Snowflake and OpenAI ink $200M enterprise pact
OpenAI and Snowflake just teamed up in a multi-year, $200 million deal to bring advanced AI models straight into enterprise data. Companies can now query OpenAI models through SQL or natural language inside Cortex AI to build custom agents without moving their data. Why it matters: It bridges the gap between frontier intelligence and secure enterprise databases.

If you thought enterprise software giants were ready to pick a single AI winner, Snowflake just proved everyone wrong.
On Monday, the cloud data warehouse giant announced a multi-year, $200 million partnership with OpenAI. The deal integrates OpenAI frontier models, including GPT-5.2, directly into Snowflake Cortex AI and Snowflake Intelligence across all three major hyperscaler clouds.
That gives Snowflake's 12,600 enterprise customers the ability to build agents and query unstructured data right where their corporate assets already live.
Calling frontier models directly from SQL
This is not just another API reseller deal. The implementation goes straight into everyday database workflows.
Using Snowflake Cortex AI Functions, technical teams can now invoke OpenAI models directly inside standard SQL queries. Analysts can run inference on rows and columns, documents, images, and audio without transferring sensitive data outside their existing governance boundary.
For non-technical employees, Snowflake Intelligence enables plain-English questions that trigger automated data retrieval and analysis without code. Brands like Canva and WHOOP are already using these capabilities to build operational agents. Moving forward, both teams plan to integrate OpenAI's Apps SDK and AgentKit, while Snowflake rolls out ChatGPT Enterprise internally.
Two checks, two competitors, two months
If this announcement feels familiar, check your calendar. Just two months ago in early December, Snowflake signed an identical $200 million enterprise pact with Anthropic.
Snowflake CEO Sridhar Ramaswamy used nearly identical talking points for both announcements. Snowflake AI VP Baris Gultekin confirmed the company remains intentionally model-agnostic, choosing not to lock clients into any single vendor. Anthropic, OpenAI, Google, and Meta all sit on the same menu.
Workflow giant ServiceNow pulled the exact same maneuver in January, inking multi-year agreements with both OpenAI and Anthropic. While VCs squabble over market share surveys—Menlo Ventures claimed Anthropic is ahead, while Andreessen Horowitz pegged OpenAI as the enterprise leader—enterprises are treating AI providers like rideshare apps, swapping between models depending on the route.
Why it matters
For enterprise builders, data gravity wins every time. Moving regulated corporate data out to third-party endpoints is a legal nightmare. Bringing GPT-5.2 into governed SQL pipelines removes that friction entirely. If your company already runs on Snowflake, testing these agents now is an obvious low-risk move.
For the frontier AI labs, this deal signals the death of platform exclusivity. Neither OpenAI nor Anthropic can lock down the enterprise tier alone. Enterprise buyers demand model redundancy, forcing AI labs to compete side-by-side inside the customer's existing data stack.
We think Snowflake's positioning is brilliant. By staying strictly neutral, Snowflake turns the brutal LLM price war into a revenue driver for its own compute and storage platform. The risk, however, belongs to the model makers. When switching between OpenAI and Claude requires little more than editing a parameter in a SQL query, brand loyalty in enterprise AI looks thinner than ever.
Sources (2)
- openai.com(opens in a new tab)Originalopenai.com
- techcrunch.com(opens in a new tab)Reporttechcrunch.com