Data Analytics News: Airflow 3.4 Soars on Any-Language Dags

Today’s data analytics news is dominated by orchestration and open table formats. Apache Airflow 3.4.0 lets teams write pipelines in TypeScript, Java or Go, Snowflake pushed streaming ingestion to 20 GB/s per table, and both Snowflake and Google BigQuery widened Iceberg catalog interoperability. Here are the five data analytics news stories that data analysts and engineers should know, each tied back to a primary source.

Apache Airflow 3.4.0 Brings Task Loops and Any-Language Dags

Apache Airflow 3.4.0 shipped on November 6 with Task Loops (AIP-111), which add a .loop() method to decorated task groups. A max_iterations value is required, an until condition is optional, and loops cannot be nested. The release also adds the AIP-85 Dag Importer, so Dags can be written in TypeScript, Java or Go with no Python file; the TypeScript runtime needs Node.js 22 or newer. The common.ai provider reaches GA at version 1.1.0 with @task.llm, @task.agent and human-in-the-loop approval.

This data analytics news matters most to analysts who lean on scheduled SQL, because loops make retry-until-fresh patterns far easier to express. If you are still building your foundation, our guide on getting started with SQL for data analysis covers the queries these pipelines run.

Snowflake Pushes Autonomous Data Engineering and 20 GB/s Streaming

At Expedition 2026, Snowflake announced its foundation for autonomous data engineering. Snowpipe Streaming reached up to 20 GB/s aggregate uncompressed per table with a median 3-second data-readiness time, based on an internal test on October 20, so treat it as a benchmark rather than an SLA, and read this data analytics news with that caveat in mind. Cboe Global Markets says it now loads almost 200 billion rows of market data per day this way, and Klarent reports a 60% TCO cut after moving Spark workloads, with its longest pipeline dropping from over 10 hours to under 3. CoCo Cloud Agents and Automations are GA, while Pipeline Builder and Write-Audit-Publish on Iceberg are in public preview.

Snowflake Horizon Makes Cross-Account Iceberg GA Without Secrets

In a November 1 release note, Snowflake made cross-account read and write of Apache Iceberg tables through Horizon Catalog generally available. Authentication uses workload identity federation with short-lived, auto-refreshing tokens, so there are no secrets to create or rotate. Vended credentials are the recommended path, which removes the need for an external volume, and the feature is available in all commercial regions.

BigQuery Expands Lakehouse Federation and Cost Attribution

Google’s BigQuery release notes show federated catalogs for Databricks Unity Catalog, AWS Glue Data Catalog, Snowflake Horizon Catalog and Workday Data Lake reaching GA on November 2, alongside GA cross-cloud caching. On November 9, compute cost attribution for BigQuery editions entered Preview, charging reservation usage back to the projects that run queries, and job-comparison tools in the jobs explorer went GA. Cross-platform catalog reads are quickly becoming table stakes, which is the most useful data analytics news trend for teams running multi-cloud estates.

Azure Databricks Changes Jobs Defaults and Ships Zerobus Ingest

The November 2026 Azure Databricks release notes say new jobs compute now defaults to Auto access mode instead of Dedicated as of November 9; existing jobs are unaffected, and compliance-security-profile workspaces follow in December. Zerobus Ingest into default storage is GA, and AI Functions in SQL will stop running on Databricks Runtime below 17.3 LTS from January 2027. Check your job definitions before then, and brush up on advanced SQL filtering if you write the queries behind those functions.

What This Data Analytics News Means for Your Stack

The common thread in today’s data analytics news is fewer moving parts: orchestration in the language your team already uses, streaming ingestion without custom plumbing, and Iceberg tables readable across platforms. Verify preview features before betting production work on them, and note that vendor performance figures come from vendor tests. We will keep tracking the data analytics news that affects day-to-day pipelines and queries.