260 things we announced at Google Cloud Next '26 – a recap

extract

260 things we announced at Google Cloud Next '26 – a recap Sarah Kennedy Vice President, Marketing, Google Cloud Peder Ulander Vice President, Product Marketing, Google Cloud Google Cloud Next ‘26 took place this week in Las Vegas, and the energy was incredible as we welcomed over 32,000 leaders, developers, and partners to explore the Agentic Era with us. Across three keynotes, 25 spotlights, 700+ breakout sessions, 490+ sponsorships, our Showcase & Expo tech playground, and innumerable impromptu networking and collaboration moments, you could really feel the excitement as organizations move from pilot to production at a scale the world has never seen. During our opening keynote and throughout the week, we shared our blueprint for the agentic enterprise and all the latest innovations for businesses to meet this moment. To name just a few: - - Gemini Enterprise Agent Platform, our new, comprehensive platform to build, scale, govern, and optimize agents - Gemini Enterprise app brings the best of Google AI to every employee, for every workflow and includes new features that let every employee build with AI - Google Cloud’s AI Infrastructure offers the industry’s widest variety of compute options and new innovations at every layer of the AI Hypercomputer, including our new 8th generation TPUs - Agentic Defense is delivered via a cybersecurity platform that combines Google’s Threat Intelligence and Security Operations with Wiz’s Cloud and AI Security Platform to detect, prevent, and respond to threats - Agentic Data Cloud introduces a new, AI-native architecture that allows your data to be utilized at the speed and scale required by agentic AI - Agentic Taskforce redefines engagement and productivity with new innovations across our Gemini Enterprise for Customer Experience and Workspace products - And so much more - That was just the tip of the iceberg from a jam-packed week of programming. So, we’ve pulled together a comprehensive summary of all 260 announcements we made across our products, partners, and customers at Next ‘26 below. Product news AI models, platforms and agents As you move toward an autonomous enterprise, where agents act with the same independence and reliability as members of your team, you need a new agentic foundation that can sustain your trust. Here’s how we do it: 1. Gemini Enterprise Agent Platform: Build, scale, govern, and optimize agents. As the evolution of Vertex AI, it offers the model selection, model building, and agent building capabilities that customers love, plus new features for agent integration, DevOps, orchestration, and security. Within Agent Platform, we announced many new features that help you move from managing individual AI tasks to delegating business outcomes with total confidence: 2. Agent Development Kit (ADK): Unlocks more powerful reasoning by organizing agents into a network of sub-agents. This new, graph-based framework allows you to define clear, reliable logic for how agents work together to solve complex problems. 3. Agent Studio: Go from building simple prompts to deploying complex agents in Agent Studio. Once you're ready for deep customization, export your logic directly into ADK to continue development in a full-code environment. 4. Agent Runtime: Delivers sub-second cold starts and allows you to provision new agents in seconds. 5. Agent Sandbox: Now available for everyone, provides a hardened environment to safely execute model-generated code and perform computer-use tasks like browser-based automation without risk to your host systems. Read more here. 6. Agent Memory Bank: Personalizes interactions by dynamically generating and curating long-term memories from conversations. Using new Memory Profiles, agents can recall high-accuracy details with low latency, so as not to lose context. Read more here. 7. Agent Sessions: Stores and manages history. With Custom Session IDs, you can use your own unique identifiers to track sessions and map them directly to your internal database and CRM records. Read more here. 8. Agent Identity: Assigns every agent a verifiable identity in the form of a unique cryptographic ID. This creates a clear, auditable trail for every action an agent takes, mapped back to defined authorization policies. Read more here. 9. Agent Registry: Maintains a central library of approved tools. It indexes every internal agent, tool, and skill, simplifying discovery and ensuring only governed, approved assets are available to your users. Read more here. 10. Agent Gateway: Manages your agent fleet from one control point, providing secure, unified connectivity between agents and tools across any environment, while enforcing consistent security policies and Model Armor protections to safeguard against prompt injection and data leakage. Read more here. 11. Agent Anomaly Detection: Detects suspicious behavior in real time and uses statistical models and an LLM-as-a-judge framework to flag unusual reasoning. It works alongside Agent Threat Detection to provide visibility into malicious activity, such as reverse shells or connections to known bad IP addresses. 