IBM has added two managed Red Hat services to its cloud portfolio, deepening its push into enterprise AI infrastructure and virtualization modernization as corporate technology buyers seek more predictable ways to run production workloads across hybrid environments.

The company said on May 12 that it is introducing Red Hat AI Inference on IBM Cloud and Red Hat OpenShift Virtualization Service on IBM Cloud. The first service is intended to help organizations deploy and govern real-time AI inference without managing the underlying GPUs, runtimes or platform operations themselves. The second gives enterprises a managed path to migrate and operate virtual machines on Red Hat OpenShift, using Kubernetes-based infrastructure and IBM Cloud bare-metal capacity.

The announcement is closely aligned with IBM’s broader positioning in the cloud market. Rather than competing primarily on general-purpose compute scale, IBM has concentrated its cloud and software strategy around hybrid infrastructure, regulated industries, open source platforms, automation and AI governance. The new services use Red Hat technology as the core platform layer and IBM Cloud as the managed delivery environment, allowing IBM to present customers with a more integrated route from existing workloads to modern AI and cloud-native operations.

IBM said Red Hat AI Inference on IBM Cloud will be generally available on May 22, 2026. Red Hat OpenShift Virtualization Service on IBM Cloud is in limited availability and is expected to reach general availability in June 2026. The timing gives IBM a near-term product launch in AI inference while establishing a second track for customers planning virtualization transitions over the coming months.

The AI inference service is aimed at a fast-growing operational issue for enterprises. Many companies have built pilots around generative AI and agentic workflows, but scaling those systems into production introduces a different set of constraints. Once models are used by employees, customers, software agents or automated workflows, inference becomes a recurring workload rather than a one-time experiment. That shifts attention to latency, throughput, cost per token, security controls, model access management and auditability.

IBM described Red Hat AI Inference on IBM Cloud as a fully managed service for running production-grade AI models. It combines Red Hat AI’s inference engine with IBM Cloud infrastructure and governance features, including IBM Cloud IAM integration, audit logging, privacy controls and service-level-backed reliability. The service is designed to let developers use familiar OpenAI-compatible APIs while avoiding direct management of GPUs and tuning runtimes.

The model catalog cited by IBM includes Granite 4.0 H Small, Mistral-Small-3.2-24B-Instruct, Llama 3.3 70B Instruct, GPT-OSS-120B and Nemotron-3-Nano-30B-FP8, with more open models and custom model support planned. IBM’s inclusion of its own Granite model family alongside third-party open models reflects a commercial strategy centered on customer choice rather than a single proprietary model stack.

That approach is consistent with Red Hat’s positioning of Red Hat AI Inference as an open, portable inference stack. Red Hat describes the product as powered by vLLM and llm-d, technologies intended to improve token economics, hardware utilization and response-time consistency across hybrid cloud environments. For customers, the practical value proposition is not simply model access, but operational control over how models are served, scaled and governed across infrastructure.

The launch also addresses a common friction point in enterprise AI adoption: the handoff from data science experimentation to IT operations. In pilot phases, teams may rely on a narrow group of developers, ad hoc cloud accounts or manually provisioned accelerators. In production, the same workloads often need centralized access controls, repeatable deployment processes, usage reporting, cost management, reliability targets and policy enforcement. IBM is using the managed-service model to absorb more of that operational complexity while keeping customers within a Red Hat-based architecture.

The second service, Red Hat OpenShift Virtualization Service on IBM Cloud, targets a separate but related modernization challenge. Many enterprises still rely heavily on virtual machines for core systems, including ERP, healthcare records, banking applications, analytics platforms and line-of-business software that cannot be rewritten quickly. At the same time, virtualization teams are under pressure from licensing changes, infrastructure consolidation, budget scrutiny and rising expectations for security and compliance.

IBM said the virtualization service allows enterprises to migrate and run VM-based workloads on Red Hat OpenShift with Kubernetes-based infrastructure, automated lifecycle management and a consistent foundation for later containerization and application modernization. It runs on IBM Cloud VPC Bare Metal, which IBM says is intended to provide predictable performance and total cost of ownership for virtualized workloads.

The service is built around Red Hat OpenShift Virtualization, which uses KubeVirt and KVM to bring virtual machines into a Kubernetes-managed environment. In practical terms, that means organizations can manage VMs and containers through a more unified platform model while preserving a VM-first operating path for applications that are not ready to be refactored. IBM provisions the environment as a managed service, reducing the need for customers to operate the virtualization platform themselves.

IBM said it will manage platform lifecycle operations for the virtualization service, including upgrades, patching, automated recovery and worker-node remediation. Those functions are significant because much of the cost and risk in virtualization platforms sits not only in the initial migration, but in ongoing day-two operations. By packaging those tasks into a managed service, IBM is targeting enterprises that want to reduce operational overhead without immediately redesigning applications for containers or serverless environments.

The company also highlighted integrated migration tooling, including the Migration Toolkit for Virtualization. That tooling is intended to help customers move from legacy virtualization environments with less disruption. IBM is pairing the service with support from IBM Technology Expert Labs, IBM Consulting, Red Hat Services and global system integrator partners, indicating that the offering is likely to be sold not only as a cloud service but also as part of broader infrastructure consulting and migration engagements.

