Snowflake Summit 2025: The biggest news in AI, agents

“These are projects that are going to define the future of the [customer’s] company,” Snowflake CEO Sridhar Ramaswamy said.

Snowflake Intelligence for natural language interaction with data. New data science AI agents that can help build machine learning pipelines. And a private preview for an Adaptive Compute service for automatic resource sizing and sharing.

These are some of the most exciting reveals to come out of the AI and data cloud vendor’s annual Snowflake Summit 2025 conference, which runs this week through Thursday in San Francisco.

Snowflake CEO Sridhar Ramaswamy told CRN that the world of IT services is undergoing “a major change” as AI takes off and top executives look to AI to create more efficient, productive companies and shorten product life cycles.

“These are no longer IT projects,” Ramaswamy said during a pre-Summit press conference in answer to a CRN reporter’s question. “These are transformation projects. These are projects that are going to define the future of the company.”

Snowflake Summit 2025

Snowflake’s 2025 channel goals include increasing the overall percentage of company revenue that comes through the channel and enabling partners to develop an AI strategy and sell AI solutions, according to CRN’s 2025 Channel Chiefs.

The vendor has more than 10,000 partners worldwide, up from 600 in 2022, according to Snowflake.

Christian Kleinerman, Snowflake’s executive vice president of product, added that opportunity in the AI era “is bigger than ever” and goes beyond traditional customer relationships with system integrators and partners.

“With the advent of AI, virtually every area of a business now has the opportunity to get improved, to increase productivity with AI,” he said. “We're in very close touch with many of our partners on how do we help organizations unlock that promise of better business outcomes, faster business outcomes and agility on AI.”

In response to a CRN reporter’s question on whether Salesforce’s pending acquisition of Informatica for $8 billion will impact Snowflake and its partners, Amy Kodl, the vendor’s interim alliances and channel lead, said in a statement that the deal “only underscores the importance of data in powering enterprise innovation in the age of AI.”

“Since inception, Snowflake has served as a leader in the data and AI landscape and will continue to deliver a unified, open data platform designed for the agility and AI innovation businesses demand,” Kodl said.

She added that Salesforce is a strategic Snowflake partner and that the two companies will “continue to empower joint customers through ongoing collaboration.”

Snowflake’s annual Summit comes on the heels of an upbeat fiscal 2026 first quarter earnings report. The vendor beat expectations on revenue during the quarter and increased the amount of expected product revenue growth for the full fiscal year–increasing the amount by $45 million and increasing percent growth year over year from 24 to 25.

Still, the vendor showed some areas that need improvement. That 25 percent growth figure is a deceleration from 30 percent growth in fiscal 2025 and 38 percent growth in fiscal 2024, according to a report by investment firm Bernstein. The company has also seen 13 quarters of decelerating product revenue and its net revenue retention (NRR) fell to 124 percent compared to 126 percent the prior quarter.

Here’s more of what to know as Snowflake Summit 2025 kicks off.

Snowflake Intelligence, data science agents

Snowflake will soon launch public previews of Snowflake Intelligence, a unified conversational experience powered by data agents and data science agents, which aim to boost data science productivity through automating routine machine learning (ML) model development tasks.

With Snowflake Intelligence, business users and data professionals alike will have the ability to ask natural language questions and uncover actionable insights in structured data tables and unstructured documents.

Combined with data science agents, users can simplify AI and ML workflows and eliminate technical overhead that can slow down decision-making.

Users can leverage these offerings to unite data silos while maintaining enterprise-grade security and compliance, according to Snowflake. Large language models (LLMs) from Anthropic and OpenAI power the offering alongside Snowflake’s own Cortex Agents–which are in public preview.

Intelligence will run inside existing Snowflake environments, inheriting all security controls, data masking and governance policies. It is designed to unify data from Snowflake, Box, Google Drive, Workday, Zendesk and more sources. Agents can generate visualizations based on enterprise data, look up internal knowledge, and access third-party knowledge through Cortex Knowledge Extensions on Snowflake Marketplace, which is slated for general availability (GA) “soon,” according to Snowflake.

Data science agents, which leverage Anthropic’s Claude AI tool, aim to create fully functional ML pipelines executed from a Snowflake Notebook to speed up the iteration process and save data science teams hours of work spent on experimentation and debugging.

