On September 3, 2026, Broadcom pushed VMware Cloud Foundation (VCF) 9.1.1 to general availability, introducing Multi-Tenant Model Sharing, enhanced Kubernetes security layers, and an AI gateway preview supporting over 150 open-weight and third-party models. This release aims to provide enterprises running regulated on-premises data a viable alternative to hyperscaler AI infrastructure.
Inside the September 3 Release and Architectural Shifts
Broadcom shipped VCF 9.1.1 just four months after the initial VCF 9.1 general availability rollout in May 2026. According to official release documentation, the point update moves several long-awaited capabilities out of technical preview. Most notably, Multi-Tenant Model Sharing is now generally available, tackling a cost and infrastructure bottleneck that enterprise private AI deployments have faced.
Previously, customers running large language models on VCF often ended up deploying a separate model instance per business unit or project. This architecture resulted in duplicated GPU capacity. By shifting multi-tenancy into the foundational layer, VCF 9.1.1 optimizes hardware allocation across workloads. It represents a different architecture from how most hyperscaler AI services handle multi-tenancy, where isolation typically happens at the service boundaries.
Alongside multi-tenancy, the release embeds five new security layers directly around AI workloads and drives a reported 2.6x performance increase across Kubernetes clusters. Administrators can also configure an optional conversational AI chat interface inside the VCF Operations console, plugging in whatever model the customer chooses to run locally.
Private Infrastructure Versus Hyperscaler Velocity
The timing of Broadcom’s push highlights an infrastructural tug-of-war. The update arrived just two days before Amazon expanded its Aurora Serverless performance gains to five new regions, showcasing how public cloud providers continue to optimize for bursty, agentic workloads.
While AWS appeals to organizations scaling fast-moving public applications, Broadcom’s value proposition targets enterprises burdened by strict data sovereignty mandates, regulatory compliance, and predictable cost constraints. By delivering an AI gateway preview designed to broker access to over 150 models right out of the box, VCF 9.1.1 attempts to bridge the gap between open-weight agility and local security.
However, the platform is not entirely feature-complete. GitOps support and native object storage remain locked in tech preview status as of the September release, meaning platform engineers waiting for end-to-end automated pipelines will need to keep external tooling in place for now. Broadcom has not yet committed to a hard general availability date for those specific features.
Enterprise Adoption and the VCF 8 Support Pressure
Organizations are evaluating VCF 9.1.1 against the backdrop of ticking clock pressures. Broadcom’s aggressive push is linked to a VCF 8 support deadline, compelling legacy VMware shops to map out their upgrade paths away from older architectures.
- Multi-Tenant Model Sharing: Generally available, cutting down redundant GPU allocation across enterprise business units.
- AI Gateway Preview: Integrates brokering access for more than 150 open-weight and third-party models directly into the private stack.
- Security Layers: Five new hardening controls wrapping around containerized AI deployments.
- Pending Capabilities: GitOps workflows and native object storage remain in technical preview.
Whether enterprises choose to consolidate their AI pipelines inside local data centers using VCF 9.1.1 or migrate entirely to hyperscaler fabrics like AWS Bedrock and Azure AI Foundry depends heavily on internal compliance budgets. For companies locked into strict regulatory frameworks, Broadcom’s latest drop makes staying on-premises far more competitive than it was just a few quarters ago.
