Palantir says customers can expect compute costs for streaming ingests on Foundry worker to fall by around 97.5% from November 1, 2026. The company posted the change on November 5, 2026, on its Foundry documentation announcements page.
Palantir said customers "can expect to see a reduction of around 97.5% in compute costs for their streaming ingests when running on Foundry worker compared to what it would have been previously." The announcement does not give a baseline price, a new rate or a pricing unit.
It also does not say whether the cut is permanent or carries an end date. The effective date falls four days before the posting date, and Palantir does not say how November usage before the announcement is treated.
What changed and who it is for
The change applies to teams that stream data from external systems into Foundry. Palantir calls Foundry worker "the recommended method for accessing external systems" and says it replaces the legacy agent worker. Palantir said the change "is intended to facilitate migration from agent worker to Foundry worker."
Palantir says it builds on a recent pricing announcement. An October 2026 note said that "From August 2026 on," Foundry worker "no longer incurs compute costs for data ingest and export, media ingest, and Iceberg extract operations." That note said no credits would be issued for prior compute charges on those operations.
The October note also said: "Streaming ingests will continue to be billed. The pricing model for streaming operations will be available soon."
How migration works
Palantir's Foundry worker vs. agent worker guide says agent worker "is in the legacy phase of development." The guide says agents now act only as network proxies.
According to the guide, "Foundry worker runs each capability in its own container, sized to the job." It adds that new capabilities and ongoing feature development target Foundry worker only.
Existing agent worker sources keep working, and the guide calls agent worker "legacy, fully supported." A guided wizard handles migration, which is reversible for 30 days.
All running jobs, including syncs, exports and tasks, are terminated during migration. Scheduled jobs run in Foundry at their next scheduled time. Unscheduled jobs must be restarted manually.
Palantir recommends that agent worker users with streaming ingests migrate "as soon as possible." It asks for feedback through Palantir Support or its Developer Community, using the data-connection tag.
Why it matters
In Foundry News's view, Palantir is using pricing to pull customers onto Foundry worker while keeping agent worker supported. The October announcement removed compute costs from several Foundry worker operations from August 2026. This step extends lower compute pricing to streaming ingests, the one category that note left out.
For buyers, the percentage is clear but the budget impact is not. Without a baseline or a rate, savings are hard to put in dollar terms. The per-unit streaming pricing model Palantir promised in October is the figure cost forecasts will depend on.
For platform teams weighing a move, the migration mechanics matter as much as the price. Because migration terminates running jobs, streaming pipelines need a planned switchover and a check that unscheduled jobs come back up.
