At Google Cloud Next, Google announced significant advancements in Google Distributed Cloud (GDC), aimed at integrating AI capabilities like Gemini into various environments without sacrificing data sovereignty. This initiative is poised to foster a new era of sovereign neocloud architecture.
GDC offers two deployment models: GDC air-gapped, which is fully disconnected and utilizes secure, Google-supplied hardware, and GDC connected, which allows for a Google-managed software lifecycle on the user's own hardware. This flexibility addresses the needs of enterprises and governments facing stringent data regulations, enabling them to access cutting-edge AI solutions without the complexities of building their own systems.
Key Features of GDC
GDC provides a comprehensive on-premises AI solution, featuring:
- Managed Infrastructure: Designed to support demanding AI workloads with peak performance across compute, storage, and networking.
- NVIDIA Blackwell GPUs: These GPUs enhance AI performance, leveraging advanced bandwidth capabilities.
- Expanded Machine Families: GDC now includes the A4 machine family, offering a 2.25x increase in peak compute, along with memory-optimized options for specific workloads.
- Enhanced Storage Capacity: GDC has increased object storage per zone from 1PB to 6PB, alongside a tenfold performance boost in IOPS.
Integration of Gemini Models
Organizations can now deploy Google’s Gemini models directly within their environments, maintaining data sovereignty while leveraging advanced generative AI capabilities. The latest Gemini Flash models are available for GDC connected customers, enhancing local processing capabilities.
"Deploying Gemini on Google Distributed Cloud has significantly improved our global manufacturing. Running frontier AI locally allows us to analyze IoT data for real-time predictive maintenance and quality control, avoiding cloud latency." - Junhee Lee, CEO, Samsung SDS
AI Gateway for Enhanced Performance
The introduction of the AI gateway simplifies infrastructure management for sovereign environments. Key features include:
- Dynamic Request Routing: Automatically directs inference requests based on various performance metrics.
- Intelligent Load Balancing: Optimizes GPU utilization for efficient processing.
- Quota Management: Ensures high-priority applications receive necessary resources.
- Observability: Provides tracing and logging for compliance purposes.
Agentic AI Applications
To fully operationalize AI, organizations require secure, autonomous agents. The new agentic AI architecture for GDC allows for the development of powerful AI agents capable of executing tasks within secure boundaries. This architecture is built on Kubernetes, ensuring compliance and security.
Conclusion
Google Distributed Cloud is positioned as a leading platform for deploying AI solutions on-premises, whether connected or air-gapped. These innovations promise to deliver the necessary flexibility and security for organizations navigating the evolving landscape of AI and data sovereignty.
