Google Cloud unveils agentic enterprise infrastructure tools

Google Cloud has introduced a set of infrastructure products for what it calls the agentic enterprise, spanning compute, networking, data management and digital sovereignty.

The changes are intended to support AI systems that act with greater autonomy and generate heavier machine-to-machine traffic than conventional applications. Google framed the rollout around four areas: fluid compute, secure cross-cloud connectivity, a unified data layer and sovereign deployment options.

Compute changes

On compute, Google Cloud outlined new virtual machine families, Kubernetes tools and storage updates to help customers run standard enterprise applications and AI-driven workloads on the same infrastructure.

Among the additions is the C4N series, which can process up to 95 million packets per second. Google Cloud also introduced the M4N series with Hyperdisk Extreme, designed for data-heavy workloads including analytics, databases and AI systems.

It also launched GKE Agent Sandbox, a managed isolation service for AI agents on Google Kubernetes Engine. The service can launch up to 300 sandboxes per second per cluster and uses gVisor isolation.

Google argued that CPUs are becoming more important again for certain AI-related tasks, especially when systems must handle branching logic, control flows and isolated execution environments. That sits alongside continued use of GPUs and TPUs for model training and other parallel workloads.

Wayfair was cited as an early customer using Google Cloud infrastructure as part of its broader AI strategy.

"Wayfair's AI strategy is built on years of systematic infrastructure modernization on Google Cloud - migrating our core eCommerce engine and databases off legacy systems, decomposing monolithic services into cloud-native architecture, and unifying our data and analytics platform. That foundation is what makes everything else possible. Today, Gemini Enterprise Agent Platform is powering everything from catalog enrichment to generative shopping experiences that help customers create a home that's just right for them - and it's the same foundation preparing us for the agentic era, where AI doesn't just assist but actively drives discovery, personalization, and commerce across every customer touchpoint and across our business," said Fiona Tan, chief technology officer at Wayfair.

Network controls

In networking, Google Cloud introduced Agent Gateway, a service intended to govern traffic between enterprise AI agents. The tool understands agent protocols including MCP and A2A, and can inspect interactions, enforce access rules, and work with identity and AI safety services.

It also added Cloud Network Insights, which gives customers visibility across hybrid and multi-cloud infrastructure, including Google Cloud, AWS, Azure, data centres and internet applications. The service uses synthetic traffic analysis to trace network paths and identify where performance problems occur.

Cloud Next Generation Firewall and Cloud Armour also received updates intended to detect and block malware and exploits generated or adapted by AI systems, while Model Armour is used to analyse the content and intent of agent communications.

Data layer

In data management, Google Cloud launched Smart Storage and Knowledge Catalog. The products are meant to help customers organise structured and unstructured data so AI agents can search, identify and use information more directly.

According to Google Cloud, Smart Storage adds automated annotation, entity extraction and semantic search to Cloud Storage objects. Knowledge Catalog is intended to map business meaning across data stored in different locations without requiring customers to move that data into a single repository.

These products form part of Google Cloud's Agentic Data Cloud strategy. Vivint was cited as a Smart Storage user in its AI development work.

"AI is critical to making our customers' smart home and security solutions more intelligent and convenient. By leveraging Google Cloud's Smart Storage, we auto-annotate rich metadata delivered in BigQuery. We've scaled and accelerated our data discovery and curation efforts, speeding up our AI development process from months to weeks, continuously delivering innovations that build trust and enhance the overall home experience," said Brandon Bunker, vice president of product, AI, at Vivint.

Sovereign options

Google Cloud also expanded its sovereign and on-premises AI offerings. It introduced Confidential External Key Management, which allows organisations to retain control over encryption keys and the rules governing access to them, while using confidential computing to host the key management endpoint within Google Cloud.

Another addition is Gemini on Google Distributed Cloud, which lets customers run Gemini models in connected or air-gapped environments. The offering is aimed at organisations with strict data sovereignty or security requirements, including those operating in restricted environments.

Google Distributed Cloud is also being updated with NVIDIA Blackwell GPU support, new A4, M2 and M3 virtual machine families, and storage changes that increase capacity per zone by 6x and improve performance by 10x.

Cirrascale Cloud Services, which works on private AI infrastructure, said the changes could help customers seeking dedicated inference environments.

"Our customers demand high-performance, private AI inference without the risks of multi-tenancy. Google Distributed Cloud allows us to provide dedicated, low-latency environments that meet strict sensitive data requirements. With the ability to run Gemini on B200s and B300s, we can significantly increase inference speeds and provide the token throughput our clients need to scale," said Dave Driggers, CEO and co-founder of Cirrascale Cloud Services.