Catch the highlights summary Google Cloud Announced at Next '25 below:

New and Enhanced AI Models (Vertex AI Model Garden)

• Gemini 2.5 Pro – High-performance model for deep reasoning and coding (now in public preview).

• Gemini 2.5 Flash – Low-latency, cost-efficient model (coming soon).

• Imagen 3 – Best text-to-image model with advanced inpainting.

• Chirp 3 – Audio model with Instant Custom Voice (10-second input).

• Lyria – First enterprise-ready text-to-music model.

• Veo 2 – Enhanced video generation with editing and camera control.

• Llama 4 (Meta) – Now available on Vertex AI.

• AI2 Models – Open models from AI2 now in the Model Garden.

Vertex AI is now the only hyperscaler with generative models across video, image, speech, and music.

Tools for Model Management

• Vertex AI Dashboards – Monitor usage, latency, and errors.

• Model Customization – Fine-tune first- and third-party models securely.

• Model Optimizer – Balances quality vs. cost for responses.

• Live API – Enables real-time audio/video input for Gemini.

• Global Endpoint – Ensures responsiveness with regional load balancing.

Multi-Agent Systems Enhancements

• Agent Development Kit (ADK) – Open-source framework to build agent systems using Model Context Protocol (MCP).

• Agent2Agent Protocol (A2A) – Open standard for agent communication across platforms.

• Agent Garden – Pre-built agent samples to speed up development.

• Agent Engine – Fully managed runtime for production deployment.

• Google Maps Grounding – Adds geospatial awareness to agents.

• Customer Engagement Suite – Emotion-aware agents, real-time video understanding, no-code agent builder.

Enterprise AI with Agentspace

• Chrome Enterprise Integration – AI agents embedded in employee workflows.

• Agent Gallery – Central hub for discovering enterprise agents.

• Agent Designer – No-code tool to create task-based agents.

• Idea Generation Agent – Autonomous ideation and evaluation.

• Deep Research Agent – Synthesizes complex topics into reports.

Scientific & Research Applications

• AlphaFold 3 – Protein folding at scale using Google Cloud.

• WeatherNext Models – High-speed, accurate weather forecasting now available in Model Garden.

Next-Gen AI Hardware & Runtime

• TPU v5p “Ironwood” – 7th-gen TPU coming in 2025, designed to efficiently power large inferential AI models.

• Google Distributed Cloud + NVIDIA Blackwell – Gemini can now run locally in connected and air-gapped environments with support from Dell and NVIDIA.

• Pathways on Cloud – Google DeepMind’s distributed AI runtime (used internally at Google) is now available on Google Cloud.

Optimized Inference & Serving

• vLLM on TPU – PyTorch-based models using vLLM can now run on TPUs without code changes, supporting mixed TPU/GPU serving.

• Dynamic Workload Scheduler – Adds support for Trillium, TPU v5e, NVIDIA B200 (A4), and H200 (A3 Ultra) via Flex Start mode; Calendar mode for TPUs coming soon.

Expanded GPU Offerings

• A4 & A4X VMs – Powered by NVIDIA B200 and GB200 Blackwell GPUs; A4X in preview. Google Cloud is the first to offer both.

• NVIDIA Vera Rubin GPUs – Coming soon, enabling up to 15 exaflops of FP4 inference performance per rack.

Accelerator Management at Scale

• Cluster Director – Manages accelerator clusters with colocated VMs and topology-aware scheduling. Upcoming updates:

  • Slurm integration

  • Advanced observability (3600)

  • Job continuity features

  • Join preview for early access

Application Development Enhancements

• Design & Management Tools

  • Application Design Center (preview): Visual, collaborative canvas to design and deploy apps with inline code.

  • Cloud Hub (preview): Central dashboard for app landscape insights, health, optimization, and support.

  • App Hub: Now integrated with 20+ GCP products for app-centric visibility.

  • Application Monitoring (preview): Auto-tags telemetry and enables app-aware alerts and dashboards.

  • Cost Explorer (private preview): Granular visibility into app costs and usage efficiency.

  • Gemini-Powered Developer Tools

• Gemini Code Assist: Helps with dev tasks (migration, testing, docs) and integrates with Android Studio, Atlassian, Sentry, Snyk.

• App Prototyping Agent: Converts app ideas into working prototypes (UI + backend) in Firebase Studio.

• Gemini Cloud Assist: Embedded across design and operations tools (e.g., Observability, FinOps Hub, Security) to enhance infra design, troubleshooting, and cost optimization.

• Developer Programs

• Google Developer Program: Offers AI-assisted tools across all tiers. Premium ($75/month/seat) for enterprise teams is now in preview.

Compute Innovations

• General-Purpose Compute

• C4D & C4 VMs: High-performance VMs powered by AMD (C4D) and Intel (C4), with up to 288 vCPUs, DDR5 memory, Local SSD, and bare metal options.

• Specialized Compute

• H4D VMs (preview): HPC-optimized, delivering record-breaking per-core and memory bandwidth performance.

• M4 VMs: Optimized for SAP HANA with 65% better price-performance over M3.

