancing generative AI.”

NVIDIA’s Integration: Catalyst for AI and Data Science Advancement
Google’s renowned framework for developing Large Language Models, PaxML, has undergone optimization for NVIDIA’s accelerated computing. This enhancement equips developers with the ability to leverage NVIDIA’s H100 and A100 Tensor Core GPUs, transforming experimentation and scalability. The optimized PaxML container for GPUs is now instantly available via the NVIDIA NGC™ software catalog. PaxML is designed on JAX, fine-tuned for GPUs through the OpenXLA compiler.

Early adopters of the NVIDIA-powered PaxML include Google DeepMind and other research teams, who have embraced it for exploratory endeavors. The optimized PaxML container will be accessible through the NVIDIA NGC container registry, catering to global researchers, startups, and enterprises shaping the future of AI-driven applications.

Furthermore, a significant stride has been made with Google’s incorporation of serverless Spark, fortified by NVIDIA GPUs, into its Dataproc service. This integration is set to expedite Apache Spark workloads, streamlining data preparation for AI development.

A Legacy of Collaboration and Innovation
The synergy between NVIDIA and Google spans multiple dimensions of hardware and software, driving groundbreaking developments:

• Google Cloud A3 VMs Powered by NVIDIA H100: Google Cloud’s bespoke A3 VMs, backed by NVIDIA H100 GPUs, are slated for imminent general availability. These VMs, set to debut next month, herald a new accessibility era for NVIDIA’s AI platform, delivering a 3x boost in training speed and significantly enhanced networking bandwidth compared to the preceding generation.
• NVIDIA H100 GPUs Empowering Vertex AI: VertexAI, Google Cloud’s platform, is poised to integrate NVIDIA H100 GPUs in the near future. This integration facilitates the swift development of generative AI Large Language Models.
• Google Cloud’s Privileged Access to NVIDIA DGX GH200: Google Cloud secures early access to the NVIDIA DGX GH200 AI supercomputer, armed with the NVIDIA Grace Hopper Superchip. This strategic access is geared toward harnessing the capabilities of generative AI workloads.
• NVIDIA DGX Cloud’s Arrival on Google Cloud: NVIDIA DGX Cloud, a supercomputing and software marvel, will be accessible directly from web browsers, enabling unparalleled speed and scalability for advanced training tasks.
• NVIDIA AI Enterprise on Google Cloud Marketplace: Users can now seamlessly access NVIDIA AI Enterprise through the Google Cloud Marketplace. This cloud-native software platform simplifies the development and deployment of enterprise-ready applications, encompassing generative AI, speech AI, computer vision, and beyond.
• Google Cloud’s Pioneer Role with NVIDIA L4 GPUs: Earlier this year, Google Cloud led the charge by becoming the inaugural cloud provider to introduce NVIDIA L4 Tensor Core GPUs via the G2 VM. The transition from CPUs to L4 GPUs for AI video workloads promises a staggering 120x surge in performance and a remarkable 99% efficiency improvement. These GPUs find extensive use in image and text generation, VDI, and AI-accelerated audio/video transcoding.

In essence, the dynamic partnership between NVIDIA and Google Cloud is unleashing unprecedented AI capabilities, ensuring businesses worldwide can unlock the full potential of generative AI models. This collaboration marks a monumental stride in technology and innovation, setting the stage for transformative advancements across industries.

Source: https://nvidianews.nvidia.com/