Photo: Tima Miroshnichenko / Pexels
The enterprise giant is betting that autonomous AI agents need to stay behind corporate firewalls, not roam free in public clouds
Broadcom is making a calculated bet on where AI agents should live, and the answer is emphatically not the public cloud. Through its VMware division, the company has been rolling out a series of Tanzu platform upgrades designed to keep autonomous AI workloads locked inside private cloud infrastructure, wrapped in layers of security that treat every agent like an untrusted actor until proven otherwise.
What Broadcom actually built #
The foundation of this push is VMware Tanzu Data Intelligence, a multimodal data lakehouse launched in August 2025. Think of a data lakehouse as a hybrid storage system that combines the flexibility of a data lake (where you can dump any kind of data) with the structured querying power of a traditional data warehouse. This one handles petabytes of both structured and unstructured data with millisecond response times, which matters when AI models need to pull information from massive corporate datasets without introducing latency bottlenecks.
Then came the more security-focused piece. In April 2026, Broadcom introduced Tanzu Platform Agent Foundations as part of Tanzu 10.4, unveiled at the AI in Finance Summit. This is where the trust architecture gets interesting.
The framework operates on a deny-by-default networking model. In practical terms, an AI agent deployed in this environment cannot communicate with anything, access any service, or reach any data unless explicitly granted permission through service bindings.
Each agent runs inside sandboxed containers with enforced CPU and memory limitations. If an agent starts consuming more resources than allocated, whether through a bug, a prompt injection attack, or simple misconfiguration, the sandbox constrains it. The system also leverages trusted Buildpacks for what Broadcom calls immutable supply chains, meaning the software components that make up each agent’s runtime cannot be tampered with after deployment.
Why private clouds, not public ones #
Broadcom is positioning VMware Cloud Foundation as the substrate that gives businesses predictability and direct control over their AI deployments. The pitch to CIOs and CISOs is essentially: you can run agentic AI at scale, but you maintain the security perimeter, the compliance framework, and the kill switch.
The financial services sector appears to be a key target market for these capabilities, given that Agent Foundations was specifically introduced at a finance-focused AI summit.
The bigger picture for enterprise AI #
The Tanzu platform’s evolution tells this story clearly. Version 10.3 integrated generative AI capabilities in a relatively straightforward manner. Version 10.4, arriving just months later, shifted the emphasis to containment and control.
VMware Explore 2026, scheduled for August 31 through September 3, is expected to double down on private AI cloud solutions.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our