For IBM, that services pull-through matters. The company’s strongest differentiation in enterprise technology often comes from combining software, cloud infrastructure and consulting. A customer evaluating VM migration may need architecture planning, application dependency mapping, compliance reviews, migration execution, performance validation and operating-model changes. IBM’s ability to attach advisory and implementation work could make the virtualization launch more commercially significant than a standalone infrastructure SKU.

The two services also reinforce the strategic importance of Red Hat within IBM’s portfolio. IBM acquired Red Hat in 2019 to strengthen its hybrid cloud position, and Red Hat OpenShift has become the centerpiece of IBM’s modernization story. By launching managed AI inference and virtualization services on IBM Cloud, IBM is extending Red Hat from container platform and Linux subscription economics into higher-level managed services tied to AI operations and VM modernization.

That expansion comes as enterprise buyers increasingly want flexibility across public cloud, private cloud, on-premises infrastructure and edge environments. Many regulated companies are unwilling or unable to move all workloads into a single public cloud. They may have data residency requirements, latency constraints, existing mainframe and on-premises estates, or sector-specific compliance obligations. IBM’s message is that Red Hat can provide a consistent platform foundation across those environments while IBM Cloud supplies managed infrastructure for selected workloads.

The AI inference launch also places IBM into one of the most competitive layers of the AI infrastructure stack. Cloud providers, model companies, GPU specialists and software vendors are all trying to control the production inference layer because it is where ongoing AI usage translates into recurring compute demand. Training large models remains expensive, but production inference can become a larger operational cost as AI agents, chatbots, code assistants and retrieval-augmented applications scale across thousands or millions of interactions.

IBM’s advantage is likely to depend less on raw accelerator availability than on governance, integration and enterprise trust. The company is aiming the service at clients that require access controls, auditability, privacy protections and predictable operations. That positioning may resonate with banks, insurers, healthcare companies, governments and industrial customers that are interested in AI but cautious about uncontrolled model usage or unmanaged public endpoints.

At the same time, IBM faces intense competition. Major hyperscalers continue to expand managed model-serving platforms, proprietary AI services and GPU-backed infrastructure. Specialized providers are also targeting the inference market with optimized accelerator clouds and open-source serving stacks. IBM will need to show that its Red Hat-based managed approach can offer enough control, portability and governance to offset any perceived disadvantages in scale or ecosystem breadth.

The virtualization service enters a market shaped by enterprise reassessment of long-standing infrastructure assumptions. Many organizations are reviewing virtualization roadmaps as they balance cost, vendor concentration, cloud migration, container adoption and operational resilience. IBM’s service gives those customers a potential intermediate path: move VM workloads into a managed OpenShift-based environment while preserving optionality for later modernization. That could appeal to organizations that view a full application rewrite as too risky, too expensive or too slow.

From a technology architecture perspective, the services also connect two trends that are often treated separately. AI adoption requires a modern platform for models, data, APIs, governance and compute orchestration. Virtualization modernization involves the estate of existing applications that still run much of the enterprise. By placing both offerings on IBM Cloud and Red Hat foundations, IBM is arguing that customers can modernize current workloads and prepare for AI-driven operations through a common platform strategy rather than separate infrastructure silos.

The announcement is not framed as a major consumer-facing AI product or a new foundation model release. Its importance is instead in enterprise deployment mechanics. IBM is addressing the less visible but commercially significant infrastructure work required to make AI and modernization projects durable: managed runtime operations, identity integration, audit trails, migration tooling, lifecycle automation and support for hybrid operating models.

For customers, the near-term questions will center on availability, pricing, performance, model support and migration execution. Red Hat AI Inference on IBM Cloud is scheduled to become generally available first, giving customers an earlier opportunity to evaluate model serving, API compatibility, governance features and cost behavior. The virtualization service remains in limited availability, with broader access expected in June, so enterprise adoption will likely depend on pilot outcomes and IBM’s ability to show smooth migration from incumbent environments.

The services also give IBM more ways to defend and expand existing enterprise accounts. A company already using Red Hat Enterprise Linux, Red Hat OpenShift, IBM Cloud or IBM Consulting may find the new offerings easier to evaluate than a separate platform from another provider. Conversely, customers considering changes to virtualization or AI infrastructure could use the services as an entry point into IBM’s broader hybrid cloud stack.

IBM said the new managed services build on existing managed offerings across Red Hat Enterprise Linux, Red Hat OpenShift, Red Hat Ansible Automation Platform and Red Hat AI. That continuity is central to the company’s message: enterprises can adopt new AI and virtualization capabilities without abandoning established Red Hat tooling or taking on a wholesale platform replacement.

The launch underscores how enterprise cloud competition is shifting from basic migration to managed modernization. Companies are no longer only asking where to host applications; they are asking how to govern AI usage, control inference economics, reduce infrastructure operations, migrate VMs without disruption and create a path toward cloud-native development at their own pace. IBM’s latest Red Hat-based services are designed to compete directly in that infrastructure decision cycle.

If IBM can convert the announcement into customer deployments, the services could strengthen its hybrid cloud narrative at a time when AI infrastructure spending is moving from experimentation toward production discipline. The company’s challenge will be proving that managed Red Hat services on IBM Cloud can deliver measurable benefits in cost control, governance, migration speed and operational resilience against alternatives from larger cloud competitors and specialized AI infrastructure providers.