According to TheRegister.com, the data platform Snowflake is adding support for Google’s Gemini models to its Cortex AI service. The company’s VP of AI Engineering, Dwarak Rajagopal, stated this allows customers to run Gemini natively within their Snowflake data environment across all supported clouds, including AWS, Azure, and Google Cloud, via cross-region inference. The rollout is part of a phased, model-agnostic strategy, and Cortex AI already supports models from OpenAI, Anthropic, Meta, Mistral, and Deepseek. A public preview for Gemini 1.5 Flash and the upcoming Gemini 1.5 Pro through Cortex AI Functions began on Tuesday, with support for Snowflake Intelligence and Cortex Agents coming soon. Pricing follows Snowflake’s consumption model, where customers only pay for what they use. The goal is to let customers experiment and scale AI usage based on value, not sunk costs.
The Model-Agnostic Play
Here’s the thing: Snowflake isn’t betting on one AI horse. It’s building the entire stable. By adding Gemini to a roster that already includes Claude, GPT, and Llama, they’re executing a classic platform play. The promise is “model choice without compromise,” as Rajagopal put it. And for enterprises sitting on petabytes of sensitive data in Snowflake, that’s a powerful pitch. The big win isn’t just access to another model—it’s that you can query it with SQL or an API without ever moving your data. That addresses a huge chunk of the security and governance headaches that have slowed corporate AI adoption. Basically, they’re selling peace of mind as a feature.
Why This Matters For Enterprises
Look, every cloud provider has their own AI studio. But what if your data is spread across AWS, Azure, and Google Cloud? Or what if you just don’t want to be locked into one vendor’s model ecosystem? Snowflake is positioning itself as the neutral ground. You keep your data in one place (their place), and you can shop for the best model for each specific task. Need a cheap, fast model for simple summarization? Maybe you pick Gemini 1.5 Flash. Need deep reasoning on a complex financial document? Perhaps you route that to Gemini 1.5 Pro or Claude 3 Opus. This flexibility is huge. It turns AI from a monolithic tool into a modular component. And the pay-as-you-go pricing lowers the barrier to experiment, which is critical. How many AI projects die in pilot because the cost to scale is terrifying? Snowflake’s trying to fix that.
The Bigger Data Platform Battle
So what’s the real endgame? It’s about becoming the indispensable layer. Snowflake doesn’t necessarily care which model wins the LLM race; they just want to be the system of record and the processing engine for all your data, regardless of where it lives or which AI touches it. This move directly counters the efforts of the big three clouds (AWS with Bedrock, Azure with OpenAI, Google with Vertex AI) to keep data and AI processing within their own walls. Snowflake is saying, “Keep your data with us, and we’ll give you safe access to all of them.” It’s a clever wedge. But it also raises questions. Can they maintain performance parity with native cloud AI services? And will the consumption model lead to bill shock as these AI experiments scale? Only time will tell. For now, it’s a compelling option for companies that are truly multi-cloud or just plain vendor-averse.
