Snowflake✓

Trusted healthcare AI starts with knowing where your data came from. 🩺 Snowflake's Alistair Hankin joins DCN to share why provenance and patient consent matter as much as the model, and how governed AI keeps sensitive patient data secure. Watch here 👉

What’s keeping you company at #SnowflakeExpedition 2026? We’re thinking coffee, a few snacks and a favorite desk buddy. Join us virtually Nov. 3–5, and tell us what’s on your checklist. Save your spot:

Keep your standard tables. Route your interactive queries to a different warehouse. Snowflake Interactive Warehouses let you serve latency-sensitive workloads from existing tables without creating a duplicate serving table. Snowflake Data Superhero Alexander Murugesan walks through proactive cache warming, fallback warehouse configuration and a selective-query benchmark on 150 million rows. Building dashboards, APIs or repetitive application lookups? See the setup and what determines the latency benefit:

68 minutes vs. 316 minutes. In an independent, hands-on evaluation, @Signal_65 found Snowflake CoCo completed a seven-phase enterprise workflow in 68 minutes, compared to 316 minutes via manual workflow. That’s 4.6x faster. See the full evaluation:

The riskiest agent behavior may not appear in the prompt. Dangerous Activity Detection in Cortex AI Gateway evaluates the full trajectory: objectives, plans, tool calls, results and effects, with provenance preserved for review. See how it works:

How far behind is your recovery environment when you need it? Snowflake’s Optimized Refresh and RPO Assurance are now GA. Control your refresh schedule with Optimized Refresh, or use RPO Assurance for continuous managed refreshes and priority access to replication resources. For eligible failover groups, RPO Assurance provides a financially backed 30-minute recovery point objective (RPO), 99.9% of the time, subject to service terms and operating conditions. That’s replication lag, not time to restore service. See the setup and onboarding requirements:

Your support team finds out which tickets are on fire one ticket at a time. This Snowflake pipeline turns emails, forms, and calls into one -1 to +1 sentiment score with AI_COMPLETE, and a Dynamic Table keeps it current. Build it:

We're bringing @OpenAI's GPT-6.1 Terra to Snowflake Cortex AI. It's available now in public preview on Cortex Inference and AI Gateway, and coming soon to Cortex AI Functions, Snowflake CoCo, Snowflake CoWork and Cortex Agents. More model choice for visual reasoning, computer use and everyday writing, right next to your governed data.

Test a new model version on live traffic before making it the default. See how Snowflake Gateway traffic splitting and monitoring let you compare model services behind one persistent client endpoint:

Automated Snowpipe setup is now GA.⚡️ Continuously ingest data from s3://my-bucket/clickstream/ That's the whole setup. Snowflake CoCo builds the Snowpipe, checks the first loads and leaves it running.

We’re expanding our partnership with @Oracle. When available, Snowflake on OCI will be a native, first-party offering that brings Snowflake’s AI Data Cloud to Oracle Cloud Infrastructure (OCI). The offering would bring analytics and AI closer to Oracle data hosted on OCI, with purchasing and billing directly through Oracle. Explore what’s planned and what it could mean for your data and AI workloads:

Eliminate training-serving skew in ML with integrated feature and model serving. Our quickstart walks through an real-time XGBoost fraud detection pipeline with Snowflake’s Online Feature Store. The guide 👉

Wake up with a demo idea. Build it in 30 minutes. Present it to a customer an hour later. That’s just another day at Snowflake for Solution Engineer Cathy Xu, who found that ownership, continuous learning, and the right AI tools completely change what’s possible from Day 1 – both for your career and Snowflake customers. See how she’s building the agentic enterprise from Sydney:

Starting the day with a summary pulled from more than 40 data sources is how our CFO Brian Robins uses Snowflake inside Snowflake. At #SnowflakeExpedition you'll get the hands-on workshops and expert guidance to build your own path to value. Save your spot 👉

41.8 GB/sec. 133.5 million rows/sec. One native Snowflake table. Our synthetic Snowpipe Streaming benchmark in AWS us-west-2 sustained 41.8 GB/sec of uncompressed throughput with Elastic Channels. Throughput grew linearly as we scaled from 15 to 455 Kubernetes workers. The setup combined Elastic Channels, 1-vCPU worker pods and Rust-based encoding to reduce client-side overhead. Building telemetry or log pipelines? See the benchmark method, latency measurements and how to start with one worker:

The most intelligent bank starts with its people. @Westpac's Andrew McMullan shares how a modern data foundation and an upskilled workforce are driving its AI push. 95% of its people now use AI tools.

A few changed rows shouldn’t mean rescanning an entire group to update a SUM. Snowflake Dynamic Tables now optimize qualifying top-level aggregates using stored state and incoming changes. We explain the mechanism, including why deleting a MAX value is harder:

Our sales team has asked its AI assistant more than 2 million questions, and it's changing how we sell. 🚀 Anahita Tafvizi and Jonathan Beaulier join Ground Level AI to share where agents add value and where people still make the call. Listen here 👉

Your gold layer doesn’t always need a second serving table. Snowflake’s zero-copy Interactive Analytics path lets an Interactive Warehouse query existing standard or Iceberg tables directly. Snowflake Data Superhero Raj Thakkar explains the setup and tradeoffs:

Still branching your schema scripts based on whether a table exists? Snowflake’s CREATE OR ALTER TABLE creates a missing table or applies supported changes to an existing one. Rerun the unchanged definition, and it’s a no-op. We walk through adding columns while retaining existing rows, plus the drops and renames you need to review carefully. Check it out: