Platform Pulse: The Agentic Foundation and Megascale Intelligence

🤖 The Agentic Enterprise Platform

Google Cloud Next '26 marked a pivotal shift from assistive AI to autonomous agency, introducing a unified platform to build, scale, and govern the next generation of business engines.

1. Gemini Enterprise Agent Platform: A new end-to-end platform that unifies model selection, building, and orchestration with native Vertex AI integration.
2. Agents CLI: From Create to Production: A specialised tool providing coding assistants with machine-readable access to the GCP stack, streamlining scaffolding and deployment.
3. Google's Official Agent Skills Repository: Equip agents with real-time expertise for BigQuery and GKE, reducing context bloat and token waste.
4. Gemini Enterprise for the Agentic Task Force: New features for long-running agents and agentic collaboration spaces to transition AI into an autonomous business engine.
5. Production-Ready AI Agents: Monolith Refactoring: Essential lessons on replacing monolithic scripts with orchestrated sub-agents and structured Pydantic outputs using the ADK.
6. Builders Hub from the Google Developer Program: A centralised workbench offering a unified project dashboard and personalised learning for AI developers.
7. Partner-Built Agents in Agent Gallery: Specialised AI agents from partners like Adobe and Atlassian are now directly accessible within the Gemini Enterprise app.
8. The $750M Agentic Innovation Fund: Google's massive investment to support partners building and deploying enterprise AI agents with forward-deployed engineering.
9. The GEAR Program for AI Learning: Hands-on AI training designed to equip professionals with the skills needed to launch agents at scale.
10. 10 Codelabs for Next '26 Tech: Functional code tutorials to translate the latest AI announcements into production workflows.

⚡ Modern Infrastructure & Megascale Networking

As the demand for AI inference explodes, the physical and virtual fabric of the data centre is being re-engineered for Petabit-scale throughput and carbon-neutral efficiency.

1. Virgo Network: Google’s Megascale AI Fabric: A new scale-out data centre fabric designed to underpin the AI Hypercomputer for the next decade of ML.
2. Inside the 8th Generation TPU Architecture: co-designed system levels featuring the TPU 8t for frontier training and the TPU 8i for large-scale reinforcement learning.
3. Cloud Run updates at Next '26: New support for NVIDIA Blackwell GPUs and deep integration with the Gemini Enterprise Agent Platform.
4. GKE performance, security, and scale enhancements: A suite of updates enabling GKE to serve as the primary infrastructure for agentic applications.
5. Firebase AI Logic: Client-First AI Updates: Server prompt templates, function calling, and hybrid on-device inference for Android applications.
6. Cloud Run Worker Pools (GA): Now generally available for stateful, pull-based agentic workloads, supporting complex back-end logic.
7. Cloud Run Ephemeral Disk: Preview of mountable volumes that persist only for the duration of a specific job or worker instance.
8. Accelerator Network Profile (GA): Automates manual VPC and subnet configuration for high-performance GPU and TPU workloads.
9. Slurm Operator add-on for GKE: Managed scheduling for customised AI and HPC platforms, enabling Slurm on any GKE cluster.
10. Hyperdisk ML: 2 TiB/s Throughput: Now GA for machine series like A3 Ultra and N4, providing the highest throughput in the GCP storage portfolio.

🔐 Identity-First Security & Governance

In the 'Agentic Web,' identity is the only perimeter. The week's updates focus on cryptographic agent attestation, fraud defence, and supply chain integrity.

1. Okta: Blueprint for the Secure Agentic Enterprise: A strategic approach to discovering, registering, and standardising security policies for AI agents across the enterprise.
2. Agent Identity (GA): Provides a strongly attested, cryptographic identity for each agent tied directly to its hosting resource's lifecycle.
3. Google Cloud Fraud Defense: The next evolution of reCAPTCHA, serving as a trust platform specifically for the agentic web.
4. Agent ID Overreach in Entra ID: An analysis of Service Principal takeover risks due to administrator scope overreach in Microsoft Entra ID.
5. Microsoft Entra Passkeys for Windows: Late April rollout of passkeys on Windows to enhance phishing-resistant authentication.
6. Bitwarden npm Supply Chain Attack: A critical reminder of supply chain risks as a popular npm package is targeted in a supply chain heist.
7. Next-gen FinOps: Explainability and Spend Caps: AI cost visibility via the Explainability Agent and absolute control through project-level Spend Caps.
8. Essential AI Security On by Default: Security Command Center Standard now activates baseline AI and cloud security by default for all users.
9. Agent Identity Auth Manager (Preview): Securely authenticate agents to third-party services using 2-legged/3-legged OAuth and API keys.
10. Privileged Access Manager for Agent Identities: New preview support for agents to act as both grant requesters and approvers within the PAM framework.

📊 The Agentic Data Cloud

The shift from isolated data silos to a unified context engine was realised this week through Spanner Omni and the evolution of Dataplex into the Knowledge Catalog.

1. Cloud Spanner Omni: Run Anywhere: A downloadable edition of Spanner allowing organisations to run Google's innovation in their own data centres and edge locations.
2. Google Cloud Knowledge Catalog: The evolution of Dataplex into a dynamic context engine, providing AI agents with secure, high-precision retrieval.
3. What’s new in BigQuery: Powering the Agentic Era: New capabilities for knowledge graphs, built-in AI, and lakehouse interoperability.
4. Oracle AI Database @ Google Cloud: New regions and enhanced AI integrations forming a foundation for the agentic enterprise.
5. Spanner Studio: Visual Graph Schema Management: Simplify graph design by mapping nodes and edges through an intuitive interface instead of manual DDL.
6. Cloud Spanner: Native Schema Design for Cost Savings: Reclaiming up to 50% of storage budgets by eliminating the 'text tax' on binary data.
7. Firestore Enterprise: GA and Text Search: Native mode and pipeline operations reach GA, alongside new previews for Text and Geospatial search.
8. Dataflow Job Builder: External Iceberg REST Catalogs: UI-driven ingestion from external Iceberg catalogs into Google Cloud Lakehouse without manual code.
9. AI.AGG: Semantic Aggregation in BigQuery: New preview function for semantically aggregating unstructured data based on natural language instructions.
10. BigQuery Graph (Preview): Model data as a graph directly from tables to find complex, hidden relationships using GQL.