Snowflake Data Cloud

Cloud data platform that separates storage and compute for elastic warehousing.

Reviewed by 7wData

On this page

Publisher review

Snowflake is a cloud-based data warehouse that decouples storage and compute, running on AWS, Azure, and Google Cloud with elastic scaling. You provision virtual warehouses in 1- to 512-credit increments, run queries, then shut down—paying for storage monthly and compute by the second. For organizations with episodic or bursty analytics workloads, this model simplifies capacity planning compared to pre-sized traditional databases.

Since 2012, Snowflake has built a polished SQL engine optimized for traditional analytics. Features like Time Travel (query historical snapshots) and zero-copy cloning (instant table copies) remain competitively strong. Secure Data Sharing allows sharing a database with other Snowflake accounts without copying data—useful for multi-tenant SaaS or vendor distribution.

In 2025, Snowflake expanded into AI and lakehouses. Cortex AI provides multimodal LLM functions (summarization, translation, sentiment analysis) callable from SQL. Cortex Code is an agent that generates SQL, Python, and dbt from natural language. Native Apache Iceberg support lets Snowflake act as a lakehouse control plane for external data on S3, GCS, or Azure Blob—historically Snowflake required external tools for ingestion, transformation, and reverse ETL.

But structural weaknesses persist. First: cost. Usage-based billing is simple conceptually but hard to forecast. Oversized warehouses or unoptimized queries can 10x a bill in days. The median annual spend is ~$96,600, yet 30% of customers cite pricing concerns, and many forecast wrong by 200–300%. Second: Snowflake remains a warehouse, not a full stack. You still need Fivetran, dbt, and Hightouch. Third: Snowflake is cloud-only with shared multi-tenant infrastructure. A 2024 security incident affected 200+ customers, raising concerns for regulated industries. Finally, concurrency caps at ~8 users per warehouse, and row/column access controls are absent, limiting use in some multi-tenant designs.

Databricks, growing at 65% YoY, has closed the competitive gap by offering ETL, analytics, and AI in one platform. Snowflake's net revenue retention fell from 158% to 125%, signaling slower expansion. For pure SQL analytics at scale, Snowflake remains solid. For teams needing a consolidated stack, the calculus has shifted.

Get the AI & data signal, daily.

48k+ subscribers read this every morning. One email, both newsletters. Unsubscribe anytime.

How it works

  1. Virtual Warehouses

    Elastic compute clusters that scale from 1 to 512 credits per hour; suspend when idle; billed per second.

  2. Snowpark

    Python and Java SDK for ML and data engineering directly in the warehouse; runs transformations without extracting data.

  3. Cortex AI

    Suite of LLM functions (summarization, translation, sentiment) callable from SQL; includes Cortex Code for SQL/Python generation and Cortex Analyst for Q&A on structured data.

  4. Iceberg Tables

    Native Apache Iceberg support for open lakehouses; manages external data on S3, GCS, or Azure Blob with Snowflake-side governance.

  5. Time Travel & Zero-Copy Cloning

    Query historical versions of tables up to 90 days back; create instant table copies without data duplication.

  6. Secure Data Sharing

    Share databases across Snowflake accounts without copying data; consumers query your account's copy-on-write view.

  7. Native Apps

    Marketplace for distributing and monetizing Snowflake applications; includes role-based access and billing integration.

Strengths and trade-offs

Strengths

  • Polished SQL engine with strong optimizer for traditional OLAP; Time Travel and zero-copy cloning are mature and rarely matched by competitors.
  • Cross-cloud portability: run on AWS, Azure, or GCP with minimal code change; native Iceberg support bridges warehouse and lakehouse use cases.
  • Secure Data Sharing avoids data duplication for multi-tenant distribution; no need to copy sensitive data to consumers.

Trade-offs

  • Cost unpredictability is endemic: 30% of customers cite pricing concerns; companies routinely forecast 200–300% wrong due to unoptimized queries or oversized warehouses.
  • Requires multi-tool stack: ETL (Fivetran, Airbyte), transformation (dbt), reverse ETL (Hightouch), and visualization (Tableau, Looker)—Snowflake alone is not a complete platform.
  • Concurrency and access control limits: capped at ~8 concurrent users per warehouse; lacks fine-grained row/column access controls, limiting multi-tenant architectures; shared infrastructure creates security risk (2024 breach affected 200+ customers).

