The data platform Snowflake is putting Google's Gemini to work inside its Cortex AI, aiming to give customers access to a foundational model within the boundaries of their data environment across supported clouds, the company told The Register.

Customers use Snowflake to store nearly any operational data, including financial records such as ERP data, CRM records including tickets and transactions, IoT and sensor data, API calls and responses, website user activity, and even content such as PDFs, document files, images, videos, and audio. Cortex AI allows users to pick the model of their choice, now including Gemini, to process that glacier of information using an LLM to get business insights.

“By running Gemini natively in Snowflake, customers can use Gemini models across all supported clouds via cross-region inference, regardless of whether their Snowflake environments run on AWS, Azure, or Google Cloud,” Dwarak Rajagopal, VP of AI Engineering and Research at Snowflake, told The Register via email.

Snowflake’s phased rollout is part of its AI model-agnostic strategy, which he said should ensure customers can use best-in-class models across the Cortex stack without re-architecting data or workflows.

“Our philosophy is model choice without compromise – customers shouldn’t have to trade off innovation, governance, or simplicity to use the best models available,” Rajagopal said.

Cortex AI already supports OpenAI, Anthropic, Deepseek, Meta, and Mistral. With Cortex AI, users can access those models directly in SQL or via APIs, analyze multimodal data, and then build AI use cases.

• Snowflake update caused a blizzard of failures worldwide
• Snowflake goes all out to woo PostgreSQL developers with lakehouse extensions
• Snowflake jumps on agentic AI train with Anthropic tie-up
• Snowflake CISO on the power of 'shared destiny' and 'yes and'

While the company is not disclosing pricing, it's apparently consistent with Snowflake’s “broader consumption model,” Rajagopal said.

“Customers only pay for what they actually use, which lowers risk, encourages innovation, and makes it easier to move AI from experimentation into production with confidence,” he said, adding that Snowflake wants customers to get to work inside the platform “to start small, experiment freely and scale AI usage based on real value” rather than sunk costs.

Snowflake customers can begin using a public preview of Gemini 3 Pro and Gemini 2.5 Flash through Cortex AI Functions as of Tuesday. Support is coming soon for Snowflake Intelligence and Snowflake Cortex Agents, which customers use to build AI agents that tap into company data. ®