12. Agent Security dashboard: Uncovers vulnerabilities automatically. Powered by Security Command Center, it unifies threat detection and risk analysis, allowing teams to map relationships between agents and models, automate asset discovery, and scan for vulnerabilities in the underlying operating system and language packages. Read more here. 13. Agent Simulation: Simulates realistic conversations and tests agents against human-like synthetic user interactions and virtualized tools in a controlled environment. Agents are automatically scored based on task success and safety across multi-step conversations. Read more here. 14. Agent Evaluation: Tracks live performance to continuously score agents against live traffic using multi-turn autoraters that can evaluate the logic of an entire conversation, not just a single response. With turnkey dashboards and Agent Observability, you can visually trace complex reasoning to debug issues as they happen. Read more here. 15. Agent Optimizer: Automatically clusters real-world failures and suggests refined system instructions to improve accuracy. Read more here. 16. Long-running agents: Execute complex business processes that can work autonomously in secure cloud sandboxes, giving agents the ability to orchestrate business logic, write code to build custom tools, and complete multi-step work — without needing constant prompting. Read more here. Then there’s Gemini Enterprise, the end-to-end system for the agentic era, which includes the Gemini Enterprise app and the following new capabilities: 17. Agent Designer: Uses the same no-code agent designer experience of Agent Platform and lets employees build sophisticated schedule- or trigger-based agents using any enterprise connector, and providing a virtual flowchart of your agent. 18. Inbox in Gemini Enterprise: Provides a central location to monitor, guide, and help manage all of your agent activity, including long-running agents. Notifications are categorized into actionable groups like "Needs your input," "Errors," and "Completed.” 19. Projects: Create a dedicated space where the agent’s memory is confined to the files and conversations your team adds, and connect it to data sources like Google Drive, NotebookLM, and Google Group Chats. 20. Skills: Create simple shortcuts using an “@” mention for repetitive tasks like applying brand guidelines, formatting a report, or accessing specific data. 21. Canvas: An interactive editor directly within Gemini Enterprise that lets teams easily create and edit Docs and Slides, and even export to Microsoft 365 files, within the same experience. 22. Enhanced Deep Research agent: Executes complex, multi-step research plans that autonomously synthesize information from the open web and internal company data. 23. Dynamic and interactive Agent UI experiences: Move beyond simple text chat with new support for the Agent-to-UI protocol (A2UI), enabling custom agents to dynamically generate rich, native UI components directly within the Gemini Enterprise app. 24. High-code agents: Allow your engineering teams and IT leaders to build sophisticated agents with full code and publish them directly into the Gemini Enterprise app for the business to use. 25. Data Insights agent for unified analysis: Uncovers deep patterns across your entire data landscape. 26. Integrations for Bring Your Own Model Context Protocol (BYO-MCP): Allows admins to securely connect the Gemini Enterprise app directly to your own custom or third-party business tools, internal workflows, and custom-built applications. 27. Cross-platform unified intelligence: Delivers unified intelligence across a fragmented data landscape where data lives in Google Workspace, Microsoft 365, or other business applications. You can try the Gemini Enterprise App today by signing up here. Our incredible AI ecosystem is also growing with Agent Gallery, where customers can browse a highly vetted set of agents built by our partners. 28. Agent Gallery: Provides access to new third-party agents from partners including Accenture, Adobe, Atlassian, Deloitte, Lovable, Oracle, Palo Alto Networks, Replit, S&P Global, Salesforce, ServiceNow, Workday, and more. AI and compute infrastructure We announced the eighth generation Tensor Processing Units, or TPUs, which consist of: 29. TPU 8t: The training powerhouse, specifically designed for high-throughput AI workloads, delivering nearly 3x higher compute performance than the previous generation to shrink training timelines for massive models. Learn all about TPU 8t here. 30. TPU 8i: A breakthrough system for inference and reinforcement learning (RL) workloads, engineered to deliver the ultra-low latency and up to 80% better performance-per-dollar advantage for agentic workflows and Mixture of Experts (MoE) models compared to the previous generation. Go deep on TPU 8i here. 31. Native PyTorch support for TPU, or TorchTPU: With TorchTPU, you can run models on TPUs as they are, with full support for native PyTorch features like Eager Mode. Now in preview with select customers. Whether you need a fully disconnected deployment, or want an integrated, Google-managed environment on your own hardware, Google Distributed Cloud (GDC) is the foundation of our sovereign neocloud architecture. Here’s the latest: 32. NVIDIA Blackwell GPUs: NVIDIA Blackwell (NVIDIA HGX B200) and NVIDIA Blackwell Ultra platforms (NVIDIA HGX B300) GPUs accelerate AI performance, leveraging fifth-gen NVIDIA NVLink to deliver data-center scale bandwidth directly to your environment. 