How Snowflake Redefined Its Data Stack With an AI-First Strategy
At last week’s Snowflake Summit in San Francisco, CEO Sridhar Ramaswamy and SVP of Products Christian Kleinerman laid out a broad AI-first strategy that redefines Snowflake from an industry-leading data warehouse provider to a full-spectrum data and AI management platform capable of mobilizing business across silos and industries.
“This is the most significant technological moment we’ve seen in two decades,” Ramaswamy told tech reporters. “And we’re evolving to reflect that.” Snowflake’s mission? To empower every enterprise to achieve its full potential through intelligent use of data and AI, the CEO said.
“At a high level, our goal is to simplify and improve how organizations manage the entire data life cycle — from ingestion to transformation, and all the way to consumption,” SVP of Products Christian Kleinerman told The New Stack. “What’s significant this week is that we’re touching every part of that life cycle. We’re not just making incremental changes; we’re broadening what’s possible.”
Reimagining the Data Life Cycle
Ramaswamy reiterated the company’s holistic view of the data life cycle — from the moment data is created (a user login, a machine sensor signal) to structuring and enrichment, to analytics and decision-making. “Our platform now spans structured and unstructured data,” he said. “From raw ingestion to semantic interpretation and agentic action.”
AI, of course, is central to this evolution. Snowflake aims to make data directly accessible — not just to data engineers and analysts, but to business users and AI agents. AI, Ramaswamy said, is bridging the divide between structured databases and the vast troves of unstructured documents, media, and conversational history previously underutilized.
For example, Snowflake’s new AI capabilities can extract receipt totals from images, parse legal contract terms from PDFs, and analyze chat transcripts from tools such as Slack or Zoom — all within the same platform used for traditional SQL queries.
“We’re enabling AI models to reason across data sources — Drive, SharePoint, Box, databases — and deliver actionable insights,” Ramaswamy said. “It’s no longer about just asking what happened, but having an agent act on your behalf.”
Product Principles: Simplicity, Connectivity, Trust
Snowflake’s approach is rooted in three longstanding principles: simplicity, connectivity, and trust.
Even with strong underlying infrastructure, user experience must be intuitive, Ramaswamy said. “There’s nothing simple about crunching a petabyte of data, but we make it feel that way.” A two-line SQL query can drive immense computation behind the scenes.
Second, Snowflake is doubling down on data connectivity, both within organizations and across them. More than 10,000 active data-sharing relationships exist across Snowflake’s “stable edges,” enabling seamless collaboration between partners.
Finally, trust remains paramount. “Our customers serve hundreds of millions of users,” Ramaswamy said. “Ensuring data privacy and AI output verifiability is non-negotiable.” Snowflake’s AI products include citation-based responses and operate fully within the customer’s deployment boundary, guaranteeing data residency and compliance.
Introducing Adaptive Compute
Kleinerman unveiled a major evolution of Snowflake’s infrastructure: Adaptive Compute, a new model that replaces manual cluster sizing with intelligent policies. Rather than specifying XS, M, or L compute sizes, customers simply define query characteristics, and Snowflake handles optimal resource allocation dynamically. It doesn’t “suggest” anything; it just orders the right configuration.
This move builds on the company’s pioneering separation of storage and compute, first introduced a decade ago. Snowflake is now eliminating yet another layer of complexity.
“We’re simplifying operations while boosting price-performance,” Kleinerman said. Adaptive Compute, currently in private preview, has already shown promising results among early adopters.
Complementing this is Generation 2 Virtual Warehouses, which leverage the latest cloud instances and performance optimizations. For some workloads, Gen 2 provides up to four times the speed improvement over the previous generation, Kleinerman said.
Expanding Data Engineering With OpenFlow
In data ingestion and transformation, Snowflake debuted OpenFlow, a managed service for flexible, secure data movement. Built on its acquisition of Datafold, OpenFlow extracts and loads both structured and unstructured data from a wide array of sources — including SharePoint, Drive, Slack, and enterprise databases.
“AI’s value depends on access to the right data,” said Kleinerman. “OpenFlow reduces the friction in making unstructured content usable.”
OpenFlow supports both fully managed and BYOC (Bring Your Own Compute) deployments, giving customers control over execution location for compliance-sensitive workloads.
Making AI SQL-Native
Snowflake is also embedding AI directly into SQL workflows through AI SQL — a suite of new functions that allow users to call AI models from within standard queries.
The feature supports sentiment analysis, classification, and multimodal data processing (including PDFs, audio, and images). The goal is to enable analysts to use powerful AI capabilities with minimal training.
“It’s AI as a function call,” said Kleinerman. “This makes AI accessible to millions of existing SQL users.”
Automating Migration With SnowConvert AI
Snowflake also announced SnowConvert AI, a next-generation migration assistant. Building on its original SnowConvert tool, the new version uses AI to handle code translation errors, verify logic, and validate data fidelity.
“Migration is often blocked by the last few percent of code that fails to translate,” Kleinerman noted. “AI closes that gap.”
The tool includes AI-powered data validation and code testing, simplifying one of the most complex steps in modernizing legacy data infrastructure.
Snowflake Intelligence: The AI Interface for Everyone
The flagship announcement of the day was Snowflake Intelligence, a conversational interface that allows business users to query their organization’s data using natural language and securely governed AI agents.
Demonstrated live by Director of Product Jeff Hollen, the system allows users to ask questions like “How are sales tracking against forecast in the last two weeks?” and receive visual, citation-backed answers in seconds.
Users can follow up with deeper queries, such as “Why is the West region underperforming?” Snowflake Intelligence will pull structured and unstructured data, analyze it, and surface a narrative. Agents can even take actions, such as drafting and sending emails based on the insights, Hollen said.
Importantly, all permissions and governance policies from the user’s Snowflake account are respected automatically, he said.
Cortex: Making Data AI-Ready
To support these new capabilities, Snowflake is launching several services under its Cortex AI brand:
- Cortex Analyst: Translates natural language to SQL and provides answers from structured data.
- Cortex Knowledge Extensions: Makes unstructured content — such as news, contracts, or customer notes — AI queryable.
- Semantic Views: Adds business context to structured models, improving the accuracy and trust of AI responses.
Snowflake is also working with publishers such as Gannett (USA Today) to bring AI-ready content to the Snowflake Marketplace, enabling richer hybrid data applications.
The ROI of Agentic AI
When asked about Snowflake’s path to profitability and the role of AI, Ramaswamy said: “You should only launch ROI-positive products.”
While Snowflake has crossed the $1 billion annual revenue mark, the company is also investing aggressively in AI to position itself for long-term platform dominance. Stock-based compensation and operating expenses are being reined in, with plans to scale profitably.
“We’re using AI internally to drive efficiency,” Ramaswamy said, “from sales enablement to product development. The same agents we offer our customers, we use ourselves.
“AI is not just for coders. Just like blogs democratized publishing, Snowflake Intelligence democratizes enterprise insights. We’re not just building tools. We’re helping our customers do more with their data.”
