Google Cloud Platform Newsletter

Check Archive for older issues

News

Make security simpler: Introducing the Google Cloud recommended security checklist - Available today is a new recommended controls checklist to help you set configurations and policies when building a resilient cloud foundation.

Grow your own way: Introducing native support for custom metrics in GKE - Native custom metrics support in the GKE HPA makes it easy to scale workloads on things like active requests, KV Cache or a game server player count.

H4D VMs, now GA, deliver exceptional performance and scaling for HPC workloads - H4D VMs designed for tightly coupled HPC workloads with significant performance gains resulting in faster time to solutions.

Unified Maintenance: A new, unified way to manage maintenance across Google Cloud - Unified Maintenance, now GA, is a centralized dashboard that lets you view and manage maintenance events across your Google Cloud services.

Fueling the autonomous network: Google Cloud and DigitalRoute simplify data readiness for AI - At MWC 2026, we launched reusable data pipelines with DigitalRoute that transform raw, chaotic network noise into high-quality, “AI-ready” data.

Scaling the autonomous network: Introducing the Data Steward and Core Network Agents - In partnership with Future Connections, new Autonomous Data Steward and the Core Network VoLTE Agents bring agentic AI to the entire telecom stack.

The rise of the autonomous network: How GraphML is redefining telecom operations - At MWC 2026, MasOrange showcased a Proof of Concept (PoC) built with NetAI that uses GraphML and digital twins to drive fully managed AIOps.

Announcing the MCP Toolbox Java SDK - Announcing the MCP Toolbox for Databases Java SDK. Build secure, stateful enterprise AI agents with type-safe orchestration for high-concurrency workloads on Spring Boot and AlloyDB.

From framework to scale: Accelerating autonomous networks at MWC 26

Calling all devs: Build the future of Multimodal AI in the Gemini Live Agent Challenge - Stop typing, start interacting! Join the Gemini Live Agent Challenge to break the text box paradigm and build immersive, multimodal AI agents.

Articles, Tutorials

Infrastructure, Networking, Security, Kubernetes

Proactive Preparation and Hardening Against Destructive Attacks: 2026 Edition - Proactive recommendations organizations must prioritize to protect against destructive attacks within an environment.

Look What You Made Us Patch: 2025 Zero-Days in Review - Our analysis of 90 zero-day vulnerabilities tracked in 2025, focusing on techniques and how AI will accelerate the vulnerability landscape.

The AI-native core: Highly resilient telco architecture using Google Kubernetes Engine - Learn how Google Kubernetes Engine (GKE) enables two telco deployment strategies: cloud-centric evolution and strategic hybrid modernization.

Coruna: The Mysterious Journey of a Powerful iOS Exploit Kit - Coruna is a powerful iOS exploit kit leveraging 23 vulnerabilities across multiple threat actors and global campaigns.

Designing private network connectivity for RAG-capable gen AI apps - Learn how do you design a secure private network infrastructure for AI applications with retrieval-augmented generation (RAG).

Decoding GCP Billing with Antigravity and BigQuery MCP - How Google’s agentic IDE and the Model Context Protocol can turn an opaque cloud bill into a granular, custom FinOps CLI in hours.

How to Apply for Google Cloud Invoice Billing - Switch from credit card (online) to invoice billing (offline).

App Development, Serverless, Databases, DevOps

Data Strategy = AI Strategy Series: Transforming Developers into AI Architects with Google Cloud - This article highlights that a successful AI strategy is inherently a data strategy, marking a transition for developers from simply integrating APIs to architecting enterprise-grade AI solutions. It details how Google Cloud, using services like AlloyDB and Cloud SQL, enables this transformation by focusing on speed, scale, and security, making the database a crucial context engine for production-ready applications.

Stop Mocking Cloud Tasks — Use a Real Local Emulator - How to run Google Cloud Tasks locally without mocks or stubs.

Deploy to Cloud Run while skipping the build - This article introduces a new and faster way to deploy applications to Cloud Run by skipping the traditional build process. Developers can provide Cloud Run with a pre-packaged archive of their application and use the `--no-build` flag, leading to dramatically faster deployment times. This method also ensures automatic base image updates as the code is dynamically mounted at runtime.

Simplifying Cloud Run Deployments with Compose - Google Cloud Run now supports direct deployments from your `compose.yaml` file, significantly simplifying the process of moving Docker Compose-defined applications to the cloud. This integration allows you to use the same local configuration for production deployments, automatically handling networking, persistent storage, and secure secret management.

Effortlessly Scaling Cloud Spanner Load Tests to Millions of ops/sec with Cloud Run Jobs - Why your load generator shouldn’t be harder to scale than your database and how to fix it with Cloud Run Jobs.

