Snowflake locks in OpenAI for $200M - GPT-5.2 runs natively on customer data
Snowflake announced a $200M multi-year partnership with OpenAI on February 2, bringing GPT-5.2 and other OpenAI models directly into its Cortex AI platform and Snowflake Intelligence agent. The pitch: run frontier models on your data without moving it outside your Snowflake environment.
For enterprises, this addresses the data sovereignty problem that's blocked many AI implementations. Customer data stays within Snowflake's infrastructure across AWS, Azure, and Google Cloud. You query it via SQL, analyze multimodal content (text, images, audio), and build agents without creating new data egress paths for security teams to worry about.
What's actually new
The technical integration matters more than the dollar figure. OpenAI's Apps SDK and AgentKit now work natively with Snowflake's governance layer, meaning you can build AI agents that act on governed data without custom plumbing. Snowflake claims 99.99% uptime for these integrations - we'll see how that holds under production load.
The deal is mutual: OpenAI uses Snowflake for experiment tracking and analytics, while Snowflake runs ChatGPT Enterprise internally. Both CEOs emphasize "agentic AI on proprietary data" as the unlock for ROI.
The implementation question
For data teams evaluating this against direct OpenAI API integration or alternatives like Tableau with embedded AI, the trade-offs are clear: you're trading some flexibility for built-in governance and zero data movement. Cost per query will depend heavily on workload - Snowflake's compute pricing layered on model inference isn't simple math.
Snowflake positions itself as model-agnostic (it already works with Anthropic), but this deal makes OpenAI's models the obvious first choice for its 12,600 customers. If you're a CIO trying to let analysts run AI queries without creating compliance headaches, this is designed for you.
The real test: how many production deployments ship in the next six months, and what the actual cost looks like at scale.