Developer Guide
FEATURE
SnowpipeStreaming
Stream large amounts of continuous data for near real-time insights with Snowpipe Streaming, supporting up to 10 GB/s per table, 5-second latency and a flat ingestion-based pricing model.
Go from ingestion to insights in seconds
Stream data directly to Snowflake with up to 10 GB/s throughput with 5-second latency. Automate your entire streaming pipeline with Snowflake CoCo.
Pay for data, not infrastructure
Spend less, predict more. One flat rate — 0.0037 credits per uncompressed GB ingested — eliminates surprises and hidden charges.
Build for production-ready AI
AI needs a continuous flow of current data. Get exactly-once semantics with ordering guarantees and automatic schema evolution so your models and agents always have what they need the moment they need it.
Benefits
High-performance streamingbuilt for the AI era
High-Performance Ingestion
Run pipelines faster with high-performance streaming
Stream up to 10 GB/s directly with 5-second latency from ingestion to queryable insights, and run queries 56% faster on average with a new Rust-based server-side SDK.
Replace hours of manual setup with a guided conversational workflow. Use CoCo to deliver AI-assisted configurations.
Govern every stream with unified security and data lineage via Snowflake Horizon Catalog so data is AI-ready the moment it lands.
Predictable Cost
Stream more for less with consumption-based pricing
Pay 0.0037 credits per uncompressed GB, a simple consumption-based rate with no hidden compute or infrastructure fees.
Cut client-side resource costs by 30% compared to Snowpipe Streaming Classic.
Eliminate infrastructure costs entirely with a fully serverless architecture. No brokers, no clusters, no compute to provision.
Avoid billing surprises. Scale throughput to your exact needs, with costs that track directly to data volume ingested.
AI Ready
Give your AI pipeline every feature it needs, built in
- Guarantee exactly-once delivery with ordering guarantees so no event is duplicated, dropped or delivered out of sequence — critical for CDC pipelines, financial data and context-aware AI agents.
- Deliver AI ready inputs to your models and agents without manual pipeline orchestration by pairing with Dynamic Tables to continuously transform and model data as it lands.
- Whether you're writing to Snowflake native tables, Snowflake-managed Iceberg tables or Interactive tables, you get flexible options to build pipelines at every speed.
Snowpipe Streaming High-Performance Architecture
Frequently Asked Questions
Find answers here about Snowflake Snowpipe Streaming High-Performance Architecture, and how it can streamline your data pipelines.
What is Snowflake Snowpipe Streaming High Performance Architecture?
Snowpipe Streaming High Performance Architecture is Snowflake's next-generation and serverless streaming ingest service. It supports up to 10 GB/s throughput per table, delivers data queryable in ~5 seconds, and uses flat-rate pricing at 0.0037 credits per GB.
How is the Snowpipe Streaming High Performance Architecture different from Snowpipe Streaming Classic?
Snowpipe Streaming High Performance Architecture introduces a Rust-based architecture that moves key processing server-side, significantly reducing client-side overhead. TPC-DS benchmarks show 56% faster query performance, and customers are reporting up to 30% lower client-side resource costs compared to Classic.
What SDKs and languages are supported?
Java, Python and REST are generally available today. We have more on the roadmap
How is it priced?
Volume-based at 0.0037 credits per GB ingested. It is identical to Snowpipe file ingestion pricing. You pay only for what you ingest. No infrastructure, cluster sizing or operational overhead.
How does Snowflake Snowpipe Streaming High-Performance Architecture support AI use cases?
It’s purpose-built for the real-time data demands of modern AI, including agentic AI applications that require a continuous flow of the latest information to make intelligent and autonomous decisions. Streaming data directly into Snowflake means AI agents, ML models, and personalization engines always reason over current reality, not stale batch exports.
Snowflake CoCo accelerates end-to-end development of streaming AI pipelines. Using natural language, data engineers can generate Snowpipe Streaming configurations, define Semantic Views to expose live data to Cortex Agents, and create Streamlit dashboards all from a single conversational interface. CoCo is deeply aware of Snowflake's data, compute, and governance semantics, reducing setup from hours of manual configuration to a guided workflow. Once pipelines are running, CoCo also helps teams diagnose issues, troubleshoot permissions and identify optimization opportunities across ingestion, transformation and serving.
Where is it available?
It’s generally available as of November 2025 across AWS, Azure and GCP commercial deployments.
How easy is it to migrate from Snowpipe Streaming Classic?
The migration process is easy. The architectural changes and API updates discussed here also apply to migrations to the Python SDK, because the high-performance architecture is available in both languages. To migrate your application, please see this guide.
