IBM announced a self-hosted deployment option for IBM Bob on October 1, bringing its agentic software development platform to on-premises, private cloud, sovereign cloud, and air-gapped environments. Bob runs on infrastructure the customer manages, with model inference either on supported models the customer installs locally or, in hybrid configurations, through approved external model services. IBM says the option is generally available, with NVIDIA Nemotron and Poolside Laguna as the first supported self-hosted models.
IBM frames the release around data residency, citing a Futurum Research projection that hybrid and edge deployments will take 44% of the AI infrastructure market by 2030 while public cloud falls to 46%, and its own Institute for Business Value study from June 2026, The Calculus of AI Sovereignty, in which 68% of surveyed executives said meeting data residency and sovereignty requirements across geographies is challenging. Those are vendor-cited figures, and the practical point behind them is that regulated shops often cannot send source code to a hosted coding assistant at all.
“The future of enterprise AI will depend on security, governance and sovereignty,” said Neel Sundaresan, GM of AI and Automation at IBM. “Organizations need AI that operates inside environments they have control over, especially when working with sensitive code and regulated data. Bob’s self-hosted deployment provides a way for enterprises to bring agentic AI directly to those environments so they can benefit from the technology while maintaining security, compliance, and operational control.”
Sovereign AI Execution and Model Management Architecture
The self-hosted build carries the core Bob feature set: the IDE experience, BobShell for command-line and CI/CD work, parallel tool calling, the agent harness, and Bob’s skills and modes for understanding, planning, developing, and modernizing applications. Developers get the same Bob experience across deployment configurations, while platform and security teams decide where inference runs. For the Premium Package for Z, IBM says Bob services run inside a Red Hat OpenShift environment the enterprise manages, with models reached through a Bob model gateway.
At general availability, the customer sources and provides the model. IBM’s supported list splits by deployment method: NVIDIA Nemotron and Poolside Laguna for self-hosted installs the customer runs on its own infrastructure, and Claude Sonnet 5.0, Claude Opus 4.8, Gemini 3.7 Flash, and OpenAI GPT 5.6 Sol for hybrid or private SaaS connections through an approved private connectivity pattern. Eligible existing model licenses carry over under a bring-your-own-license approach, and IBM says multi-model routing and a longer supported list are planned for later releases.
The two paths can coexist in one deployment. IBM’s example is a bank running on-premises inference for work on core banking applications while pointing Bob at an approved external model service for a less restricted workload. The modernization case rests on the optional Premium Packages for Java Modernization, IBM i, and IBM Z, which add platform-specific workflows, connectivity, and context on top of the self-hosted core, each under its own licensing and deployment requirements.
Deployment Status and Licensing
The self-hosted option is generally available now. IBM has not published pricing, infrastructure sizing, or the full list of supported IDE and infrastructure integrations for the self-hosted build, and the model list above is what IBM supports at GA. We covered Bob’s general availability and the Premium Package for Z preview at Pensa 2026, and the Premium Package for Z gained its own self-hosted option on September 30, adding an Application View for top-down application discovery and zContext for IBM Z-specific knowledge.
Amazon