Snowflake Summit 26: Iceberg v3 generally available, Datastream and a $6 billion AWS commitment

2026-06-02 · Publisher : Snowflake · Domain : Cloud & DevOps

At Snowflake Summit 26 (1-4 June), Snowflake announces general availability of Apache Iceberg v3 and managed Iceberg storage, a preview of Datastream (Kafka-compatible streaming) and a $6 billion commitment with AWS.

### Background From 1 to 4 June 2026, **Snowflake Summit 26** gathered in San Francisco more than 20,000 expected attendees, according to the vendor's press release. On 2 June, at the platform keynote, Snowflake detailed its announcements. Analysts single out three areas: data openness (Iceberg), agentic AI and context governance. ### Key facts - **Apache Iceberg v3** generally available (deletion vectors, row lineage, multi-engine read and write access); **Snowflake Storage for Iceberg Tables** generally available; Horizon catalog based on Apache Polaris. - **Snowflake Datastream**: an Apache Kafka-compatible streaming service, reported as a private preview. - **Horizon Context** (a semantic governance layer for AI), **Cortex Training** (tuning open models on managed GPUs), **AI Agent Identity** (per-agent access control, generally available), **Adaptive Compute**. - Snowflake Intelligence is renamed **CoWork** and Cortex Code **CoCo**. - Before the Summit (27 May), Snowflake announced a **$6 billion, five-year commitment** with AWS, the acquisition of Natoma (an MCP platform) and an expanded partnership with Anthropic. ### Why it matters Iceberg v3 reaching general availability at Snowflake, then at Databricks two weeks later, marks the convergence of the two platforms on an open table format, which makes it easier to choose the engine without moving data. For IT leaders, the $6 billion commitment with AWS is a reminder that these platforms remain tied to a hyperscaler, with consequences for infrastructure costs and sovereignty. ### Takeaway Snowflake aligns its platform on the Iceberg format and agentic AI, and secures its compute capacity at AWS.

Tags : Données, Conférence, Open Source, AWS