Pricing context

Snowflake uses consumption-based pricing: compute is billed in credits at $2–4 per credit on-demand (or $1.50–2.50 with annual commitments), storage at ~$40 per TB per month on-demand, and cloud services at ~10% of daily compute when exceeded. Three editions (Standard, Enterprise, Business Critical) charge 0–50% premiums over Standard rates. No free tier; pricing varies by cloud provider (AWS, Azure, GCP) and region.

The median customer pays ~$96,600 annually, though organizations committing to 100,000+ credits annually can negotiate 25–45% discounts. Typical small teams spend $500–2,000 monthly; medium businesses $2,000–10,000 monthly; enterprises $10,000–50,000+ monthly.

Getting started with Snowflake Data Cloud

  1. Sign up and choose configuration

    Create a Snowflake account and specify your cloud provider: AWS, Azure, or Google Cloud. Select your region and edition—Standard is adequate for trials; Enterprise adds higher concurrency. Snowflake will assign your initial compute warehouse.

  2. Provision virtual warehouse

    Configure your first warehouse by selecting a size: X-Small (1 credit per hour) suits evaluation; larger sizes scale up to 512 credits. You pay per second of compute time. Snowflake suspends idle warehouses automatically to minimize cost.

  3. Load or connect data

    Decide how to ingest data: upload a CSV, connect Fivetran or Airbyte for ongoing pipelines, or point Snowflake to S3/GCS objects via Iceberg tables. Start with sample data to learn the SQL interface before integrating production sources.

  4. Run your first query

    Open the Snowflake web console, write a SQL SELECT query against your loaded table, and execute it. Verify the data returns correctly and record the execution time. This baseline helps you identify performance issues later.

  5. Set up monitoring and schedules

    Enable Time Travel (snapshot history) to query past table states, and set cost alerts in your account settings. Schedule recurring transformations via dbt or SQL jobs, then monitor warehouse credit usage weekly to control spend against your budget.

Frequently Asked Questions

What is Snowflake Data Cloud?

Snowflake is a cloud-based data warehouse decoupling storage and compute across AWS, Azure, and Google Cloud. You pay monthly for storage and per-second for compute via virtual warehouses. This elastic model suits organizations with episodic analytics workloads better than traditional pre-sized databases.

How much does Snowflake cost?

Snowflake uses consumption-based pricing: compute costs $2–4 per credit on-demand ($1.50–2.50 annually), storage costs ~$40 per TB per month. The median customer pays ~$96,600 yearly. Pricing varies by cloud provider and edition, with three tiers (Standard, Enterprise, Business Critical) charging 0–50% premiums.

What are Snowflake's key features?

Snowflake's core features include Time Travel for querying historical snapshots up to 90 days back, zero-copy cloning for instant table copies, Secure Data Sharing for cross-account database access, and Cortex AI for LLM functions like summarization and translation callable from SQL.

Why is Snowflake cost unpredictable?

Snowflake's usage-based billing is simple conceptually but hard to forecast. Unoptimized queries and oversized virtual warehouses can multiply bills by 10x. Companies routinely forecast 200–300% incorrectly. Thirty percent of customers cite pricing concerns, and bill surprises stem from inefficient settings rather than data volume alone.

Does Snowflake include ETL and reverse ETL?

No. Snowflake is a warehouse, not a full-stack platform. You still need separate tools: Fivetran or Airbyte for ETL, dbt for transformation, Hightouch for reverse ETL, and Tableau or Looker for visualization. Databricks offers consolidated alternatives with built-in ETL, analytics, and AI.

Should I choose Snowflake or Databricks?

Choose Snowflake for pure SQL analytics at scale with mature Time Travel and cloning. Choose Databricks for unified platforms needing ETL, analytics, and AI integrated. Databricks grows 65% YoY with rising net revenue retention; Snowflake's NRR fell from 158% to 125%, signaling slower expansion.

Alternatives

Integrations

dbt Tableau Fivetran Airbyte

How Snowflake Data Cloud compares

Direct head-to-head against 3 competitors. Picked by 7wData.

