Red Hat AI Inference on IBM Cloud adds private network endpoint

Overview

Updated 54 minutes ago

IBM Cloud customers can now send Red Hat AI Inference requests over IBM's private network instead of the public internet. The API is unchanged except for the hostname: private.us-east.rhai.ibm.com replaces us-east.rhai.ibm.com.

The private path runs through a virtual private endpoint gateway, a virtual interface that sits in the customer's own virtual private cloud (VPC) subnet. That clears a common security-review blocker for banks, hospitals and government agencies, though only in the Washington, D.C. (us-east) region for now.

Why it matters

If a security review blocks AI prompts from crossing the public internet, this removes that blocker for IBM Cloud workloads in one region.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

IBM Cloud regions with a private inference endpoint

Washington, D.C. (us-east) is the only region listed.

Hostname change needed to switch paths

Developers swap us-east.rhai.ibm.com for private.us-east.rhai.ibm.com.

Code changes required

Project IDs, API keys, model IDs and request bodies stay the same.

Connectivity paths per project

Public endpoint or private endpoint, chosen by configuration.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Luca Ioffredo

Author of the IBM Community post on the private endpoint

Published the developer walkthrough on 2026-11-02

Organizations Involved

International Business Machines Corporation (IBM)

Cloud provider

Operates Red Hat AI Inference on IBM Cloud and the virtual private endpoint gateways

IBM runs the managed inference service and the private network path that connects customer VPCs to it.

Red Hat, Inc.

Software company

Supplies the inference platform behind the managed service

Red Hat provides the AI inference software that IBM sells as a managed cloud service.

Timeline

May 2026 November 2026

4 events Latest: Today

Tap a bar to jump to that date

November 2026
IBM publishes private-endpoint developer walkthrough
Today Documentation

Post shows the hostname swap and a Code Engine pattern.
Today
October 2026
Private endpoints launch in us-east
Capability

VPE gateways connect VPCs to inference; embeddings API also added.
October 5th, 2026
May 2026
Inference service goes generally available
Launch

OpenAI-compatible chat completions reach production, with a console playground.
May 22nd, 2026
IBM announces managed Red Hat AI Inference
Announcement

IBM sets general availability for May 22, 2026.
May 12th, 2026

Scenarios

IBM extends the private inference endpoint to a second region

Possible Resolves by Q2 2027

Discussed by: IBM's release notes, which list us-east as the only region with a private endpoint

IBM's pattern with new cloud capabilities is to launch in us-east and add regions later. A second region would need its own private hostname and endpoint gateway in that region's service catalog. The release notes page records region availability for each feature, so the change would be visible there first.

Private inference stays a one-region feature through 2027

Possible Resolves by End of 2027

Discussed by: IBM's documentation, which describes the private endpoint as a us-east capability

VPE gateways are per-region infrastructure, and IBM has not said when or whether it will build them elsewhere. If demand concentrates in Washington, D.C., the feature could stay there while IBM invests in other parts of the service. The absence of a second region in the docs would confirm this path.

IBM names a customer running inference over the private endpoint

Possible Resolves by Q2 2027

Discussed by: IBM's case-study library and IBM Community blog, which regularly feature customer walkthroughs

A named case study would show the feature clearing a real security review, not just a documentation page. IBM publishes customer stories for cloud services, and a regulated buyer describing its network setup would be the clearest signal that private inference is in production.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

November 2017

AWS PrivateLink (2017)

Amazon Web Services introduced PrivateLink, which let customers reach AWS services and third-party software through private IP addresses inside their own virtual network instead of over the public internet. The connection used network interfaces tied to a service endpoint.

Then

AWS services and software vendors began offering private endpoints as an option.

Now

Private connectivity became a standard procurement requirement for regulated cloud workloads.

Why this matters now

The Red Hat AI Inference private endpoint follows the same design: the service stays shared, the network path becomes private.

2020

Azure Private Link (2020)

Microsoft made Azure Private Link generally available, giving Azure customers private IP access to platform and partner services from inside their virtual networks.

Then

Azure services, including later AI services, added private endpoint support.

Now

Private endpoints became a checkbox in enterprise security reviews for cloud AI.

Why this matters now

IBM is adding the same checkbox to its managed inference service, one region at a time.