Modernize your apps with the new Cloud Run universal OS only runtime - Google Cloud Run has introduced a new `osonly24` runtime, enabling faster and more flexible application deployments. This innovation allows developers to directly deploy pre-compiled binaries like Go, Rust, and Dart, bypassing traditional container build processes entirely.

Achieving Exactly-Once File Processing with Google Cloud Functions and Cloud Storage - Google Cloud Functions triggered by Cloud Storage events operate with "at-least-once" delivery, meaning a function may be invoked multiple times for a single file upload, potentially causing duplicate processing. To achieve "exactly-once processing semantics," strategies like idempotent processing and various deduplication methods using event IDs, object generations, or atomic database inserts are crucial.

Your Deleted VMs in GCP Left Behind 500GB of Orphaned Disks (They’re Still on Your Bill) 👻 - This article highlights how deleted Google Cloud VMs often leave behind orphaned persistent disks and unmanaged snapshots, leading to significant, hidden billing costs for unused storage. It details practical steps using `gcloud` commands, Terraform configurations, and automated Cloud Functions to identify, clean up, and prevent these resources, emphasizing potential savings of 15-25% on cloud storage spend.

Testing Google’s Workspace CLI Without the REST API Headache (With Claude Integration) - Installation, Auth, and MCP Commands.

AI Model Transfers on Google Cloud Storage: Parallel Uploads for 15–20x Higher Throughput - Large AI model artifacts frequently encounter slow single-stream uploads to Google Cloud Storage, creating significant bottlenecks. Utilizing parallel composite uploads with `gcloud storage cp` can dramatically boost throughput by 15-20x, efficiently transferring large training checkpoints and model weights.

Google Workspace CLI: Agent Native Alternative to Zapier, n8n and Make - Google has launched `gws`, a new command-line interface that offers programmatic access to all Google Workspace APIs. This tool dynamically generates commands from Google’s live API catalog, ensuring it stays current and greatly simplifying integrations for developers and AI agents.

Big Data, Analytics, ML&AI

Small models, high quality: Inside BMW Group’s experiments evaluating domain-specific language models - Inside BMW Group’s experiments with building and evaluating domain-specific language models for the automotive sector.

Ditching Batch: How to Build a Live Data Pipeline with BigQuery Continuous Queries and Spanner - This article explores how Google Cloud's BigQuery Continuous Queries revolutionize data pipelines by enabling real-time processing of new data directly into Cloud Spanner using simple SQL. This powerful combination allows for immediate analytical insights and high-concurrency transactional reads, effectively moving beyond traditional batch processing for live applications.

Materialised Views - BigQuery Materialized Views offer a powerful way to significantly reduce costs and improve query performance by pre-calculating and storing query results as physical data. Unlike standard views, these materialized views eliminate the need to re-run complex queries against raw data, with BigQuery intelligently handling incremental refreshes and even rerouting base table queries for optimal efficiency. This functionality is crucial for optimizing data processing and dashboards, despite some syntax limitations and storage costs.

Hands-On ELT Pipeline with BigQuery, GCS, and Pub/Sub - From Raw Data to Analytics: Building an ELT Pipeline on GCP.

Escaping Airflow Migration Chaos: Building an AI-Powered Autonomous Migration Agent - The Jira ticket had been sitting in the “Backlog” column for six months, staring at me like a unexploded ordnance.

From network APIs to network AI agents — building the agentic future of connectivity with Nokia - At Mobile World Congress, we announced that we’re integrating Nokia’s Network as Code (NaC) platform with Google Cloud’s optimized agentic AI stack.

The ultimate Nano Banana prompting guide - Learn prompting best practices for Nano Banana Pro and Nano Banana 2, from tech specs to prompting frameworks.

Building a context layer from the ground up - Gorgias built a sophisticated context layer for their internal AI agent, enabling it to accurately answer business questions by generating SQL queries. They unified all data, including unstructured text, within Google Cloud's BigQuery, leveraging its native vector search capabilities. This robust system employs structured metadata and hierarchical instructions to guide the agent in data navigation and understanding, significantly enhancing its reliability and consistency.

Google MCP Servers Tutorial Series - This resource explores Google MCP Remote Servers, hosted implementations of the Model Context Protocol that enable AI agents to securely interact with various Google Cloud services like databases, analytics, and infrastructure management. It provides practical tutorials and a repository of resources for building "Agentic AI" that can perform complex tasks, leverage official documentation, and operate with robust security.

