For companies built on data distribution, like providers sharing data with consumers using platforms like Snowflake, Databricks, and BigQuery, these egress fees quickly become the single largest operational expense.
The solution? Embracing no-egress, high-performance object storage that works seamlessly with your existing modern data architecture.
The Rude Awakening: Why Traditional Cloud Storage Is Data’s “Hotel California”
Amplify, a company that provides a white-label solution allowing SaaS providers to offer data access features to their customers, experienced this cost shock firsthand. They initially stored file-based data (like CSV exports) on Amazon S3.
When their startup credits ran out, they realized that egress costs were consuming their budget. As one speaker noted, Amazon allows you to check in your data, but they never want it to leave.
Consider the sheer expense of data distribution:
- High Cost Per Gigabyte: Amazon S3 charges approximately 9 cents per gigabyte downloaded (egress).
- Scale Multiplies the Pain: When distributing a single 2-terabyte data product, that single download costs roughly $180.
- Costs Dwarf Storage: For a company storing 10 terabytes of data, the storage cost might be manageable (e.g., $230/month). But if that 10TB dataset is downloaded only three times in a month, the egress fees can exceed $2,800—more than 10 times the storage cost.
This model severely penalizes a company’s ability to democratize data and meet customer demand for recurring data consumption.
Integrating the Data Distribution Layer: Snowflake, BigQuery, and Databricks
Amplify recognized that to solve the one-to-many and one-to-any data sharing problem, they needed a robust hub. They chose Snowflake as their core data storage and transformation layer due to its cross-cloud compatibility, dynamic scalability, and leading position among OLAP (Online Analytical Processing) databases.
Snowflake is central, but external object storage is still required for the “last mile” of delivery, especially for file-based outputs like CSVs, or for data transfers to other environments. This is where services like BigQuery, Databricks, Redshift, and Synapse come into play—all requiring efficient, cost-effective data staging.
Amplify’s customers want data delivered directly to their own analytics platforms, whether that’s a native Snowflake share or a transfer to a BigQuery environment. This flexibility means that the underlying object storage must be high-performance, S3-compatible, and open.
The Solution: Seamless Integration with Backblaze B2
In search of an S3-compatible, cost-effective alternative, Amplify selected Backblaze B2. The philosophical approach of B2 aligns with the “Open Cloud.” pushing back against egress fees and encouraging the use of best-of-breed tools.
The financial shift was dramatic:
- Free Egress Buffer: B2 offers free egress up to three times the amount of data stored each month.
- Affordable Overage: If the consumption exceeds that free allowance, the egress rate is only a penny per gigabyte, which is roughly one-ninth (1/9th) the cost of the standard AWS egress fee.
For Amplify, this switch transformed a cost structure where distribution fees were astronomical into one where subsequent months involved minimal egress charges, dropping their costs substantially. Backblaze also offers unlimited egress when integrating with specific CDN and compute partners like Fastly, BunnyNet, Cloudflare, Vultr, and CoreWeave.
Technical Harmony: B2 as an External Stage
The migration from S3 to B2 was surprisingly quick and painless, taking only about two weeks from idea to production implementation. This speed was possible due to Backblaze B2’s S3 compatibility .
Key technical benefits of this integration include:
- Snowflake External Stage: B2 integrates directly with Snowflake as an external stage. Enabling this only required a single, one-day support ticket with Snowflake.
- Fast Data Unloading: Snowflake uses the fast and scalable COPY INTO command to unload tables into B2 storage.
- Format Flexibility: B2 supports exporting data in multiple formats, including CSV, JSON lines, and Parquet. Parquet is especially crucial for faster transfer and loading when moving data from Snowflake to external platforms like BigQuery.
Amplify was able to transition data sets in batches by simply regenerating the files from Snowflake and deleting the expensive files from S3, minimizing disruption (customers reportedly didn’t even notice the change).
Beyond the Basics: Cost Savings Enable Innovation
The savings generated by moving to a no-egress model directly fuels business growth and innovation. Instead of spending thousands of dollars monthly on bandwidth, Amplify redirected those funds toward development, including expanding their offerings to support BI/analytics environments like PowerBI, Tableau, and Looker. They are also exploring becoming a Snowflake native application.
If you want to see the platform in action, Amplify offers an interactive live demo and a free tier to get started. For more details on the technology, visit amplifidata.com.
By eliminating punitive egress fees, organizations gain operational freedom to serve their customers better, knowing that every download doesn’t chip away at their bottom line. If you are currently leveraging Snowflake, BigQuery, Databricks, or any modern data platform that frequently distributes data to consumers, it is time to reassess your storage provider and ensure you are not paying the download tax. The cost savings are immediate and the technical integration is seamless.