33. New machine families: In addition to the N2 and N3 machine families for general-purpose workloads, GDC now supports the new A4 machine family for inference, and the memory-optimized M2 and M3 machine families for workloads that require higher memory-to-vCPU ratios. 34. Enhanced storage: GDC now supports 6PB object storage per zone — 6x the previous storage capacity. In addition, it now offers 30 IOPS/GB per zone, a 10x performance boost. We also announced several new instances and features across our Compute Engine family: 35. A5X with NVIDIA Vera Rubin platform: A5X will be powered by NVIDIA Vera Rubin NVL72 when it becomes available later this year, including concepts from the open-source Falcon networking protocol. Learn more here. 36. Google Axion N4A: Google’s custom Arm-based Axion CPU achieves up to 2x better price-performance than comparable current-generation x86-based VMs for cost-sensitive workloads, now GA. Read more here. 37. Google Axion N4A supports GKE Agent Sandbox. GKE Agent Sandbox lets agents safely execute untrusted code and tool calls, and running on Axion N4A instances provides up to 30% better price-performance than the next leading hyperscale cloud provider. Now GA. Read more here. 38. Google Axion C4A.metal: Our first Axion bare metal instance is designed to power Android development, automotive simulation, CI/CD pipelines, security workloads, and custom hypervisors. Now in preview. Read more here. 39. C4 instances support for Intel Xeon 6 (Granite Rapids): Achieve high-performance for AI workloads with Intel AMX and native FP16 support. C4 VMs are available with Intel Xeon 6 processors across all shapes. Read more here. 40. Flexible CUDs for more VM families: Flex CUDs are now available for memory-optimized (M1-M4) and HPC-optimized (H3, H4D) VM families, as well as Cloud Run. Learn more here. 41. C4N VMs: For high-volume network applications, C4N with Titanium adapters provides nearly 400 Gbps of VM-to-VM bandwidth, and C4N with Hyperdisk Extreme provides 25 GiB/s of block storage throughput and nearly 1M IOPS. Now in preview; sign-up here. 42. M4N instances: Designed for memory-intensive databases, M4N with Hyperdisk Extreme offers 26.57 GiB of RAM per vCPU. Sign-up for the preview here. 43. Z4D instances: Optimize I/O-intensive workloads such as SQL, NoSQL, and vector databases with up to 84 TiB of high-performance local SSD directly on the node. Z4D virtual machines and bare metal instances will be in preview soon. 44. Hyperdisk Balanced improvements: Hyperdisk Balanced High Availability now offers a 4x performance improvement for high-availability databases like SQL Server or PostgreSQL by dynamically routing full-disk performance to the active VM. Learn more here. 45. Hyperdisk ML performance improvements: With 2 TiB/s of aggregate throughput (up from 1.2 TiB/s), Hyperdisk ML helps eliminate AI storage bottlenecks, with more than 200x higher throughput per disk than competitive offerings. Learn more here. 46. Hyperdisk Exapools: For large-scale training, Hyperdisk Exapools offer the highest aggregate block storage performance and capacity per AI cluster of any hyperscaler. Now GA, learn more here. 47. Z4M: For distributed parallel file systems and large-scale AI/ML workloads, Z4M provides up to 168 TiB of local SSD, up to 400 Gbps of network bandwidth, and support for RDMA. Z4M VMs and bare metal instances are expected in preview in Q3 2026. If you want help deploying agents in your environment, you can partner with our experts for a diagnostic of your data and infrastructure by signing up here. Agentic Data Cloud As AI shifts from reactive intelligence to proactive action, legacy data architectures are breaking down under the agentic load. Our Agentic Data Cloud closes the gap between thinking and doing, merging analytical history with transactional power in a real-time loop, so your agents can securely act on your behalf. The Agentic Data Cloud comprises many products and platforms, starting with: 48. Knowledge Catalog: A universal context engine that maps and infers business meaning across your data estate using aggregation, enrichment, and search to help agents execute tasks accurately. Read all about Knowledge Catalog here. 49. Deep Research Agent powered by the Knowledge Catalog: Synthesizes business data, internal docs, and web research to autonomously answer your hardest, most complex questions. 