Cloudflare Basin is now generally available. Announced on 1 October, it is a serverless data platform built on Apache Iceberg and Cloudflare R2. It lets teams ingest, store, catalog and query analytical data without managing clusters.
That is a product launch. What sits behind it is a category move, and it lands in the middle of a fight the data industry is still having.
What Cloudflare Basin actually is
Cloudflare Basin has three parts, all generally available. Basin Pipelines ingests events from Workers, HTTP endpoints and Cloudflare Logpush, transforms them with SQL, and writes Iceberg tables to R2. Basin Catalog manages the Iceberg metadata and maintains the tables automatically. The third, Basin SQL, is a serverless, distributed engine that queries those tables directly.
The platform was formerly called the Cloudflare Data Platform, launched in beta during Birthday Week 2025. Its rename marks general availability and pulls the three products under one name.
The design choice that matters is Iceberg. Because the data is stored in an open table format, any Iceberg-compatible engine can read it directly — PyIceberg, DuckDB, Snowflake, Spark, StarRocks, Trino. Nothing is copied or converted. Cloudflare’s pitch is that teams own their data, not just access to it.
Pricing is usage-based, with no hourly charges. There are no clusters to provision. Cloudflare cites Bobsled, a data-product platform, as a customer.
Why Cloudflare Basin is a category move
Cloudflare is not a data-warehouse company. It built its business on networking, security and edge compute. Basin puts it in direct competition with Snowflake, Databricks and the hyperscaler data stacks — a crowded and well-funded market.
Its differentiator is not a query engine. It is the economics of egress. Cloudflare owns R2, its object storage service, which charges nothing for egress. Amazon S3, Google Cloud Storage and Azure Blob Storage all charge to move data out. Because Basin stores data in open Iceberg on R2, teams can query it from any cloud or region without transfer fees.
That is a genuine structural advantage, and analysts see the logic. Constellation Research’s Michael Ni noted that Cloudflare already has the global infrastructure, serverless compute and an egress-free storage model, and already sits in the path of application, log and event data. He also set expectations carefully. He does not expect large companies to rip out Databricks or Snowflake soon. Cloudflare can capture workloads at the edge of the traditional warehouse market, then move up the stack.
Cloudflare makes the same caveat. Basin is not positioned as a blanket replacement for every warehouse workload.
The Iceberg wars are over — the catalog fight is on
To understand what Cloudflare is really competing on, you have to look at where the industry has settled.
The table-format war is finished. Apache Iceberg won, and even Databricks — creator of the rival Delta Lake format — now treats Iceberg as a first-class citizen. Snowflake, Google BigQuery, Microsoft Fabric and AWS have all adopted it. The question is no longer which format; it is which catalog governs the data.
That is the new battleground, and it is crowded. Databricks’ Unity Catalog supports managed and foreign Iceberg tables, Iceberg v3 features and federation across AWS Glue, Snowflake Horizon and Google Cloud. AWS offers S3 Tables and the Glue REST catalog. Snowflake embeds Iceberg REST endpoints into every account through Horizon and runs a managed version of Apache Polaris. Polaris itself graduated to a top-level Apache project in February 2026, and Dremio positions as Iceberg-native.
Gartner projects that by 2028, 80% of enterprise data warehouses will run on a lakehouse format. That format is built on open table formats like Iceberg. That is the market Cloudflare is entering.
Cloudflare’s advantage in this specific fight is that it has no proprietary catalog to defend. It can present Basin Catalog as a neutral Iceberg REST catalog, and its openness argument costs it nothing. For a buyer worried about lock-in, that is a cleaner story than one from a vendor whose flagship format competes with Iceberg.
The caveats in “zero egress”
The headline claim deserves scrutiny, because “zero egress fees” is not absolute.
Cloudflare is accurate that retrieving data directly from R2 — through the Workers API, the S3 API or r2.dev domains — incurs no transfer charges. That is the model R2 has run since 2021.
But there are two edges. First, when Basin Pipelines delivers streaming data to R2 in Iceberg or Parquet beyond the included free tier, a billing line item called “Sinks (egress)” applies. Second, if you connect other metered services to an R2 bucket, those services bill you separately. The zero-egress promise holds for the direction that matters most to most teams — reading your own data — but it is not the same as a platform with no transfer costs anywhere.
That nuance matters because the whole pitch rests on it.
The developer-tools story behind the launch
Cloudflare paired the data launch with two developer-tools releases that reveal its broader thinking.
The first is cf, an agentic command-line tool covering the entire Cloudflare API — more than 3,000 operations, up from roughly 280 in Wrangler. It uses a TypeScript configuration format and adds a natural-language command search, so an agent can describe a task and find the command.
The second is Forge, open-sourced under Apache 2.0 in late September. Forge is a pipeline that runs in CI and generates SDKs, CLIs, documentation and MCP servers from API definitions. Cloudflare built it because its API spans more than 3,500 operations across hundreds of services, and the hosted generators it used could not handle that scale.
The common thread is that Cloudflare is designing for agents as customers. Its framing is that CLIs, SDKs and MCP servers are now table stakes for every product, not just developer tools. Basin’s SQL engine, described as a readily available API “for you and your agents,” fits the same thesis.
What to watch
Three markers will show whether Basin lands.
First, named enterprise deployments. Bobsled is a start-up reference. A large regulated or data-heavy enterprise adopting Basin for real workloads would signal that the openness pitch converts beyond the edge of the market.
Second, governance. The catalog fight is about control, not just storage. If Basin Catalog cannot match Unity Catalog and Snowflake Horizon on fine-grained access control and lineage, openness alone will not carry enterprise deals.
Third, whether the egress economics hold at scale. The advantage depends on customers actually moving data across clouds and regions. If they do not, the savings are smaller than the pitch implies.
Cloudflare has the infrastructure and the pricing model to make Basin credible. What it does not yet have is proof that the warehouse market’s edge is big enough to matter.
Editor’s Note
Sources: Cloudflare’s Basin press release of 1 October 2026, Cloudflare’s Basin launch blog, the Basin product documentation, and Cloudflare’s earlier Data Platform announcement of September 2025.
Egress and pricing detail is from Cloudflare’s R2 pricing documentation. Competitive context is from Databricks, AWS and Snowflake documentation and blogs, plus material on Apache Polaris and Dremio. Independent coverage and analysis are from SiliconANGLE (including Constellation Research’s Michael Ni), Techzine, and LavX News. The cf CLI and Forge details are from Cloudflare‘s developer blog and documentation and from independent analysis. The “zero egress” caveats, the Gartner lakehouse projection and the catalog-fight framing are drawn from the cited sources; TechRecast has not independently benchmarked Basin’s performance.