# Air Gaps Don't Have to Mean Dark Data

_Bright Machines streams factory data out one auditable path with zero inbound access, queryable in minutes._

**Published:** 2026-11-12
**Source:** https://www.startuphub.ai/ai-news/technology/2026/air-gaps-don-t-have-to-mean-dark-data

---

I think Bright Machines proved air-gapped factories can still serve live customer data without opening inbound access.

[Databricks](https://www.databricks.com/blog/air-gapped-not-data-dark-how-bright-machines-gives-customers-live-factory-dashboards-without) says Bright Machines captures PostgreSQL changes with Debezium at the edge and reads Iceberg tables in place through Unity Catalog federation.

Bright Machines designs software-defined manufacturing that combines robotics, machine vision and automation software for electronics assembly at scale per TipRanks.

A row committed on the line is queryable in [Databricks](https://www.startuphub.ai/ai-news/technology/2026/databricks-unveils-omnigent-meta-harness) in one to two minutes per the company account, while a Lakeflow job rebuilds silver models every ten minutes.

The old path stitched nine loosely coupled services from broker to BI tool per Databricks, and it silently dropped schema changes while dead-letter queues hid misses.

The new path keeps operators on the on-prem assembly dashboard while one edge writer copies Iceberg objects to the site cloud account for cross-account replication.

Tables register in AWS Glue only when metadata fully replicates and only forward, so Unity Catalog governs the validated version with no copy job.

Teradyne is pushing test in the other direction with Omnyx Edge inline board test for compute, sensing and motion boards per Quantisnow, which keeps validation at the line.

The fair pushback is this design moves a few hundred gigabytes a day as hundreds of thousands of small objects per the company account, so object rate matters more than bytes.

That risk partly survives because a five-minute Debezium heartbeat fails the job red when any source lags over thirty minutes, down from days of silent drift per Databricks.

If you run locked-down plants, copy this pattern and push open formats out one auditable door while you enforce roles where analysts query.

---

Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.

© StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.