NVIDIA and Google Enhance AI Ecosystem with Blackwell and Gemini Innovations, Expanding On-Premises and Cloud Capabilities

NVIDIA and Google's partnership, focused on advancing artificial intelligence and empowering developers, has taken a significant leap forward with the introduction of NVIDIA Blackwell and the latest updates to Google’s Gemini models. The collaboration goes beyond mere infrastructure, delving into deep engineering integration to optimize the entire computing stack for AI. Key Innovations and Integration JAX, OpenXLA, MaxText, and llm-d The partnership includes contributions to community-driven software efforts such as JAX, OpenXLA, MaxText, and llm-d. These projects aim to enhance the efficiency and capabilities of AI models, particularly benefiting the deployment and performance of Google's Gemini family of models. JAX, a machine learning framework, is being optimized for seamless scaling and performance gains on Blackwell GPUs, allowing it to handle massive distributed workloads effectively. Performance-Optimized NVIDIA Software NVIDIA has developed several specialized AI software tools like NeMo, TensorRT-LLM, Dynamo, and NIM microservices, which are now tightly integrated with Google Cloud services such as Vertex AI, Google Kubernetes Engine (GKE), and Cloud Run. This integration accelerates performance and simplifies the deployment process, making it easier for organizations to leverage advanced AI capabilities. NVIDIA Blackwell in Production Google Cloud was the pioneer among cloud providers to deploy NVIDIA's HGX B200 and GB200 NVL72 systems with its A4 and A4X virtual machines (VMs). These VMs are accessible through managed services like Vertex AI and GKE, providing a scalable and flexible platform for developing and deploying agentic AI applications. A4 VMs: General Availability A4 VMs, powered by NVIDIA HGX B200, are now generally available, offering robust performance for a wide range of AI tasks. They are part of Google Cloud's AI Hypercomputer architecture, which supports the development of agentic AI solutions at scale. A4X VMs: Exascale Compute A4X VMs, equipped with the NVIDIA GB200 NVL72, deliver over one exaflop of compute power per rack, facilitating seamless scaling to tens of thousands of GPUs. This is made possible through Google Cloud's Jupiter network fabric and advanced networking with NVIDIA ConnectX-7 NICs. The third-generation liquid cooling infrastructure ensures high performance even for the most demanding AI loads. On-Premises Deployment of Gemini While Google's Gemini models have been driving cloud-based agentic AI applications, certain sectors like public, healthcare, and financial services have faced challenges due to strict data residency, regulatory, and security requirements. To address this, NVIDIA Blackwell is now being incorporated into Google Distributed Cloud, a fully managed on-premises, air-gapped environment, and edge solution. This integration allows organizations in regulated industries to deploy Gemini models securely within their own data centers. NVIDIA Blackwell's breakthrough performance and confidential computing features ensure that user prompts and fine-tuning data are protected, enabling customers to innovate while maintaining full control over their information. This expansion of Gemini's reach through Google Distributed Cloud opens up new opportunities for a broader set of organizations to benefit from next-gen agentic AI. Optimizing AI Inference Performance Google's Gemini family of models represents the company's most advanced and versatile AI technology, excelling in complex reasoning, coding, and multimodal understanding. NVIDIA and Google have collaborated to optimize Gemini-based inference workloads, ensuring they run efficiently on NVIDIA GPUs, especially within the Vertex AI platform. This optimization is crucial for handling large volumes of user queries, enhancing the responsiveness and scalability of AI applications. Similarly, the Gemma family of lightweight, open models has been optimized for inference using the NVIDIA TensorRT-LLM library and is expected to be available as easy-to-deploy NVIDIA NIM microservices. These optimizations maximize performance, making advanced AI accessible on diverse deployment architectures, from data centers to local NVIDIA RTX-powered PCs and workstations. Strengthening the Developer Community To further support the developer community, NVIDIA and Google Cloud have launched a joint developer program. This initiative brings together experts and peers to foster cross-skilling and innovation. By combining their engineering expertise, leadership in open-source projects, and a strong developer ecosystem, the two companies are lowering barriers for developers to build, scale, and deploy AI applications. Industry Insight and Company Profiles Industry insiders praise the strategic partnership between NVIDIA and Google for its comprehensive approach to AI innovation. The combination of hardware advancements like NVIDIA Blackwell with Google Cloud's robust infrastructure and managed services is seen as a pivotal step in democratizing AI technology. Both companies are leaders in their respective fields, with NVIDIA known for its cutting-edge GPU technology and Google Cloud for its scalable and secure cloud platforms. This collaboration is expected to significantly drive the adoption of advanced AI solutions across various industries, particularly those with stringent data and security requirements.