Google Plots BigQuery Path from Data to AI

A little under a year ago, I explained how Google was positioning its BigQuery product as a unified data platform for processing data in multiple formats, across multiple locations, for multiple use cases—including business intelligence (BI) and artificial intelligence (AI)—using a combination of multiple data engines, including SQL, Spark and Python. The evolution of BigQuery as the focus of Google’s analytics and AI offerings continued at the recent Google Cloud Next ‘25 event, with Google now describing BigQuery as an autonomous data-to-AI platform and highlighting its ability to support multi-mode data processing, addressing batch and stream processing of structured and unstructured data, with enhanced support for AI workloads and new data-focused agents.

Google Cloud, a subsidiary of Alphabet Inc., is a cloud service provider offering a portfolio of infrastructure as a service as well as SaaS applications, including, among others, a variety of AI, analytic and data services. The company today has a broad portfolio of data and analytics offerings that spans data storage and processing, data management and data governance, as well as analytics, AI and generative AI (GenAI) and machine learning (ML). My colleague David Menninger has provided a deep dive into the GenAI and agentic related announcements made at Google Cloud Next ‘25, including those related to Agentspace, Agent Builder and AI Agent Marketplace, as well as the Gemini family of models and AI infrastructure. In this Analyst Perspective, we’ll focus on the analytic and data platforms products. The flagship offering in Google’s analytics and data portfolio is the Google BigQuery distributed serverless analytic data platform. This has long been a critical component of Google Cloud, having been the third product launched on the platform in 2010, following Google App Engine and Google Cloud Storage. More than one-half (58%) of participants in ISG Research’s Market Lens Cloud Study are using cloud for the majority of their data platforms. I have previously written about the critical importance of trusted data and data management to agentic AI, and this is also reflected in Google’s increasing focus on BigQuery as an autonomous data-to-AI platform that can provide a foundation for multimodal data processing, as well as native integration with Google’s Looker and Vertex AI environments for BI and AI workloads.

The level of focus Google is placing on BigQuery was highlighted by the sheer number of announcements made at Google Cloud Next ’25. There were too many to cover in detail, so I’ll pick out some of the most significant. This includes the addition of the BigQuery AI Query Engine, which enables co-processing of SQL queries alongside Gemini models in BigQuery to facilitate the integration of trusted proprietary enterprise data and GenAI output at runtime. Google also announced the preview of BigQuery Knowledge Engine, which uses Gemini models to analyze data schema, table descriptions and query patterns to automatically generate metadata, identify relationships and recommend business glossaries. BigQuery knowledge engine provides the foundation for grounding AI models and agents with trusted business context and is also the foundation for semantic search across BigQuery. BigQuery Vector Search and BigQuery continuous queries also both reached general availability, with the former providing the ability to generate and manage vector embeddings and the latter enabling the continuous processing of SQL statements to transform and query event data as it is ingested into BigQuery. I assert that through 2027, almost all enterprises developing applications based on GenAI will invest in data platforms with vector search and retrieval-augmented generation to complement foundation models with proprietary data and content.

Google also made a series of announcements related to data management and governance, including the preview release of BigQuery Multimodal Tables to enable unified storage and querying alongside structured data. Google also enhanced BigQuery governance to provide a single view for data stewards to manage and govern data discovery, classification and sharing. The new BigQuery universal catalog combines the company’s existing Dataplex Catalog with the BigQuery metastore, which is designed to work with multiple engines, including BigQuery, Apache Spark, Apache Hive and Apache Flink, and supports the Apache Iceberg table format. This enables multiple analytics engines to query data regardless of whether it is stored in BigQuery storage tables, BigQuery tables for Apache Iceberg or external tables in the BigLake storage engine. Google also announced the preview of serverless execution of Apache Spark workloads within BigQuery as well as the general availability of Google Cloud for Apache Kafka. Given the increased enablement of multiple engines within BigQuery, the GA of advanced workload management capabilities is also significant, as is BigQuery Spend Commit pricing that enables users to manage spending across the various data processing engines, streaming and governance environments.

BigQuery was by no means the only focus of attention at the Google Cloud Next ‘25 event. There were multiple announcements related to the AlloyDB operational data platform, including the ability to search structured data in AlloyDB with Google Agentspace, enhanced AlloyDB support for natural language query, vector search and structured filters and joins, as well as improvements to AlloyDB AI, with new models for vector search, multimodal and text embedding. AlloyDB is an important component of the Google Distributed Cloud air-gapped private cloud offering for regulated customers with data sovereignty requirements, and will be complemented by the imminent availability of Gemini models, Vertex AI and Agentspace on GDC. The company also announced MCP Toolbox for Databases, enabling developers to use Model Context Protocol to create connections between AI agents and multiple Google databases, including AlloyDB for PostgreSQL, Spanner, Cloud SQL for PostgreSQL, Cloud SQL for MySQL, Cloud SQL for SQL Server, and Bigtable. Google also announced the preview of three new specialized agents targeted at data engineers, data scientists and analytics users. Embedded in the BigQuery pipelines functionality, the data engineering agent capabilities are designed to facilitate pipeline generation, data preparation and anomaly detection as well as automate metadata generation. The data science agent is embedded within Google’s Colab notebook to automate feature engineering, facilitate model selection, and accelerate model training and iteration. Looker conversational analytics adds natural language query capabilities to the Looker BI platform, including integration with Looker’s semantic layer for business context, while Google also delivered a new conversational analytics API, enabling developers to embed conversational analytics into applications and workspaces.

Google was rated as an Exemplary provider in our 2024 Buyers Guides for Data Platforms, Analytic Data Platforms and Operational Data Platforms, and continues to enhance its overall data portfolio, with a particular emphasis on BigQuery and AlloyDB. Not all the new announcements are generally available, but the Google Cloud Next ‘25 event showcased the potential advantages of tight integration between Google’s data platforms and AI capabilities. I would encourage any enterprises evaluating potential providers for data platforms to support BI and AI initiatives to include Google in their assessments.

Regards,

Matt Aslett