50. Zero-copy federation: Provides direct access to data and context across applications, operating systems, and AI platforms — including Palantir, Salesforce Data360, SAP, ServiceNow, and Workday. Read more here. 51. LookML Agent: Autonomously reads strategy documents, spreadsheets, and reports to generate business-ready semantics, ensuring AI and analysts use the exact same enterprise definitions. 52. BigQuery measures: Redefines data consistency by embedding programmatic business logic directly into the SQL engine, ensuring calculations are reusable, accurate, and governed. 53. Data Products: Packages data assets with built-in intent, SLAs, and governance constraints. These self-contained intelligence blocks ground agents reliably in production. 54. Reusable data quality rules: Scales consistent, high-quality data governance by combining reusable, expert-defined test logic across multiple assets to ensure trusted data. 55. Structured approval workflows: Validates new business terms and data products with structured workflows, ensuring changes are verified by domain experts before reaching AI. 56. Column-level lineage: Tracks exactly where your agent's truth originated. We’ve expanded end-to-end lineage tracking to include Looker, plus column-level lineage for BigQuery and Spark. An AI-native, cross-cloud lakehouse is another foundation element of the Agentic Data Cloud. Here’s what’s new in ours: 57. Cross-cloud Lakehouse (formerly BigLake): Breaks down data silos with a borderless lakehouse. Powered by Iceberg REST Catalog, agents seamlessly access data across AWS, Azure, and a vast partner ecosystem. 58. Lakehouse catalog federation: Enables easy discovery, analysis, and zero-copy sharing of your data seamlessly across AWS Glue, Databricks, Snowflake, and SAP. 59. Lightning Engine for Apache Spark: Accelerates your agentic data science workloads with an engine that delivers up to 2x the price-performance over proprietary market alternatives. Read more here. 60. Lakehouse federation for AlloyDB: Enables live joins between your transactional data and massive historical analytical insights without moving data or building complex ETL pipelines. 61. Managed Iceberg storage & REST Catalog: Benefits from automatic management and multi-table transactions, plus read/write interoperability across BigQuery, Spark, and OSS engines. 62. Cross-cloud caching: This intelligent cache stores cross-cloud data on the first read to slash massive egress fees and accelerate follow-on queries for AWS and Azure data. For practitioners, we announced several products with enhanced automation: 63. Google Cloud Data Agent Kit: A portable suite of skills, MCP tools, and plugins that turns environments like VS Code and Gemini CLI into native data workspaces that autonomously orchestrate outcomes. Read more here. 64. Data Science Agent: Simply state your goals in plain English to automatically execute complex plans for loading, cleaning, and visualizing data in BigQuery notebooks. Read more here. 65. Data Engineering Agent: Autonomously builds complex Dataform pipelines from scratch while enforcing governance rules and testing to keep bad data out of your production systems. Read more here. 66. Database onboarding and observability agents: Lets agents guide you through database provisioning, or proactively monitor health and execute automated troubleshooting across your database fleet. Read more here. 67. Conversational Analytics: Empowers teams to query complex datasets using natural language and publish custom analytical agents directly into Gemini Enterprise. Read more here. We also announced new products and features across our database portfolio: 68. Spanner Omni: This edition brings Spanner’s globally consistent, multi-model database beyond Google Cloud. Run it anywhere — across multiple clouds, on-premises, or on your laptop. Read more here. 69. Spanner Columnar Engine: Accelerates analytical queries up to 200x on live operational data using vectorized execution, without impacting your production transactional workloads. 70. Managed remote MCP servers for databases: Securely manages the infrastructure to connect AI models directly to your operational data, eliminating the burden of hosting MCP servers. 71. Vibe coding integrations with AI Studio: Agent-led workflow that generates live applications from simple text prompts, easily connecting to trusted database services. 72. Tools for Data Agents: Modular building blocks that provide ready-to-use functions, like high-accuracy text-to-SQL, giving custom agents secure and direct database access. 73. Bigtable in-memory: A new hybrid storage architecture natively integrates frequently accessed data in-memory to deliver sub-millisecond read latency for real-time apps. 74. Memorystore for Valkey 9.0: Provides a fully managed migration path from self-managed Redis, complete with enterprise-grade security features like Access Control Lists (ACLs). 