SnowConvert AI, Cortex AISQL

Snowflake is using Summit 2025 to unveil its SnowConvert AI agentic automation offering to accelerate migrations from legacy platforms to Snowflake and modernize data infrastructure faster–plus, a public preview for Cortex AISQL, a tool to bring Structured Query Language data transforming power to unstructured text.

Snowflake cited Oracle, Teradata and Google BigQuery as particular “legacy” data warehouses that users can migrate off of through SnowConvert AI. SnowConvert AI can apply to data warehouse migrations, business intelligence (BI) migrations, and extract, load, transform (ELT) computing migrations.

SnowConvert AI can migrate business intelligence and ETL tools as well as data ecosystems without disrupting critical workflows, according to the vendor. Snowflake estimates code conversion and testing phases become twice or three times as fast with SnowConvert AI.

With Cortex AISQL, users can extract insights across multi-modal data and build pipelines with SQL and AI, according to Snowflake. AISQL can work on text, images, audio and more. Models from a variety of leading vendors power AISQL, which also uses functionality and performance optimizations from Snowflake’s SQL engine.

Snowflake is offering a private preview for Cortex AISQL performance optimization capabilities, which promises to improve performance up to 70 percent depending on datasets, with up to 60 percent cost savings when filtering and joining data across thousands of records.

Analysts can enrich customer tables with chat transcripts, correlate sensor readings with inspection photos, merge sales figures with social media sentiment, and perform other tasks that combine traditional structured data and unstructured data, according to Snowflake.

During Summit 2025, Snowflake said it has moved semantic model sharing into private preview while promising to make Cortex Knowledge Extensions GA “soon” on its Marketplace so that users can bring in proprietary unstructured data from third-party providers with intellectual property (IP) protection and attribution.

The vendor has also put Agentic Snowflake Native Apps on its Snowflake Marketplace. These interoperable agentic products are standalone or can act as building blocks for apps created on Cortex Agents or within Snowflake Intelligence.

Semantic model sharing is designed to allow users to integrate AI-ready structured data within Snowflake Cortex AI applications and agents, according to the vendor. Semantic models give LLMs definitions and an understanding of structured data and business concepts for asking questions in natural language and getting back quality answers.

Users can share and access semantic models internally with the Internal Marketplace or with third-party providers through Snowflake Marketplace and maintain their governance and version control.

Cortex Knowledge Extensions will work with a variety of business articles and content from The Associated Press, USA Today, CB Insights, Stack Overflow and other providers, according to Snowflake. The move will create enterprise AI system real-time feeds of external content and knowledge for improved AI responses. The extensions apply to textbooks and research papers as well, which enterprises can purchase.

Extensions leverage retrieval-augmented generation (RAG) to permit access. Snowflake’s Zero-ETL Sharing functionality allows content owners to revoke access at any time.

Snowflake Openflow GA, dbt project building

Snowflake has made its Openflow multi-modal data ingestion service generally available on Amazon Web Services (AWS) and plans to “soon” allow native dbt project building, running and governing in Snowflake.

Openflow aims to simplify data accessibility and AI readiness and eliminate fragmented data stacks and hours of manual data ingestion labor, according to the vendor. It works on open standards to allow data integration in a single, unified platform without vendor lock-in.

With native dbt project building and governing, Snowflake users don’t need to spend as much time on infrastructure maintenance and can leverage Snowflake Workspaces with dbt for in-line AI Copilot code assistance, native git integration and side-by-side visual differencing.

Gen2 Standard Warehouse, Adaptive Compute

Snowflake has made the second generation (Gen2) of its Standard Warehouse GA and moved its Adaptive Compute into private preview.

More platform-related private previews coming “soon” include external data in universal search for finding data in external relational databases such as PostgreSQL and MySQL and a Copilot for Horizon Catalog for governance and security questions through a Cortex AI-powered chat interface.

Coming to public preview are snapshots for immutable, point-in-time data backups. Now generally available are new AI observability tools. Coming to GA are catalog-linked databases and extensions to Snowflake’s Trust Center.

Standard Warehouse Gen2 offers hardware and software enhancements to speed up analytics performance, according to Snowflake. The goal is for customers to feel confident about future-proofing their systems as artificial intelligence continues to proliferate.