One plugin for endless model choices - This article introduces a Genkit plugin that allows developers to seamlessly integrate and switch between various AI models, including Claude, Mistral, Gemini, and Llama, all through a single provider. Leveraging Google Cloud's Model Garden, this solution eliminates the hassle of managing multiple API keys or tracking quotas from different consoles. It provides a streamlined and secure way to access diverse AI capabilities within a unified environment.

Maximize Your Agent’s Output: Leveraging Multi-Provider Models in Genkit - This article details how to maximize AI agent output by implementing a continuous improvement loop within Google Cloud's Genkit, leveraging multi-provider models. By assigning different AI models roles like "writer" and "critic," their distinct biases and weights work together to significantly refine and enhance the final results.

Beyond the Browser: “Computer Use” and Deep Research for Financial Intelligence - Bridging the gap between AI insights and legacy systems with Gemini Computer Use and Google ADK.

What's new in TensorFlow 2.21 - Google has officially launched LiteRT, the successor to TFLite, which offers significantly faster GPU and NPU acceleration alongside seamless support for PyTorch and JAX. The update also introduces lower-precision data type support for increased efficiency and a commitment to more frequent security and dependency updates across the TensorFlow ecosystem. This transition solidifies LiteRT as Google's primary high-performance framework for deploying GenAI and advanced on-device inference.

Various

You can't stream the energy: A developer's guide to Google Cloud Next '26 in Vegas - While keynotes are available online, Google Cloud Next '26 in Las Vegas offers an irreplaceable in-person experience centered on networking, hands-on problem solving, and the transition to agentic AI.

Slides, Videos, Audio

Security Podcast - #265 Beyond Shadow IT: Unsanctioned AI Agents Don't Just Talk, They Act!

Releases

AlloyDB - Feature: AlloyDB enhanced backups are generally available ( GA ). You can now select the Enhanced tier during cluster creation, manage your project-level backups with tiered tabs, and delete an enhanced backup. For more information, see Manage enhanced backups. Feature: The following AlloyDB AI features are now generally available ( GA ): Auto vector embeddings provide a scalable, automated solution for managing the lifecycle of vector embeddings for large-scale datasets, eliminating the need for manual reindexing or custom scripts. This feature keeps embeddings in sync with transactional data and now supports incremental refresh in manual mode, ensuring that embeddings are only generated for new or updated rows. Additionally, you can perform incremental table refreshes or migration up to 130x faster than traditional row-by-row processing using bulk mode, improving efficiency for semantic search and Retrieval Augmented Generation (RAG). AI functions integrate LLMs like Gemini to bring 'world knowledge' to your AlloyDB data and incorporate advanced semantic search and ranking capabilities directly into your SQL workflows. This feature includes out-of-the-box functions for filtering ( ai.if ), semantic ranking ( ai.rank ), generation ( ai.generate ), and forecasting ( ai.forecast ). Experience higher performance in AlloyDB AI by utilizing array-based AI functions. You can perform batch processing of natural language prompts directly within your SQL queries, significantly improving efficiency for large-scale AI operations. For more information, see Perform intelligent SQL queries using AI functions.

Apigee API Hub - Feature: gcloud CLI support for API hub The gcloud CLI now supports Apigee API hub, allowing you to manage your organization's API catalog, versions, and lifecycle metadata directly from the command line. For more information see gcloud CLI for API hub.

Apigee Integrated Portal - Announcement: On March 5, 2026 we released a new version of the Apigee integrated portal. Feature: You can now publish APIs using AsyncAPI documents to render documentation for asynchronous APIs in your portal. For more information, see Publishing your APIs.

AppEngine Standard - Feature: Support for migrating from the legacy Mail API to an SMTP-based email service, such as SendGrid, Mailgun, or Mailjet is in General Availability.

AppEngine Standard Ruby - Feature: Support for Ruby 4.0 runtime is in Preview.

Cloud Architecture Center - Feature: Migrate on-premises VMs to Google Cloud: Links to resources to help you migrate on-premises VMs from VMware vSphere, Nutanix AHV, and Microsoft Hyper-V to Google Cloud. Feature: (New guide) Classify multimodal data: A high-level architecture to design a multi-agent AI system that analyzes disparate multimodal data and produces a high-confidence classification.

Backup and DR Service - Feature: AlloyDB enhanced backups are now generally available ( GA ). Enhanced backups are managed and stored by Backup and DR Service and provide immutable and indelible backups through backup vault, fine grained scheduling, backup protection against source project deletion and source cluster deletion, PiTR using logs and centralized monitoring and reporting. For more information on protecting AlloyDB clusters using Backup and DR Service, see Protect and recover an AlloyDB cluster to a backup vault. Feature: Backup vault support for Compute Engine instances and Persistent Disk encrypted with customer-managed encryption keys (CMEK) is now generally available.