75. AlloyDB AI-powered search at scale: Scales enterprise search to 10 billion vectors using Google’s ScaNN index, offering up to 6x faster queries than standard PostgreSQL. 76. New AI functions for AlloyDB: Uses optimized modes for ai.if , ai.analyze_sentiment , and ai.summarize to embed Gemini’s powerful reasoning directly into database queries. 77. Firestore full-text and geospatial search: Google’s search technology is now natively integrated into Firestore, delivering highly relevant keyword and geospatial queries. 78. Compute Engine-to-managed migrations: Minimizes migration complexity with automated networking and replication to move PostgreSQL workloads to Cloud SQL or AlloyDB easily and securely. 79. Database Center with BigQuery: Monitors your entire data estate — from transactional databases to analytical warehouses — from a single management plane powered by Gemini fleet analytics. BigQuery, now more than ever, is the linchpin of modern, agentic data analytics, and includes many new capabilities: 80. Reverse ETL for BigQuery: A one-click solution to sync heavy analytical insights from your lakehouse back to AlloyDB or Spanner for low-latency, high-QPS serving to end users. Read more here. 81. BigQuery Graph: Maps entities, relationships, and business logic directly within your data platform to anchor AI agents in a governed reality for complex reasoning tasks. 82. BigQuery fluid scaling: Executes highly variable, agentic workloads with per-second billing. This premier autoscaling model instantly adjusts resources. 83. ObjectRef: Processes unstructured data alongside structured data using SQL and Python. Build rich, multimodal context on your Knowledge Catalog without moving data. 84. AI.PARSE_DOCUMENT: Part of BigQuery AI, simplifies complex document processing with a single SQL function that automates OCR, layout parsing, and chunking using Gemini’s layout parser. 85. TabularFM model: Part of BigQuery AI, brings high-quality, zero-shot regression and classification natively to BigQuery without requiring extensive feature selection or model management. 86. Optimized mode for SQL-first AI: Part of BigQuery AI, trains task-specific models on the fly for functions like AI.CLASSIFY and AI.IF, delivering up to a 230x reduction in token consumption and cost. 87. BigQuery-native Gemma embeddings: Part of BigQuery AI, generates high-quality vector embeddings at scale on standard CPUs directly within your data warehouse to simplify enterprise AI preparation. 88. Autonomous embedding generation: Part of BigQuery AI, fully manages pipelines for unstructured data, automatically keeping vector indexes in sync asynchronously as new data is ingested at BigQuery scale. 89. BigQuery hybrid search: Part of BigQuery AI, unifies semantic and full-text search into a single powerful function to deliver superior retrieval precision for RAG and complex exploration tasks. 90. Python UDF: Part of BigQuery AI, enrich, transform, or clean data using fully managed Python scalar functions that autoscale serverlessly to millions of rows directly in BigQuery. 91. Connected Sheets with TimesFM: Part of BigQuery AI, brings the scale of BigQuery to Google Sheets, featuring state-of-the-art zero-shot forecasting with the TimesFM model and robust anomaly detection. 92. Geospatial analytics datasets: Part of BigQuery AI, combines native Google Earth AI and roads management data directly with your enterprise data for deep, real-world agentic analysis without wrangling. 93. Proactive agentic workflows: With BigQuery, agents can go beyond basic Q&A to detect metric shifts, perform root-cause analysis, and deliver scheduled research briefings directly to your inbox. 94. BigQuery Agent Analytics plugins: Logs detailed interactions from frameworks like ADK and LangGraph directly to BigQuery to audit, troubleshoot, and optimize long-term agent performance. 95. BigQuery Studio productivity tools: Brings a contextually aware assistant, SQL and Visualization Cells, Files Explorer, and Git workflows directly into BigQuery. 96. Advanced runtime, small query, and history-based optimizations: These BigQuery optimizations have accelerated performance by 35% year-over-year while cutting processing costs. 97. New BigQuery workload management features: Granular cost controls arrive with reservation groups, dynamic assignments, and project-level slot controls, simplified by rules-based management. 98. Colab Data Apps: Transforms your notebook analyses into shareable, fully managed interactive Python applications that business teams can access directly from Data Studio. With Looker Business Intelligence, you get not just agentic AI, but agentic BI. Here’s the latest: 99. Data Studio (formerly Looker Studio): Available in Free and Pro versions, this data visualization engine now also provides a unified interface to browse data sources, apps, and agents. Read more here. 100. Looker Dashboard Agents: Transforms static views into interactive hubs. Ask natural language questions within dashboards for instant summaries and context-aware, governed answers. 