Succeeding in the agentic era requires a transformation in your data strategy: moving from human-scale to agent-first workloads, evolving from reactive intelligence to proactive action, and shifting from raw data to semantic knowledge that agents can use to reason accurately.
For over a decade, BigQuery’s continuous innovations have helped tens of thousands of organizations build a scalable data and AI foundation and navigate several industry and technological transformations. BigQuery has evolved into an autonomous data-to-AI platform, growing over 30x in data processed with Gemini, 25x in AI functions processing unstructured data, and 20x in agent-building tools with Model Context Protocol (MCP).
Customers like Definity are building data platforms to enhance their customers’ experience, improve back-office operations, and boost data team productivity.
“We stood up our data platform in Google Cloud and ingested all critical insurance data in 10 months, which is about half of the time that people see in the industry. The technology that BigQuery provides, processing large amounts of data very quickly, is giving our practitioners and engineers tools that are advanced and a platform that has AI and ML built in. We have doubled the number of users [in a very short period of time].” — Tatjana Lalkovic, Chief Technology Officer, Definity
Today, we are announcing new BigQuery capabilities in lakehouse, built-in AI processing and reasoning, and agentic experiences, all anchored by our commitment to industry-leading price-performance and enterprise readiness.
Open, cross-cloud lakehouse
Enterprise data is often scattered across applications, multiple cloud environments and on-premises. While early lakehouse solutions reduced data duplication, the agentic era demands a foundation that is natively multimodal, cross-cloud, and AI-ready. Our approach blends Apache Iceberg’s interoperability and Google’s differentiated infrastructure with new capabilities, including:
Managed Iceberg tables in Lakehouse (GA, formerly BigLake) enables the openness of Iceberg with advanced BigQuery capabilities, including automatic table management, Iceberg partitioning, multi-table transactions, change data capture, enhanced vectorization, and history-based optimizations.
Iceberg REST catalog enables read/write interoperability (preview) on Iceberg tables between BigQuery, Spark, and other OSS and third-party engines, so you don’t have to make complex engine trade-offs.
Cross-cloud Lakehouse (preview) brings BigQuery AI and analytics to other clouds, starting with AWS and Azure. Using open standards like the Iceberg REST Catalog, high-bandwidth networking via Cross-Cloud Interconnect, and transparent caching, BigQuery achieves performance and total cost of ownership comparable to native warehouses, enabling true cross-cloud for enterprises.
Catalog federation (preview) enables easy discovery, analysis, and zero-copy sharing of data across AWS Glue, Databricks, SAP, Salesforce, Snowflake, and Confluent Tableflow (coming later this year).
Real-time data replication that closes the loop between raw data and operational action by allowing you to replicate data from Spanner, AlloyDB, and Cloud SQL instantly into BigQuery tables (GA) and Iceberg (preview).
Analyze data across BigQuery and Iceberg table on AWS using SQL or natural language
Refer to this blog to learn more about our latest lakehouse innovations.
Graph-based reasoning for enterprise agents
For AI to solve truly complex, multi-hop operational problems, business logic must extend all the way down to the data platform layer. Defining logic at this layer ensures that definitions remain consistent and governed from ingestion to consumption. BigQuery Graph (preview) provides the foundation to activate this context, allowing data practitioners to map entities, relationships, and business logic directly within the data platform. This anchors AI agents in a governed reality, enabling them to solve sophisticated challenges at scale with consistent accuracy.
Today, we are announcing new graph capabilities to enhance AI reasoning:
Native support for measures in BigQuery Graph (preview) enables you to unify analytical metrics and relationships into a single, governed entity. It transforms data into a “business map” for multi-hop structural reasoning, allowing agents to move beyond simple searches to trace the ripple effects of business events.
Graph support in BigQuery Conversational Analytics (preview) allows conversational analytics agents to navigate a deterministic business map instead of raw tables to provide answers with higher accuracy. Graphs enable dual reasoning: agents can instantly calculate precise KPIs using measures while simultaneously traversing complex relationships to uncover the “why” behind the numbers.
BigQuery Graph and Looker integration (preview) allows you to reuse measures defined in BigQuery across your data stack by exposing graphs as Looker views. You can also define BigQuery Graphs using Looker with source control and validation. This interoperability ensures that metrics, such as Churn Rate, remain identical across both dashboards and AI agents.
Visual modeling experience in BigQuery Studio (preview) provides an intuitive interface to easily build and manage the entities, relationships, and business logic that power your agentic context.
Analyze data in natural language using Conversational Analytics Agents with BigQuery Graph
Native AI processing to unlock structured and unstructured data
Your data is no longer confined to rows and columns; agents demand a platform that can work across structured and unstructured data at scale without requiring data copies or movement. BigQuery AI makes it easy to perform predictive machine learning tasks (like forecasting, anomaly detection, and recommendations) and generative AI tasks (like entity extraction from images and text, content generation, and data enrichment), with built-in access to over 170 foundational models. New capabilities in BigQuery AI include:
AI.PARSE_DOCUMENT (preview) simplifies complex document processing workflows with a single SQL function that automates Optical Character Recognition (OCR), layout parsing, and chunking.
TabularFM model (preview) brings high-quality regression and classification to BigQuery without the need for extensive feature selection, tuning, training or model management.
ObjectRef (GA) allows you to process unstructured data alongside structured data using SQL and Python. This establishes the foundation for building rich, multimodal context directly on your Knowledge Catalog.