BigQuery - Feature: You can create a remote model based on the Vertex AI gemini-embedding-001 model, or a remote model based on an open embedding model from Vertex Model Garden or Hugging Face that is deployed to Vertex AI. You can then use the AI.GENERATE_EMBEDDING function with these remote models to generate embeddings. You can also use the AI.EMBED function directly with the gemini-embedding-001 model endpoint. These features are generally available (GA). Feature: You can now use the Pipelines & Connections page to streamline your data integration tasks by using guided, BigQuery-specific configuration workflows for services like BigQuery Data Transfer Service, Datastream, and Pub/Sub. This feature is in Preview. Feature: You can now use an alternate syntax when you call the VECTOR_SEARCH function to improve query performance when you search for a single vector. This feature is in Preview. Change: An updated version of the Simba ODBC driver for BigQuery is now available. Feature: Monitor dataset replication latency and network egress bytes in Cloud Monitoring for BigQuery cross-region replication and managed disaster recovery. These metrics are generally available (GA). Feature: You can now use continuous queries to stream BigQuery data to Spanner in real time. This feature is generally available (GA).

BigTable - Feature: Bigtable tools in the Agent Development Kit (ADK) are generally available (GA). With these tools, you can build AI agents that interact with Bigtable to discover metadata about Bigtable tables and instances and execute LLM-powered SQL queries.

Cloud Build - Fixed: The authorization vulnerability CVE-2026-3136 was fixed.

Chronicle - Change: Google SecOps has updated the list of supported default parsers. For more information, see Supported log types and default parsers on the webpage.

Cloud Composer - Announcement: A new Cloud Composer release has started on March 03, 2026. Get ready for upcoming changes and features as we roll out the new release to all regions. This release is in progress at the moment. Listed changes and features might not be available in some regions yet. Feature: Airflow 3.1.7 is available in Cloud Composer 3. Feature: Highly Resilient environments are available in environments with Airflow 3 (Preview) starting from Airflow version 3.1.7. Feature: Per-folder roles registration is available in environments with Airflow 3 (Preview) starting from Airflow version 3.1.7. Change: Airflow 3.1.0 is no longer included in Cloud Composer images and builds. Change: New Airflow builds are available in Cloud Composer 3: composer-3-airflow-3.1.7-build.0 composer-3-airflow-2.10.5-build.29 (default) composer-3-airflow-2.9.3-build.49 Change: New images are available in Cloud Composer 2: composer-2.16.6-airflow-2.10.5 (default) composer-2.16.6-airflow-2.9.3

Compute Engine - Issue: For Red Hat Enterprise Linux (RHEL) operating system, VM Manager provides vulnerability scanning results based on the latest minor version for each major version released. If your VM runs an earlier minor version of RHEL, you might get inaccurate results in the vulnerability reports. For more information about supported operating systems for vulnerability reports, see supported operating systems. Feature: Generally available: You can use managed constraints with Organization Policy Service for centralized, programmatic control of your Compute Engine resources. Managed constraints replace legacy compute.* constraints and are identifiable by the compute.managed.* prefix. They also include built-in support for safe rollout tools like Policy Simulator and dry run mode. For more information, see Organization policies for Compute Engine and Managed constraints. Feature: Generally available: You can apply compact placement policies to standalone Flex-start VMs. These policies let you colocate your standalone Flex-start VMs as close to each other as possible. Applying a compact placement policy minimizes network hops and improves the performance of latency-sensitive workloads. For more information, see About compact placement policies and About Flex-start VMs.

Cloud Deploy - Feature: Cloud Deploy is now available in the following region: asia-southeast3 (Bangkok)

Document AI - Feature: Custom classifier model pretrained-classifier-v1.5-2025-08-05 is available as General Availability (GA). For more information about available models, see the custom classifier page.

Cloud Firestore - Feature: Firestore Enterprise edition now supports Native mode in all supported regions. For a list of supported regions, see Locations.

Cloud Functions - Feature: Support for Ruby 4.0 runtime is in Preview.

Gemini - Feature: File outline and finish changes features in VS Code general availability The file outline and finish changes features are now available in VS Code and generally available (GA).

IAM - Feature: Service account principal sets are generally available. You can use service account principal sets to reference all service accounts or service agents in a project, folder, or organization when writing allow policies, deny policies, and access policies.

KMS - Feature: Cloud KMS deletion of keys and key versions is generally available. Keys and key versions must meet deletion criteria before they can be deleted. Names of deleted keys can't be reused. For more information, including deletion criteria, see Delete Cloud KMS resources.