This tool

Snowflake Data Cloud

Pricing
Snowflake uses consumption-based pricing: compute is billed in credits at $2–4 per credit on-demand (or $1.50–2.50 with annual commitments), storage at ~$40 per TB per month on-demand, and cloud services at ~10% of daily compute when exceeded. Three editions (Standard, Enterprise, Business Critical) charge 0–50% premiums over Standard rates. No free tier; pricing varies by cloud provider (AWS, Azure, GCP) and region. The median customer pays ~$96,600 annually, though organizations committing to 100,000+ credits annually can negotiate 25–45% discounts. Typical small teams spend $500–2,000 monthly; medium businesses $2,000–10,000 monthly; enterprises $10,000–50,000+ monthly.
Target
Snowflake is a cloud-based data warehouse that decouples storage and compute, running on AWS, Azure, and Google Cloud with elastic scaling.
Deployment
cloud
Strength
Polished SQL engine with strong optimizer for traditional OLAP; Time Travel and zero-copy cloning are mature and rarely matched by competitors.
Watch for
Cost unpredictability is endemic: 30% of customers cite pricing concerns; companies routinely forecast 200–300% wrong due to unoptimized queries or oversized warehouses.

Databricks Lakehouse

Pricing
DBU rates $0.15-$0.70/DBU by workload type; cloud VM, storage, and networking billed separately, typically adding 1-2x DBU cost.
Target
ML and data engineering teams wanting ETL, SQL analytics, and AI without assembling a separate multi-tool stack.
Deployment
SaaS on AWS, Azure, or GCP.
Strength
Single platform covers ETL, SQL analytics, and ML without external tools, the main reason teams switch away from Snowflake.
Watch for
Idle clusters accumulate cost silently; total bill equals DBU fees plus cloud infra, often 2-3x the DBU estimate.

Google BigQuery

Pricing
$6.25/TiB on-demand queries; $0.02/GB active storage, $0.01/GB long-term; 1 TiB free monthly; capacity slots $0.04/slot-hour.
Target
GCP-native organizations wanting serverless SQL analytics with no warehouse provisioning or cluster management.
Deployment
SaaS, GCP only.
Strength
Genuinely serverless: auto-scales to petabyte queries without provisioning clusters; no idle-compute billing between query runs.
Watch for
GCP-only deployment creates hard cloud lock-in; February 2026 multi-region storage transfer fees added billing surprises; on-demand pricing spikes on unoptimized queries.

Microsoft Fabric

Pricing
F2 capacity $262/month to F32 at $4,198/month; Power BI content creators add $10-$20/user/month; 41% discount with 1-3 year reservation.
Target
Microsoft 365 or Azure organizations consolidating BI, data engineering, and warehousing under one Azure capacity subscription.
Deployment
SaaS, Azure only.
Strength
OneLake unifies BI, pipelines, warehouse, and ML on one Azure storage layer, eliminating data copies between services.
Watch for
May 2026 global outage with delayed incident response exposed SaaS reliability gaps; access permissions inconsistent across Fabric workload types.

User reviews

No user reviews yet. Be the first to write one.

Sources

Reporting on this tool draws on these publicly available sources.

  1. www.revefi.com — Detailed pricing components: compute credits ($2–4 on-demand), storage ($40–45/TB/month), edition premiums, typical cost ranges by organization size.
  2. peliqan.io — Identified weaknesses: cost unpredictability (200–300% forecast errors), multi-tool stack requirement, lack of built-in ETL/reverse ETL, cloud-only deployment, lack of fine-grained access controls.
  3. mammoth.io — Credit costs by warehouse size (X-Small 1/hour through 2X-Large 512/hour), storage and Time Travel/Fail-safe costs, serverless feature charges (Snowpipe, Auto-clustering, Search Optimization).
  4. inquisitiveintellect.medium.com — Technical limitations: cloud-only SaaS model, concurrency cap (~8 users per warehouse), lacks row/column/label-level access controls, limited query optimization tools, no index creation, migration challenges due to lack of database constraints.
  5. www.cloudzero.com — Bill shock patterns: oversized warehouses, inefficient queries, unoptimized settings; median annual spend ~$96,600; capacity commitment discounts (25–45% for 100,000+ credits).
  6. qrvey.com — Competitive positioning: Databricks net revenue retention growing 65% YoY vs. Snowflake's NRR decline from 158% to 125%; market shift toward consolidated platforms.
  7. salimaouamri.medium.com — 2026 competitive dynamics: Snowflake and Databricks converging on features; Databricks gaining ground with full-stack (ETL, analytics, AI) offering; market segment shift toward unified platforms.