IBM is moving its agentic software-development platform Bob closer to customers' own infrastructure, announcing general availability of a self-hosted deployment option that lets organizations run AI-assisted development on-premises, in private clouds or in sovereign clouds.
By IBM's description, Bob goes beyond code generation: the platform is meant to apply AI to the whole job of software delivery and modernization — analysing existing code, planning changes, executing migrations. The self-hosted mode is aimed at customers with sensitive or mission-critical systems that cannot send their data to a public AI cloud.
The announcement fits a broader industry movement. As governments and enterprises frame data residency, auditability and sovereignty as requirements, vendors of agentic coding tools are increasingly offering variants that run on customer-controlled hardware. IBM is also extending a premium package for its Z mainframes with on-premises deployment.
For IBM, the move is strategically useful twice over. It opens Bob to regulated sectors that avoid public AI services, and it positions the company as a vendor that meets governance requirements rather than sidesteps them.
For buyers, the decisive question is less availability than operating burden: a self-hosted agentic platform needs GPUs, model updates, access control, and audit trails — operational work that a cloud customer delegates to a provider. Whether companies absorb that cost is likely to depend less on technology than on their regulatory environment.