GKE new features - Feature: GKE Inference Quickstart (GIQ) now offers recommendations for distributed AI inference. This enables you to deploy optimized, full configurations for advanced models, such as the Qwen and gpt-oss model families, on NVIDIA GPUs and Cloud TPUs. This release introduces GKE Inference Gateway by integrating llm-d inference scheduling. You can select optimized configurations for workloads like Advanced Customer Support, Code Completion, and Deep Research. This tunes your infrastructure to meet the specific latency and throughput requirements of these applications. For more information, see Analyze model serving performance and costs with Inference Quickstart. Feature: You can use automated disk type selection for Hyperdisk volumes on GKE. This feature allows GKE to automatically select the most appropriate disk type based on the machine type of the node where your workload is scheduled. With this feature, you can create a single StorageClass that supports clusters with mixed VM generations. For example, GKE can provision Hyperdisk on compatible instances (such as C3 or C4) while automatically falling back to Persistent Disk on other generations. For more information, see Automated disk type selection. Feature: The H4D machine series, designed for high performance computing (HPC) workloads, is generally available for GKE clusters. Based on 5th generation AMD EPYC Turin with Cloud RDMA 200 Gbps networking, H4D VMs offer 192 cores (SMT disabled), up to 1,488 GB of memory, and 3,750 GiB of Local SSD. H4D is optimized for tightly-coupled applications that scale across multiple nodes and offers RDMA-enabled 200 Gbps networking. You can use H4D with GKE clusters in Standard, or with the Performance compute class in Autopilot. For more information, see Run high performance computing (HPC) workloads with H4D.

Cloud Monitoring - Feature: The SQL queries issued by Observability Analytics can now use a system-defined variable which resolves to the project ID. If a dashboard template uses the project ID variable, then you don't need to update the SQL query after installing the template. For more information, see the following documents: Google Cloud console: Charts with SQL queries API: Charts with SQL queries

NetApp - Feature: Google Cloud NetApp Volumes supports Producer VPC spokes with Network Connectivity Center, a generally available feature. For more information, see Connect additional networks with NCC.

Cloud Interconnect - Feature: Support for 400 Gbps connections is Generally Available for the following Cloud Interconnect types: Dedicated Interconnect Cross-Cloud Interconnect Feature: VLAN attachments with maximum capacities of 400 Gbps are Generally Available for the following Cloud Interconnect types: Dedicated Interconnect Cross-Cloud Interconnect

Network Intelligence Center - Feature: The following features are available in Connectivity Tests: Google Kubernetes Engine (GKE) Pod as an endpoint: you can specify a GKE Pod as the source or destination endpoint for a connectivity test. IP masquerading evaluation: Connectivity Tests evaluates whether IP masquerading is applied to traffic sent from a GKE Pod endpoint. If IP masquerading is applied, the translated address is used for the test. Network policy evaluation: for GKE clusters where FQDN network policies aren't enabled, Connectivity Tests evaluates GKE network policies that apply to a GKE Pod endpoint. For more information, see Considerations for Google Kubernetes Engine.

Cloud Run - Feature: Support for Ruby 4.0 runtime is in Preview.

Security Command Center - Feature: AI Protection is generally available ( GA ) in the Security Command Center Premium tier at the organization level. For regional availability, see Locations for AI Protection.

Cloud Spanner - Feature: You can use continuous queries to stream BigQuery data to Spanner in real time. This feature is generally available (GA). Feature: Spanner supports the ON CONFLICT clause for GoogleSQL INSERT statements to handle unique constraint violations. For more information, see ON CONFLICT DO NOTHING and ON CONFLICT DO UPDATE.

Cloud SQL MySQL - Feature: Cloud SQL for MySQL 8.4.7 is upgraded to MySQL 8.4.8. For more information, see the MySQL 8.4.8 Release Notes.

Cloud Storage - Feature: You can now identify publicly accessible objects ( preview ) by using Storage Insights datasets. The object metadata schema includes fields that provide the public access status of your objects, which lets you understand, organize, and manage publicly accessible objects at scale. Feature: You can now analyze object data and metadata from your Storage Insights datasets by using BigQuery ObjectRef functions. The ref column in the dataset contains a reference to the Cloud Storage object which you can use with ObjectRef functions to analyze object content. For example, you can build queries to help detect sensitive information in documents or to generate image descriptions. For more information, see Analyze object data and metadata using BigQuery.

Cloud Tools for Visual Studio - Deprecated: Effective immediately, Cloud Tools for Visual Studio is deprecated and is removed from distribution. For more information, see Cloud Tools deprecations.