• Z3 VMs (preview): Storage-optimized, up to 72TB Titanium SSDs with bare metal access.

• Hybrid & Integrated Compute

• Nutanix NC2 on GCP: Hybrid cloud platform now in public preview.

• Google Cloud VMware Engine: Now supports 26 node shapes.

• Infrastructure Upgrades

• Titanium ML Adapter: Enhanced NIC integration for 3.2 Tbps GPU bandwidth.

• Titanium Offload Processors: Connect GPU clusters to Jupiter fabric for scalable compute.

• Managed Instance Groups (MIGs): Now manageable as a unit and support CUDs/reservation sharing with Vertex AI.

AI-Enhanced & Natural Language Capabilities

• AlloyDB AI Upgrades:

  • Natural language querying, vector search via ScaNN, and cross-attention reranking.

  • New AI models: multimodal embeddings (text, images, video), Gemini Embeddings.

  • Agentspace integration enables structured data search with AI agents.

  • The new AI query engine supports natural language in SQL queries.

• MCP Toolbox:

  • Connects AI agents to enterprise databases using Model Context Protocol.

Database Compatibility & Migration

• Firestore + MongoDB:

  • MongoDB API compatibility with Firestore’s global scale and low latency.

• Oracle on Google Cloud:

  • Base Database Service and Exadata X11M (GA), now with CMEK support.

• Database Migration Service (DMS):

  • Now supports SQL Server → PostgreSQL (Cloud SQL & AlloyDB).

Performance, Availability & Management

• Cloud SQL & AlloyDB on Axion (C4A): Uses Arm-based processors for better price-performance.

• Database Center: GA release — unified, AI-powered database fleet management.

• Spanner Updates: Vector search (GA), graph visualization (GA), Repeatable Read isolation (preview), MySQL migration tools.

• Aiven for AlloyDB Omni: Fully managed AlloyDB Omni across AWS, Azure, and GCP (GA).

Bigtable, Memorystore & Firebase

• Bigtable: Continuous materialized views (preview), Cassandra-compatible APIs & migration tools.

• Memorystore for Valkey: Now GA, supports versions 7.2 & 8.0.

• Firebase Data Connect: GA release — combines PostgreSQL’s reliability with GraphQL APIs and type-safe SDKs.

BigQuery Platform Innovations

• Data Pipelines & Preparation:

  • GA: Pipelines, Data preparation, Contribution analysis, Pipe syntax, Semantic search.

• Data Quality & Detection:

  • Preview: Anomaly detection, Knowledge engine powered by Gemini.

• AI-Powered Enhancements:

  • GA/Preview:

    • Natural language & real-world context querying via Gemini.

    • Intelligent SQL cells, scheduled analysis, native viz.

    • AI.GENERATE_TABLE, vector search index (ScaNN), TimesFM forecasting model.

    • Multimodal tables with ObjectRef type.

• Governance & Cataloging:

  • GA: Business glossary, Metastore, Universal catalog, Catalog metadata export, BigLake cataloging, Disaster recovery.

  • Preview: Governance console, Iceberg table support.

• Workload & Cost Management:

  • GA: Advanced workload mgmt, Spend commit model.

  • Preview: Fair slot sharing and better billing attribution.

AI & Natural Language for Analysts & Scientists

• Colab Notebook Integration:

  • Data science agent, intelligent SQL cells, native EDA, and AI code assist in DataFrames.

• LLM & Model Integration:

  • Supports Anthropic Claude, Llama, Mistral in BigQuery ML.

  • TimesFM model, real-time Kafka pipelines, serverless Spark.

• Location & Geospatial:

  • Google Maps datasets + Earth Engine data now in BigQuery.

Looker Platform Updates

• Conversational & AI Tools:

  • Preview: Conversational Analytics (129), API (130), Code Interpreter.

  • GA: Gemini features like formula/visualization assistants, slide generation.

• Reporting Enhancements:

  • New drag-and-drop reports with real-time collaboration.

• DevOps for BI:

  • SQL/LookML testing automation via Spectacles.dev.

Partner Ecosystem & Integrations

• Notable Partner Launches:

  • GrowthLoop: AI-driven marketing journeys.

  • Informatica: Expanded governance/AI capabilities.

  • Fivetran: Managed Data Lake Service with Iceberg/Delta.

  • DBT: Integrated with BigQuery DataFrames & hosted on GCP.

  • Datadog: Expanded BigQuery monitoring.

Firebase Updates

• Firebase Studio (Preview):
  A Gemini-powered, all-in-one cloud development environment for building full-stack AI apps. Includes Gemini Code Assist agents.

• Genkit Framework:
  Open-source AI framework now supports Python and Go. Firebase Studio includes templates to start building with it.

• Vertex AI in Firebase:
  Now supports Gemini Live API for conversational, multimodal app interactions (e.g., audio Q&A).

• Firebase Data Connect (GA):
  Production-ready PostgreSQL backend with GraphQL APIs and type-safe SDKs.

• Firebase App Hosting (GA):
  Git-centric hosting for full-stack web apps.

