# Google Cloud and NVIDIA Launch A5X for Advanced AI Infrastructure

> The launch of A5X instances marks a pivotal moment for cloud AI, enabling businesses to harness cutting-edge technology for large-scale reinforcement learning. Early adopters like Ineffable Intelligence are already pushing the boundaries of what’s possible with this advanced infrastructure.

**Source**: blogs.nvidia.com | **Published**: 2026-10-21 | **Type**: article

## Key Facts

- Google Cloud's A5X instances will scale to 960,000 GPUs, enhancing competitive cloud offerings.
- Ineffable Intelligence's use of A5X reveals demand for advanced AI infrastructure in tech innovation.
- WPP's success with A5X shows potential for cost savings by reducing multi-node cluster needs.
- Google Cloud's Confidential Computing enhances data security, addressing vulnerabilities in AI deployments.
- Energy-efficient strategies with NVIDIA DSX MaxLPS indicate a shift towards sustainable cloud solutions.

## Summary

Google Cloud and NVIDIA have announced the general availability of their A5X instances, set to launch next month, marking a significant advancement in cloud-based artificial intelligence infrastructure. This integrated platform is designed for training, inference, and large-scale reinforcement learning, catering to the growing demand for sophisticated AI solutions across various industries. The introduction of A5X instances, featuring NVIDIA's Vera Rubin NVL72 GPUs and advanced liquid-cooling technology, positions Google Cloud as a key player in the competitive landscape of cloud computing and AI.

The A5X instances will leverage NVIDIA’s cutting-edge technology, including ConnectX-9 SuperNICs and Google Cloud’s Virgo Network, to support demanding workloads. Google Cloud aims to scale this infrastructure to accommodate up to 960,000 Rubin GPUs across multiple locations, enhancing its capacity to deliver high-performance computing resources. This expansion is particularly relevant as businesses increasingly seek robust platforms for AI-driven applications.

Ineffable Intelligence, a technology firm founded by David Silver, known for creating AlphaGo, is the first customer to utilize the A5X instances. The company is focused on advancing large-scale reinforcement learning to develop "superlearner" systems that continuously improve through interaction. Silver emphasized the need for substantial computational power and efficient connectivity to facilitate real-time learning and model updates. This reflects a broader trend in AI where organizations are prioritizing infrastructure that can support iterative learning processes at scale.

WPP, a prominent marketing and communications agency, has also begun utilizing the A5X instances to enhance its creative workflows. By employing NVIDIA's Nemotron models and OpenShell architecture, WPP is transforming conversational requests into complex creative tasks, streamlining processes that previously required extensive multi-node clusters. This capability not only demonstrates the potential of A5X instances in creative industries but also highlights the increasing importance of AI in enhancing operational efficiency.

As organizations deploy more sophisticated agentic systems, the need for data protection and model security becomes paramount. Google Cloud is addressing these concerns by offering NVIDIA Confidential Computing on bare metal instances, including A4X Max, which are powered by NVIDIA GB300 NVL72 GPUs. This service aims to provide a secure environment for sensitive data and proprietary models, a critical consideration for enterprises leveraging cloud infrastructure.

In addition to security, Google Cloud and NVIDIA are exploring energy-efficient deployment strategies through NVIDIA’s DSX MaxLPS capabilities. This initiative seeks to optimize performance per watt, allowing companies to expand their computing capacity while adhering to data center power budgets. As environmental sustainability becomes a focal point for technology investments, such innovations will be essential for attracting clients who prioritize energy efficiency.

The launch of A5X instances signals a pivotal moment in the AI and cloud computing markets, as it reflects a shift toward more integrated and powerful solutions. As competitors like Amazon Web Services and Microsoft Azure continue to innovate, the ability to offer comprehensive, high-performance AI infrastructure will be crucial for maintaining market leadership. The advancements in reinforcement learning and agentic systems indicate a growing trend toward AI applications that are not only more capable but also more adaptable to real-world challenges.

Looking ahead, the collaboration between Google Cloud and NVIDIA may set new standards for AI infrastructure, prompting other cloud providers to enhance their offerings. As businesses increasingly rely on AI for critical operations, the demand for secure, efficient, and scalable solutions will likely drive further innovations in this space. This development could lead to a more competitive landscape where companies must continuously evolve their technology to meet the sophisticated needs of their clients.

## Entities

- **Companies**: Google Cloud, NVIDIA, Ineffable Intelligence, WPP, Vertebrae, Polaris Cloud, Confidential Core AI
- **Products**: A5X instances, NVIDIA Vera Rubin NVL72, NVIDIA ConnectX-9 SuperNICs, Google Cloud Virgo Network, NVIDIA Nemotron models, NVIDIA OpenShell, NVIDIA GB300 NVL72, NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs
- **Technologies**: reinforcement learning, AI Hypercomputer, Confidential Computing, energy-efficient deployment
- **People**: David Silver

## Key Concepts

general availability of A5X instances, large-scale reinforcement learning, superlearner systems, agentic systems, Confidential Computing, creative workflows, energy-efficient deployment, data-center power budgets

## Definitions

- **A5X instances**: A5X instances are integrated platforms designed for demanding inference and reinforcement learning workloads, combining advanced NVIDIA hardware and Google Cloud technologies.
- **reinforcement learning**: Reinforcement learning is a type of machine learning where agents learn to make decisions by interacting with their environment and receiving feedback.
- **Confidential Computing**: Confidential Computing is a technology that protects data in use by isolating it in a secure environment, ensuring privacy and security during processing.
- **superlearner systems**: Superlearner systems are advanced AI models that continuously improve their performance by learning from their own experiences.
- **energy-efficient deployment**: Energy-efficient deployment refers to strategies and technologies that optimize power consumption while maintaining performance in data centers.

## Use Cases

- Advancing large-scale reinforcement learning
- Creating multistep creative workflows
- Ensuring conversational privacy
- Verifiable agent execution
- Policy-governed enterprise intelligence
- Optimizing performance per watt in data centers

## Frequently Asked Questions

**What are A5X instances used for?**

A5X instances are designed for demanding workloads in reinforcement learning and inference, providing an integrated platform that combines compute, networking, and storage.

**Who is the first customer of A5X instances?**

Ineffable Intelligence, founded by David Silver, was the first customer to access A5X instances, utilizing them to enhance their large-scale reinforcement learning efforts.

**What is Confidential Computing?**

Confidential Computing is a technology that safeguards data while it is being processed, ensuring that sensitive information remains secure even in cloud environments.

**How does WPP utilize A5X instances?**

WPP has demonstrated its creative agent reference architecture on A5X instances, enabling the conversion of conversational requests into complex creative workflows with human oversight.

**What benefits does Google Cloud offer with its infrastructure?**

Google Cloud provides a scalable infrastructure that supports advanced AI applications, including energy-efficient deployment options and robust security features through Confidential Computing.

## Links

- [Read on Welcome.AI](https://welcome.ai/content/google-cloud-and-nvidia-launch-a5x-for-advanced-ai-infrastructure)
- [Original source](https://blogs.nvidia.com/blog/google-cloud-vera-rubin-infrastructure/)
- [NVIDIA](https://welcome.ai/company/nvidia): Featured company

---

Source: Welcome.AI | https://welcome.ai/content/google-cloud-and-nvidia-launch-a5x-for-advanced-ai-infrastructure