Broadcom VMware Cloud Foundation 9.1.1 scales Kubernetes clusters 2.6x for AI

The release of VMware Cloud Foundation 9.1.1 signals a strategic push to capture AI workloads from banking and healthcare sectors that are restricted from using public cloud APIs. By enabling multi-tenancy at the namespace level, Broadcom addresses the high capital costs of on-premises GPUs while maintaining strict data residency. This move directly challenges the dominance of hyperscalers like AWS and Google by offering managed-service convenience within private data centers. As enterprises weigh the 2023 licensing shift against these new capabilities, the platform's ability to bridge the gap between legacy infrastructure and AI-native performance will determine its retention rate. Watch for the general availability of GitOps and native object storage to complete the private cloud stack.

Broadcom has been aggressively repositioning VMware Cloud Foundation as a private AI platform since completing its $69 billion acquisition of VMware in late 2023. The company launched its Private AI Factory with NVIDIA in November 2024, bundling validated reference architectures with NVIDIA GPUs and networking to give enterprises a turnkey path for on-premises AI inference and training. That partnership has expanded through 2025, with Broadcom announcing support for NVIDIA Blackwell GPUs in its Private AI Factory reference designs at VMworld 2025, signaling intent to keep pace with hyperscaler GPU refresh cycles. The multi-tenant model sharing feature in VCF 9.1.1 builds directly on this foundation by reducing per-workload GPU allocation, a critical cost lever for enterprises that cannot send sensitive data to public cloud endpoints.

The competitive landscape for private AI infrastructure is intensifying. AWS has responded to on-premises demand with Outposts and its Local Zones program, which now spans more than 30 metropolitan areas globally, offering lower-latency compute closer to regulated workloads without requiring full data-center ownership. Meanwhile, Nutanix reported in its fiscal Q2 2025 earnings call that AI-related pipeline had grown 150% year over year, driven by enterprises seeking alternatives to both hyperscalers and legacy virtualization stacks. Broadcom's licensing overhaul, which consolidated perpetual SKUs into subscription bundles, initially triggered customer backlash, but the company has used VCF feature velocity to justify the pricing shift. The 2.6x Kubernetes cluster scaling improvement in VCF 9.1.1 is a direct answer to critics who argued the platform could not match cloud-native elasticity.

On the technical side, Broadcom's Kubernetes scaling claims align with broader industry benchmarks for GPU-dense workloads. CNCF's 2025 annual survey found that 68% of organizations running AI workloads in production use Kubernetes as their orchestration layer, up from 54% the prior year, confirming the platform's centrality to AI infrastructure decisions. Broadcom's decision to push cluster scale limits in VCF 9.1.1 addresses a known bottleneck: NVIDIA's own DGX SuperPOD reference architecture recommends minimum cluster sizes of 256 GPUs for large-model training, a threshold that previously required hyperscaler-grade orchestration. By bringing that scale to on-premises environments with multi-tenant isolation, Broadcom positions VCF as a viable alternative for financial services and healthcare organizations that face strict data-residency mandates but still need to run frontier-class models internally.