Snowflake

Overview

Snowflake is a cloud data warehouse sold as a managed service on AWS, Azure and Google Cloud. It describes its own architecture as a hybrid of shared-disk and shared-nothing designs, split into three layers: storage, query processing and cloud services.

Storage is columnar and compressed, divided into micro-partitions that Snowflake manages for you. Compute is bought as virtual warehouses, independent clusters sized XS to 6XL that suspend when idle and resume on demand. Nothing is shared between them, so one team's heavy query cannot slow another team's.

That separation is the thing to model when comparing cost. Compute is billed per second after a one-minute minimum, storage per terabyte a month, so a suspended warehouse costs nothing at all.

Editions matter more than they first appear. Time Travel is one day on Standard but up to ninety on Enterprise and above, and HIPAA, PCI DSS and HITRUST only arrive at Business Critical.

Openness now runs through Iceberg. Tables can be Snowflake-managed or governed by an external catalogue such as AWS Glue or any Iceberg REST catalogue, with the data in storage you own, and Delta files can be read.

Governance is packaged as Horizon Catalog: masking and row policies that follow data into external tables and shares, automatic classification, column lineage and a Trust Center for misconfiguration.

The AI layer is Cortex AISQL, a set of SQL functions calling models from OpenAI, Anthropic, Meta, Mistral AI and DeepSeek, billed in separate AI credits rather than warehouse credits.

Key features and capabilities

The same headings are used for every cloud data warehouses entry, so two tools can be read side by side.

Architecture and storage

• Storage, query processing and cloud services are separate layers that scale independently
• Columnar compressed storage in micro-partitions, managed for you
• Virtual warehouses from XS to 6XL, each an independent cluster that shares nothing
• Multi-cluster warehouses on Enterprise scale out extra clusters for concurrency

Performance and scaling

• Credits double with each size, from 1 an hour at XS to 512 at 6XL
• Per-second billing after a one-minute minimum, with auto-suspend and auto-resume on by default
• Materialised views and search optimisation are gated by edition
• Adaptive warehouses let Snowflake choose the configuration instead of you

SQL and workloads

• Snowflake SQL on virtual warehouses, with Snowpark for code execution
• Time Travel: one day on Standard, up to 90 days on Enterprise and above
• Serverless ingestion through Snowpipe and Snowflake Openflow
• Secure sharing of governed data products between accounts

Open formats and sharing

• Iceberg tables either Snowflake-managed or governed by an external catalogue
• External catalogues include AWS Glue, any Iceberg REST catalogue and Snowflake Open Catalog
• Writes are supported to externally managed REST-catalogue tables, and Delta files can be read
• External volumes keep the data in your own S3, Azure or Google Cloud storage
• Egress is charged, free in-region, $20 a terabyte across regions and $90 to another cloud

Security and governance

• Role-based access control enforced across every access path
• Masking and row-access policies follow data into external Iceberg tables and shares
• Automatic sensitive-column classification, column lineage and data quality monitoring
• Compliance by edition: SOC 2 throughout, HIPAA, HITRUST and PCI DSS at Business Critical
• Government workloads run in separate SnowGov regions

AI and machine learning

• Cortex AISQL adds SQL functions for completion, classification, extraction and transcription
• Embedding and similarity functions for semantic search
• Models from OpenAI, Anthropic, Meta, Mistral AI and DeepSeek
• AI usage is billed in separate AI credits, metered by token or message

Where it runs and what it costs

• AWS, Azure and Google Cloud across the Americas, EMEA and Asia Pacific
• SnowGov on AWS GovCloud and Azure Government; China through an authorised partner
• No self-hosted option
• Credit and storage prices vary by region as well as edition

Pricing

Usage-basedStandard from $2.00 a credit

On-demand credits in AWS US East run $2.00 on Standard, $3.00 Enterprise, $4.00 Business Critical and $6.00 for Virtual Private Snowflake, with storage at $23 a terabyte a month and higher rates elsewhere. The cloud services layer costs 4.4 credits an hour but is waived below 10% of daily warehouse credits. Serverless features bill at one credit per compute-hour times a multiplier, and AI credits are $2.00. A 30-day trial needs no card, though the value of its balance is not published.

Vendor pricing page →

Demos and videos

• What is Snowflake? An eight-minute demo Product demo · YouTube
• Simplify AI for your data pipelines with Cortex AISQL Webinar or talk · YouTube

About Snowflake

Snowflake Inc. is a public company listed on the New York Stock Exchange as SNOW, with its head office in Menlo Park, California and co-founders Benoit Dageville and Thierry Cruanes still named on its leadership page. It reported product revenue of $1.23bn for its 2026 financial year, up 30% year on year, with 12,062 customers and more than 10,000 employees. There is no open-source or self-hosted edition; the platform is proprietary and priced in consumption credits, with openness delivered through Iceberg rather than the engine.

Founded 2012 · Menlo Park, California · snowflake.com