Cloud Covered: What was new in Google Cloud in February

AI Takes Center Stage with Major Advancements

February was a significant month for Google Cloud's AI initiatives, marked by a strong presence at the AI Impact Summit in India and the release of several powerful AI tools and upgrades. The summit underscored Google's commitment to leveraging AI for real-world problem-solving, featuring new partnerships and investments aimed at advancing science, education, and scalable AI solutions. Key highlights include the release of Nano Banana 2, which merges high-quality image generation with impressive speed, making advanced visual creation more accessible to developers and users across products like the Gemini app and Google Search. Enhancements to SynthID also continue to aid in the identification of AI-generated content.

Beyond image generation, Google unveiled Lyria 3, its most advanced music creation tool, enabling users to generate custom music within the Gemini app with descriptive prompts and even custom cover art. The integration of ProducerAI into Google Labs further expands creative possibilities in music production. For developers and creative professionals, Flow now incorporates top AI capabilities for generating, editing, and animating images and videos within a single, streamlined workspace, simplifying asset management and creation workflows.

Empowering Scientific Discovery and Performance Analysis

Google Cloud's AI prowess extended to complex scientific endeavors with the release of an upgraded Gemini 3 Deep Think. This enhanced version is specifically engineered to tackle the intricate data challenges faced in science and engineering, moving beyond theoretical concepts to deliver practical, actionable results for technical problems. Available to Google AI Ultra subscribers in the Gemini app and via the Gemini API for early access, Deep Think promises to accelerate research and development. In the realm of sports, Google Cloud's AI video analysis tool, developed with Google DeepMind, provided Team USA and U.S. Ski & Snowboard athletes with a competitive edge. This tool analyzes athlete movements from 2D video, even with bulky gear, offering near real-time feedback to optimize performance for the Olympic Winter Games.

Enhanced Developer Tools and Infrastructure

February also brought significant updates for developers and infrastructure management. The Public Preview of Datastream’s metadata integration with Knowledge Catalog was announced, offering a unified view of Datastream assets like Streams and Connection Profiles. This integration streamlines discoverability and governance by allowing users to search Datastream assets alongside BigQuery tables. For those modernizing their infrastructure, a guide was released on upgrading Apigee OPDK to v4.53 with OS modernization, detailing a stable, zero-downtime transition. Furthermore, the General Availability of Cloud Run worker pools and the open-sourcing of the Cloud Run External Metrics Autoscaler (CREMA) empower serverless AI at scale by enabling efficient, queue-aware autoscaling for non-HTTP workloads.

Streamlined API Management and Data Handling

Google Cloud is simplifying API governance with the General Availability of native OpenAPI v3 (OASv3) support for API Gateway and Cloud Endpoints. This eliminates the need to downgrade modern API specifications and allows for direct definition and enforcement of policies, including telemetry, quotas, and security, using Google-specific extensions within OASv3 files. This ensures better security by design and seamless integration with the broader developer ecosystem. In data management, the release notes highlight the general availability of the Cloud Logging API MCP server, enabling AI applications to interact with log entries. Additionally, Database Center support for the Model Context Protocol (MCP) is now generally available, allowing AI applications like Gemini CLI to connect to Database Center for fleet health reviews, inventory audits, and security checks.

Improvements in Compute and Storage

For compute resources, February saw the preview of A4X VMs powered by NVIDIA GB200 NVL72, offering high performance and efficiency for next-generation AI reasoning models, positioning Google Cloud as a leader in providing advanced AI hardware. On the storage front, Google Kubernetes Engine (GKE) introduced Dynamic Default Storage Classes. This feature simplifies storage management across mixed-generation VM clusters by automatically selecting the appropriate storage (Persistent Disk or Hyperdisk) based on node hardware compatibility. This abstraction reduces operational overhead and ensures optimal performance and cost-efficiency without complex scheduling rules.

User Experience and Security Enhancements

Enhancing the user experience within the Google Cloud console, Dark Mode is now generally available. This feature provides a modern, comfortable, and productive environment, especially for extended use or in low-light conditions, with options for automatic or manual activation. Security and identity also saw advancements, with the preview of quantum-safe digital signatures in Cloud Key Management Service (Cloud KMS). This aligns with NIST's new PQC standards, helping developers implement quantum-resistant cryptography to protect against future quantum computing threats. The ongoing support for legacy Google Security Operations SIEM infrastructure will end on April 30, 2027, encouraging migration to Google Cloud for improved reliability and security.

The Continued Evolution of Generative AI Models

The evolution of generative AI models continued with advancements rolling out in preview. Gemini 3.1 Flash-Lite became available to enterprises via Vertex AI and developers via the Gemini API, offering enhanced image generation capabilities. The 3.1 Pro model, described as noticeably smarter and more capable for complex problem-solving, also began its rollout in preview via Vertex AI and Gemini Enterprise, empowering businesses for the agentic future. Furthermore, BigQuery ML now supports open-source generative AI models from Vertex AI Model Garden, including those from Hugging Face, expanding model choices for SQL-based tasks like sentiment analysis and text generation. This broad integration of cutting-edge AI across various Google Cloud services solidifies its position as a leader in intelligent cloud solutions.