Optimized mode (preview) for SQL-first, AI co-processing-managed functions like AI.CLASSIFY and AI.IF trains task-specific models on the fly, delivering a 230x reduction in tokens consumed compared to row-by-row gen AI processing.
BigQuery-native Gemma embeddings (preview) allow you to generate high-quality embeddings at scale on standard CPUs.
Autonomous embedding generation (GA) fully manages the pipeline for unstructured data, automatically, keeping vector indexes in sync as new data gets ingested.
BigQuery hybrid search (preview) unifies retrieval by integrating semantic and full-text search into a single function, delivering superior precision for RAG and complex exploration.
Python UDF (GA) allows you to enrich, transform or clean data with fully managed Python scalar functions. You can bring your own code or libraries and the functions will autoscale to millions of rows with serverless, scale-out execution. This will be rolled out in phases over the next few weeks.
Connected Sheets brings the scale of BigQuery to the familiar Google Sheets interface, now supporting forecasting with TimesFM model (GA) and anomaly detection (preview).
Geospatial analytics datasets (preview) can now be accessed directly from BigQuery, allowing you to combine geospatial data — such as infrastructure assets and road management insights — with your enterprise data for deeper analysis, without complex manual data wrangling.
Access, process and activate unstructured data in BigQuery
Agentic experiences
AI enables highly skilled data teams to focus on high-impact strategic initiatives by offloading tedious and time-consuming tasks. We are pioneering this space by integrating Google’s foundational research, capable models, and specialized tools directly into BigQuery, providing automation and assistance at every stage of the data lifecycle. New agentic experiences in BigQuery include:
Conversational Analytics in BigQuery (GA) allows teams to query complex datasets using natural language. It provides a secure and transparent path to insights while supporting advanced features like predictive analytics and reasoning across structured and unstructured data. In addition to BigQuery Studio, BigQuery Conversational Analytics agents can be published to and consumed from Data Studio (Preview), Gemini Enterprise (Preview), or via an API for custom applications (GA).
Gain insights from your data using natural language with Conversational Analytics in BigQuery
Proactive agentic workflows (preview) go beyond simple questions to detect metric shifts, perform root-cause analysis to explain why a change occurred, and deliver scheduled research briefings directly to your inbox.
Proactive agentic workflows in BigQuery
BigQuery Agent Analytics offers plugins for ADK (GA) and LangGraph (Preview) frameworks to record agent activity to BigQuery and analyze it for troubleshooting, optimization, and agent evaluation.
BigQuery Studio gets new productivity tools, including a contextually aware assistant (preview) for resource discovery and troubleshooting, SQL Cells (GA) for blending SQL and DataFrames, and Visualization Cells (GA) to create visuals directly within notebooks. The Files Explorer (GA) allows developers to organize, share, and manage code assets in folders. Additionally, we are introducing Git integration and workflows (preview) in BigQuery Notebooks, providing complete SCM coverage across GitHub, GitLab, Bitbucket, and Azure DevOps for data science workflows.
Data Science Agent (GA) in BigQuery notebooks allows you to simply state goals in plain English to automatically execute plans for loading, cleaning, and visualizing data using BigQuery ML, DataFrames, or Spark.
Colab Data Apps (preview) bridge the gap between analysis and action by transforming notebook analyses into shareable, fully managed interactive Python applications that business teams can access from Data Studio.
BigQuery remote MCP server (GA) and BigQuery ADK toolset (GA) reduce the need for manual database connectors for agents.
Google Cloud Data Agent Kit (Preview) provides a portable suite of skills, Model Context Protocol (MCP) tools, environment-specific extensions, and native plugins. By meeting you where you build, like VS Code, Gemini CLI, Codex, and Claude Code, the Data Agent Kit turns your IDE, notebook, or terminal into a native data environment. This enables your environment to orchestrate a massive range of data workflows, automatically selecting the right frameworks (like BigQuery, dbt, Apache Spark, or Apache Airflow) and generating production-ready code.
Unparalleled performance and scale
Modern analytical workloads are increasingly unpredictable and distributed. We continue to invest in enhancing BigQuery’s core engine to ensure that as data grows in your environments, your costs and operational overhead do not.
Fluid scaling (GA) enables you to execute highly variable workloads with a premier autoscaling model that does not require a cost-and-performance trade-off. Fluid scaling in BigQuery enables true per-second billing, offering up to 34% cost savings.
Advanced runtime, small query, and history-based optimizations (GA) accelerate native and Iceberg workloads without code or schema changes. BigQuery has improved query speed by 35% year over year while reducing query processing costs by 40% year over year.
New workload management features — including reservation groups (GA), flexible dynamic assignments (preview), and project-level slot and concurrency controls (preview) — provide granular cost attribution and price-performance control, all simplified by declarative, rules-based workload management (preview).
Enhanced observability (GA) and intersection routing (preview) provide flexible disaster recovery capabilities for mission-critical workloads. Agent-powered observability (preview) for turnkey troubleshooting further simplifies operations, while the agent-ready security center (preview) provides a unified, fine-grained access control experience.
BigQuery fluid scaling for unpredictable workloads
In the agentic era, BigQuery isn’t just where your data lives — it’s where your data thinks, reasons, and acts. The future is here, and it’s more open, autonomous, and capable than ever.
Get started on your data and AI journey by taking advantage of our BigQuery data migration offer. We can’t wait to hear how you’re innovating with data.