At Google Cloud Next '26, Google Cloud showcased advancements in its Cross-Cloud Network, emphasizing the transition from standalone AI models to multi-agent autonomous systems. This evolution highlights the network's role as a vital integration layer for enterprises navigating the complexities of AI-driven environments.
The Cross-Cloud Network is now utilized by 65% of Fortune 100 companies, managing up to 27 exabytes of data monthly. The latest innovations aim to accelerate AI infrastructure, bolster security, and streamline operations.
Enhanced Networking Infrastructure for AI
As enterprises embrace an agentic world, the network must facilitate large-scale inference and reinforcement learning. Google has refined its networking architecture to enhance workloads across agents, inference, training, and more.
Networking for Agents
The Gemini Enterprise Agent Platform is designed to support the development and management of autonomous agents. Key features announced include:
- Agent Gateway: Acts as an air traffic control system for agentic traffic, enabling secure connections across various protocols.
- Ambient Networking: An integrated data plane for Google Kubernetes Engine (GKE) and Cloud Run that simplifies service discovery and traffic management, reducing operational overhead.
New capabilities under ambient networking include:
- Service Bindings: Automatically establishes service-to-service connectivity.
- Network Services Monitoring: Enhances observability for quicker troubleshooting.
Additionally, Service Extensions will support identity, governance, and AI security for agent communications.
Networking for Inference
Google's GKE Inference Gateway offers optimized inference capabilities, including:
- Multi-Region Support: Enables scaling of inference services across regions.
- Predictive Latency Boost: Enhances request routing based on performance targets.
- Disaggregated Serving: Offers flexibility in model serving.
Networking for Training
Google's network supports the largest AI models globally, with enhancements such as:
- Virgo Network: A non-blocking data center fabric capable of linking up to 134,000 chips.
- Accelerator Network Profiles: Improves bandwidth for distributed AI/ML workloads.
Cross-Cloud Network for Core Applications
The Cross-Cloud Network is designed for secure connectivity across users, data, and applications globally. Recent improvements include:
Ultra Low Latency Solutions
In collaboration with CME Group, Google Cloud has developed an ultra-low latency networking solution tailored for financial exchanges, featuring:
- Deterministic high-performance compute options.
- Scalable multicast data distribution.
- Advanced network observability tools.
Cross-Cloud Observability
New tools like Cloud Network Insights provide performance monitoring and visibility into network infrastructure, enabling proactive optimization.
Enhanced Connectivity Services
Recent updates to foundational networking services include:
- Private Service Connect: Supports secure access to over 40 services.
- Cloud Number Registry: A new IP address management solution.
AI-Powered Security and Operations
The evolving threat landscape necessitates robust security measures. Google Cloud's offerings include:
- Cloud NGFW: Advanced malware protection utilizing AI.
- Cloud Armor: Managed rules for automated threat protection.
AI-driven tools in Gemini Cloud Assist aim to streamline network operations, improve security, and enhance reliability.
Conclusion
Google Cloud's Cross-Cloud Network is built on infrastructure that supports its largest services, ensuring a secure, scalable foundation for enterprises. As organizations adapt to the agentic era, these innovations will facilitate the integration of AI technologies while maintaining operational reliability.
