As a leading expert in business intelligence and data science, Chloe Maraina has a unique vantage point on the seismic shifts occurring in the enterprise data landscape. With a passion for translating complex data into compelling visual stories, she has closely followed the race to integrate powerful AI capabilities directly into the platforms where enterprise data lives. We sat down with her to discuss the recent, significant expansion of the partnership between Snowflake and Google Cloud, exploring what the native integration of the Gemini 3 model means for developers, the competitive landscape, and the future of the AI data cloud.
The announcement emphasizes that making Gemini 3 “natively available” in Cortex AI is a substantial development. Could you paint a picture for us of the challenges developers were facing with non-native models and how this new integration changes their day-to-day workflow?
Absolutely. You have to imagine the friction developers were dealing with before. To use a powerful external model like Gemini, they were forced into this complicated, multi-step dance. First, they’d have to manage API keys and set up connections, which is always a headache. Then came the real problem: data movement. To feed the model, they had to pull data out of their secure, governed Snowflake environment. Suddenly, you have compliance and security teams breathing down your neck, asking about data residency, encryption, and potential exposure. A client I know of even tried this with other LLMs and ran into a wall of issues with data security and unacceptable latency. It just wasn’t production-ready. Now, with native integration, that entire painful process vanishes. The model lives inside the architecture. A developer can now build an application and call Gemini as if it’s a simple function, all within Cortex AI. The data never leaves the platform. This means development cycles are drastically faster, security is inherent, and you get the performance needed for real-world applications without all the infrastructure gymnastics.
Snowflake has been deliberate in its choice of partners, integrating Anthropic’s Claude LLM before this latest move with Google’s Gemini. Can you give us some insight into the strategic thinking behind which models get this native integration, and what criteria are used to gauge a model’s value for enterprise customers?
The strategy is really a blend of market demand and rigorous performance evaluation. First and foremost, Snowflake is listening intently to its customers. The decision to integrate Gemini was directly driven by customer feedback; they saw a huge appetite for using Google’s highest-performing models directly on their enterprise data. But it’s not just about popularity. The team evaluates these models based on hard metrics and benchmark testing. They look at things like reasoning ability, accuracy, and overall performance to ensure they are offering truly state-of-the-art tools. The goal isn’t to have the longest list of available models, but to provide a curated, powerful selection. They want to give customers the choice and interoperability to use the best tool for the job, whether it’s from Anthropic, Google, Meta, or another leader, all within that trusted, secure environment.
This deeper alliance with Google Cloud could certainly be seen as a strategic move to apply pressure on other major cloud providers. From your perspective, what are the most critical capabilities, like catalog unification or integrated AI services, that Snowflake should pursue with partners like AWS and Microsoft to cement its position as a truly cloud-agnostic platform?
That’s a fantastic point. This move doesn’t weaken Snowflake’s cloud-agnostic stance; it actually strengthens it by setting a new standard for partnership. It gives them leverage and a template for what a deep integration should look like. To maintain that advantage across all hyperscalers, the most critical capability is creating a seamless, unified experience. Catalog unification is at the top of that list. Imagine having a single, authoritative view of all your data assets, regardless of which cloud they’re stored in. That’s a game-changer. The second piece is tighter marketplace and AI service integration. Customers want to be able to procure and deploy solutions from AWS or Microsoft through their existing Snowflake workflows without feeling like they’re switching between disconnected worlds. By pushing for these deeper integrations, Snowflake ensures that it remains the central, indispensable hub for data workloads, making the choice of underlying cloud provider a secondary, tactical decision for their customers.
Snowflake was perceived as being a bit slower to enter the generative AI race initially. Since the leadership change in February 2024, what foundational platform shifts, aside from just adding more models, have been most pivotal in transforming it into a competitive AI data cloud?
It’s true they were more measured at the start, but that period of observation allowed them to be incredibly deliberate and aggressive when they did make their move. The shift has been far more profound than just plugging in third-party models. The most crucial change has been the deep investment in building out their own comprehensive development suite and infusing AI into the core of the platform itself. Look at the development of Snowflake Intelligence, an agent that lets users query both structured and unstructured data with natural language. That’s not a superficial add-on; it’s a fundamental rethinking of how users interact with data. This commitment to building a robust environment for both generative and agentic AI is what makes the platform truly competitive. We see proof of this with complex customers like BlackLine, a company that lives and breathes financial automation. For them to commit to this ecosystem is a massive endorsement. It shows the platform is now robust enough to handle the most intricate automation roadmaps, which goes far beyond simple AI model access.
What is your forecast for the evolution of data cloud platforms over the next five years?
I believe we’re moving rapidly from general-purpose platforms to highly specialized, industry-specific AI ecosystems. In five years, it won’t be enough to just offer a generic data cloud. The winning platforms will be those that provide tailored, vertical-specific solutions—think pre-built agents for financial compliance, or specialized models for pharmaceutical research. This will require an even deeper reliance on a rich partner ecosystem to provide those complementary, industry-specific capabilities. The focus will be on abstracting away complexity and enabling businesses to move from experimentation to production with measurable impact, faster than ever before. The future is about providing not just model choice and performance, but also the trust, simplicity, and deep industry context needed to truly transform a business with AI.