101. Agentic Workflows in Looker: Moves beyond reactive queries to proactive results. Agents monitor critical metrics for irregularities and identify hidden correlations so you can instantly address business shifts. 102. Embedded conversational experiences in Looker: Leverages Looker's API and SDK to integrate Gemini-powered, multi-turn conversational experiences into custom data applications and internal workflows. 103. Looker self-service explores: Blends personal files with enterprise data using an updated AI-Explore experience that automatically generates LookML, giving you the agility of a spreadsheet with enterprise governance. 104. Looker AI assistants: Simplifies BI with Visualization, Expression, and Insight Assistants. Use natural language to create charts, generate custom filters, and automatically summarize key trends. 105. Open BI and MCP: Empowers external AI agents to autonomously query and analyze data. We offer both open-source MCP via the MCP Toolbox and an all-new Google-managed MCP server that’s native to Looker for ease of management. 106. Continuous Integration: In Looker, automates SQL validation and content testing within your development workflow to identify potentially breaking changes before they reach production, for reliable model updates. Application development, management and runtime platforms Cloud Run, Google’s fully managed application platform, now supports: 107. Full-stack apps in Google AI Studio: Build and deploy apps vibe-coded in AI Studio, including a Firestore database and user authentication, to Cloud Run, now GA. 108. Integration with Gemini Enterprise Agent Platform: Through its integration with Cloud Run, Agent Platform helps you transition agents from experimental environments into fully managed, production-grade systems. Now in preview with select customers. 109. Fully managed remote MCP server: Enable easy deployments from developers and agents, now GA. 110. NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs: Run high-end inference on Cloud Run — up to 70B+ parameter models — without having to manage any underlying GPU infrastructure, and scaling to zero when the resource is not in use. 111. Billing caps: Define your maximum spend per month, after which point, Cloud Run resources are de-activated. Coming soon. 112. Cloud Run instances: Host long-running background jobs, in preview to select customers. 113. Cloud Run sandboxes: Spin up an ephemeral sandbox that’s strictly isolated from your agent code using a built-in sandbox tool, coming soon. 114. SSH: Gain secure shell access (SSH) directly into a running Cloud Run container, now in preview with select customers. 115. Cloud Run service bindings: Service-to-service communication for scalable microservices architectures, coming soon. 116. Ephemeral disk: Now in preview, ephemeral disk storage is created when an instance starts and deleted when it stops. And for mobile and web developers, we added new capabilities to Firebase, our premier platform for client developers: 117. Firebase SQL Connect: Firebase Data Connect is now Firebase SQL Connect, letting you connect your mobile and web client apps to the power of Cloud SQL for PostgreSQL, now with support for realtime syncing, offline cache, and native SQL— all available via a no-cost trial. 118. Firebase integration with Application Design Center: The new Firebase integration with Application Design Center (ADC) offers a unified model for deploying and managing the resources used by your mobile and web client apps, such as Firestore, App Check, Authentication, and Firebase AI Logic, in the same way you manage the rest of your Google Cloud infrastructure. 119. Firestore Enterprise edition: Introducing native full-text search, geospatial queries, relational-style JOIN using subqueries, and a robust data manipulation language (DML) as part of the new query engine in Firestore Enterprise. 120. Firebase AI Logic: Build generative AI features directly into your mobile and web client apps with no server-side setup. New features, like auto function calling and JSON schema mapping, help reduce technical overhead and give you more control to optimize your resources and costs. Read more here. 121. Firebase Phone Number Verification: Helps onboard new users to your app by retrieving the phone number assigned to the SIM in a device over cellular or WiFi. Now supported by 10+ carriers across six regions. Then, there’s Google Kubernetes Engine, or GKE, which has emerged as the platform of choice for complex, demanding AI and agentic applications. Here’s what’s new in GKE: 122. GKE Agent Sandbox: A GKE add-on based on the open-source Agent Sandbox controller project that manages isolated, stateful, and single-replica workloads on GKE, optimized for AI agent runtimes. Learn more here. 123. GKE hypercluster: A single, conforma