David Energy had nearly a year of free data pipeline services and decided to pay somebody else. In the familiar arithmetic of software purchasing, this looks peculiar. Free is supposed to be the winning number. But a pipeline has another account, one that rarely appears on an invoice: the attention required to keep it working.
- Estuary moves database changes, streams, and batch data through one managed platform.
- Its distinguishing idea is to capture data once and reuse it across destinations.
- The practical test is total cost and useful freshness, including the engineer’s time.
The free pipeline with an expensive habit
David Energy needed a PostgreSQL-to-Snowflake connection for analytics. According to Estuary’s published account, Fivetran worked, but its bill reached eight times the cost of the company’s other infrastructure combined. Airbyte came next. Bugs brought compensating credits, then more bugs brought more credits. Eventually, the team preferred a paid connection that needed less nursing.
CTO Sam Strasser described Estuary as “it was just a better technical approach.” The reported appeal included dependable replication, straightforward setup, and responsive support. This is a customer story published by the seller, so it establishes a particular experience rather than a verdict on every competing product. Still, the sequence is revealing: price failed first; reliability changed the buying decision.
Estuary occupies this rather unromantic corner of enterprise software. It transports information between the systems where a business creates it and the systems where people analyse or act on it. Its customers include energy providers, workplace software companies, manufacturers, and consumer brands. The audience is the data team that would like to spend Tuesday doing something besides repairing Monday.
One capture, several clocks
The product, also known as Estuary Flow, combines change data capture, streaming, and scheduled batch movement. CDC reads database changes rather than repeatedly copying everything. SaaS applications and files can arrive on different schedules. A warehouse, an operational application, and an AI workflow can consume the resulting data at different speeds.
Underneath is a useful separation. Captures bring records in. Collections retain them as durable, structured data. Materializations deliver them to destinations. Transformations can reshape those records along the way. The destination does not have to be decided forever at the moment the source is connected.
changes + history
retain + reuse
Application
AI workflow
That design matters when a team adds another consumer. Reusing retained data can avoid another independent extraction system and another conversation about how to recover its history. Estuary’s managed service also takes on infrastructure work that a team assembling its own streaming stack would otherwise inherit.
The platform offers more than 200 managed connectors and supports transformations in SQL, TypeScript, and Python. Public cloud suits teams seeking a hosted service; private and bring-your-own-cloud options address teams needing control over the data plane. These are consequential choices for regulated workloads, where the location of processing belongs in the purchasing conversation.
The bill has two columns
Estuary charges for data volume and connector instances. Its published Cloud rate is $0.50 per GB, plus $100 monthly for each of the first six instances; additional instances cost $50 each. An instance connects a source or destination. Billable volume covers sourced, transformed, and delivered data, so source size alone is an incomplete estimate.
For a simple illustration, 100 billable GB and two connector instances imply $250 per month at the published rates. Warehouse charges, storage, and other infrastructure still belong in the budget. A free Developer tier allows 10 GB monthly and two concurrent instances; enterprise contracts handle private deployments and negotiated requirements.
The larger savings stories involve architecture too. Prodege paired Estuary replication with Apache Iceberg tables on S3 and Starburst Galaxy for transformations. The published case reports 60% lower replication costs and an estimated 30% reduction in Snowflake ingestion expenses. Those figures describe a combined migration; attributing the entire result to a connector would be convenient arithmetic.
A year spent on the engine
Estuary’s technical ancestry starts before its 2019 founding. Gazette began inside Arbor in 2014 as a streaming framework. That history helps explain why the company’s expertise lies in distributed data movement, rather than merely arranging attractive boxes on a dashboard.
Co-founders David Yaffe and Johnny Graettinger turned that foundation into a managed business. In October 2025, Estuary announced a $17 million Series A led by M13, with FirstMark and Operator Partners participating. The investment thesis linked dependable enterprise data movement to analytics, operations, and AI.
In August 2026, Yaffe described a year spent rebuilding the runtime. The transaction loop moved from Go to Rust, staging moved from memory to local disk, and coordination moved to a per-task leader. The requirements included coordinated checkpoints and recovery without gaps or duplicates. The company also sought an atomic view of transactions spanning multiple logs.
For an operational system, that last detail can matter more than a speed claim. Acting on part of a transaction risks interpreting an unfinished change as a finished event. Estuary’s rewrite makes correctness a concrete engineering subject, with recovery and transaction boundaries to examine.
Copy the experiment, not the percentage
Envoy supplies the most useful buying lesson. Its data team tested painful sources first, ran critical pipelines alongside the old system, then expanded. Estuary’s account reports roughly 75% lower pipeline costs. More instructively, near-real-time Salesforce ingestion coexisted with dashboards refreshing about every three hours. Moving data faster leaves downstream schedules to fix.
“The streaming capability came at no additional cost.”Envoy data team, in Estuary’s customer account
Readers can copy that experiment: choose a costly source, measure billable traffic, compare records during parallel operation, and follow freshness all the way to the decision. A source API that releases updates slowly, an unsuitable connector, or a warehouse job that runs overnight can limit the benefit. The purchase deserves a workload test. Free credits, milliseconds, and impressive percentages are all less persuasive than a quiet Tuesday.