• App Testing Agent (Preview):
  Automatically generates and runs end-to-end tests in Firebase App Distribution to prep mobile apps for release.

Google Cloud Consulting

• Agentspace Accelerator:
  Prebuilt service to help integrate AI-powered search across organizational data.

• Optimize with TPUs:
  Helps migrate workloads to Google’s Tensor Processing Units for AI performance gains.

• Oracle on Google Cloud:
  Combines Oracle databases with Google Cloud’s AI & infra for improved performance.

• Delivery Navigator (Preview):
  Expanded to customers and partners—offering best practices and proven delivery methods for cloud implementations.

Network Infrastructure & Performance

• Cloud WAN (GA):
  Cross-Cloud Network backbone with 40% better performance and TCO savings.

• 400G Interconnect (Coming Soon):
  4x bandwidth boost for Cloud and Cross-Cloud Interconnects.

• 30,000 GPU Support (Preview):
  Enables massive AI clusters with non-blocking GPU configurations.

• High-Performance RDMA Networking (GA):
  Up to 3.2 Tbps GPU-to-GPU bandwidth with RDMA (186), and Zero-Trust RDMA security coming later in the year.

AI-Optimized Networking & Load Balancing

• LLM Inference Optimizations:
  Cloud Load Balancing now supports AI/LLM workloads across clouds/on-prem.

• Inference Gateway & AI Security Integrations:
  Combine with tools like Model Armor, NeMo Guardrails, and Palo Alto via Service Extensions.

• WebAssembly-based Service Extensions (GA):
  Plugin system now in Cloud Load Balancing; Cloud CDN support coming soon.

Content Delivery & Performance Enhancements

• Cloud CDN Fast Cache Invalidation & TLS 1.3 0-RTT (Preview):
  Faster content delivery and quicker resumed connections.

Service Management & Discovery

• App Hub:
  Streamlined service discovery, with cross-regional failover coming soon.

• Private Service Connect for GKE (Coming 2025):
  Publish multiple services in a single GKE cluster for broader access.

Security Enhancements

• DNS Armor (Preview): Detects advanced DNS-based attacks like tunneling and DGAs.

• Cloud Armor Hierarchical Policies: Granular protection enforcement.

• Cloud NGFW Updates: New network types and firewall tags, plus Layer 7 domain filtering coming 2025.

• Inline Network DLP (Preview): Real-time protection using third-party DLP (Symantec) via Secure Web Proxy and Load Balancer.

• Network Security Integration (GA): Consistent hybrid/multi-cloud security policies without routing changes.

• Imperva App Security: Now integrated with Cloud Load Balancing via Marketplace.

Security Innovations

• Unified Security: A single platform combining threat detection, AI-driven security ops, Mandiant expertise, and more.

• Security Agents:

  • Alert Triage Agent: Automates alert investigation and decision-making.
    Malware Analysis Agent: Uses Code Insight to detect and analyze malicious code.

• Google Security Operations: Enhanced data pipeline management for scale, cost, and compliance.

• Cyber Insurance Expansion: Risk Protection Program now includes Beazley and Chubb.

• Phishing Protection: Chrome Enterprise Premium uses Safe Browsing to defend against credential theft.

• Mandiant Enhancements:

  • Retainer: On-demand access to Mandiant’s services using prepaid credits.

  • Partnerships: New recovery solution with Rubrik and Cohesity for cyberattack resilience.

Storage Innovations

• Hyperdisk:

  • Storage Pools: Now supports up to 5 PiB.

  • Exapools: Industry-leading block storage at exabyte scale.

  • ML: Integrates with GKE for fast data hydration.

• Rapid Storage: Ultra-low latency cloud storage with 6 TB/s throughput and 20x faster data access.

• Anywhere Cache: Reduces latency up to 70% by caching data near GPUs/TPUs.

• Managed Lustre: High-performance file system with <1ms latency and massive throughput for AI.

• Storage Intelligence: Uses LLMs to query metadata and provide actionable insights into data estates.

Startups

• Lightspeed VC Partnership: Lightspeed-backed startups get up to $150K in Google Cloud credits.

• Startup Perks Program: Early-stage startups get benefits from partners like Datadog, GitLab, NVIDIA, and more.

• Vertex AI Credits: Additional $10K credits for using partner foundation models (e.g., Anthropic, Meta).

Google Workspace: AI-Powered Productivity

• Help me Analyze: A feature in Google Sheets that automatically identifies insights from data, helping users make data-driven decisions.

• Docs Audio Overview: Converts Google Docs into human-like audio summaries or podcast-style overviews of documents.

• Workspace Flows: Automates daily tasks like managing approvals, organizing emails, researching customers, and summarizing agendas.

Additional Launches

• Cloud Run Updates:

  • Cloud Run GPUs: Now generally available.

  • High Availability Apps: Public preview of multi-region deployment.

  • Kafka Consumer Worker Pools: Private preview for pull-based workloads.

• SaaS Runtime: A lifecycle management service for building, deploying, and operating AI-infused SaaS services at scale.