Red Hat, targeting enterprise AI "Inference Operations," releases AI 3.4... expanding into hybrid cloud, automotive, and space sectors

IBM ($IBM) subsidiary Red Hat has released a major update aimed at enterprise needs for “real-world operations” of artificial intelligence (AI). While rolling out a new AI platform that supports large-scale inference in hybrid cloud environments, the company also plans to expand the scope of its open-source platform to software-defined vehicles and even space data centers.

This release was announced at the “Red Hat Summit” held in Atlanta, USA. The core is “Red Hat AI 3.4.” The platform is designed to enable enterprises to run AI models and agents across multiple cloud environments and on-premises environments, with a focus on bringing AI applications that are still in the experimental stage into actual business use.

In a pre-briefing, Joe Fernandez, Vice President and General Manager of Red Hat AI, summarized the company’s strategy into four points: providing fast and flexible inference environments, connecting enterprise data with models and agents, deploying and managing agents in hybrid cloud environments, and building a unified AI platform that can run any model across various hardware and cloud environments.

This version adds a “Model as a Service” feature. Administrators can control access permissions to AI models through a central gateway, track usage, and apply policies in a unified manner. Red Hat also expanded support for distributed inference. At the same time, optimization technologies such as “speculative decoding” were introduced to improve text generation speed by up to 3 times, with the goal of lowering operating costs.

Red Hat believes that in the future, the focus of enterprise AI will shift from “training” to “inference.” In particular, with the widespread adoption of AI agents, inference demand may grow exponentially. This explanation also points out that enterprises are more inclined to connect existing models to their own data to achieve business automation, rather than developing large foundational models from scratch.

Strengthening AI agent management… Expanding cooperation with NVIDIA

Red Hat also strengthened its agent management and observability capabilities. It can track inference call records and tool usage records, and includes support for MCP (Model Context Protocol) gateways and directories. In addition, it added prompt management, automated evaluation tools, and AI security testing functions. Some functions use technology from Chatterbox Labs, which was acquired recently.

Cooperation with NVIDIA ($NVDA) has also been further expanded. Red Hat announced plans to support NVIDIA’s “Blackwell” architecture and the next-generation “Vera Rubin” platform. It will also participate in NVIDIA’s “OpenShell” project for sandboxing and secure execution of AI agents. This move is intended to meet enterprise needs to run AI agents in environments handling sensitive data in a controlled manner.

A series of these releases indicates that the enterprise AI market is shifting from “model competition” to “operations competition.” This means that, compared with which model is smarter, it becomes more important how to reliably deploy, control, and run them efficiently.

Targeting “Linux in Space”… Announcing collaboration with the International Space Station and Nissan

Red Hat has also significantly expanded the scope of its applications in partner releases. The company plans to collaborate with Voyager Technologies to deploy Red Hat Enterprise Linux 10.1 and Universal Base Image in the “Space Edge” micro data center onboard the International Space Station (ISS).

The project aims to perform data processing and AI workloads in an orbital environment, while extending DevSecOps approaches used on the ground to the space environment. Considering the characteristics of space computing—limited power, intermittent connectivity, and constrained hardware resources—it will run applications in immutable containers, migrate container-based workloads, and incorporate quantum-resistant cryptography technologies.

In the automotive sector, it will co-develop the next-generation software-defined vehicle platform with Nissan. The foundation is the “Red Hat Vehicle Operating System.” Nissan plans to use this to lay a standardized Linux foundation for future centralized vehicle computer architecture, and to drive software updates and AI feature deployment across the full vehicle lifecycle.

Nissan stated that this collaboration is a strategic decision aimed at increasing its control over its own software development stack. This shows that, as the full-vehicle industry shifts from being hardware-centered to being software-centered, leadership in vehicle operating systems is becoming a key competitive advantage.

This week, Red Hat will also continue to publish more announcements related to Red Hat Enterprise Linux, OpenShift, and Ansible automation. Overall, this release indicates that Red Hat aims to strengthen its position as an “operations platform” in the enterprise AI and hybrid cloud market, while extending its open-source influence into new computing domains such as automotive and space.

TP AI note: This article uses language models based on TokenPost.ai for summarization. The main content in the body may be omitted or may not be consistent with facts.