English
Broadcom unveiled VMware Private AI Cloud at VMware Explore, a private cloud architecture built to run AI models close to enterprise data. Instead of sending sensitive data to external AI providers, the platform brings models into on-premises private infrastructure to preserve data control and privacy.
It unifies VMware Cloud Foundation (VCF), AI Factory, Tanzu Platform, vDefend, and Avi Load Balancer into a single environment for AI inference, agentic applications, containerized services, and traditional VMs. The stack delivers a unified infrastructure and operational model for AI and general workloads, letting enterprises retain full oversight of data location, security policies, and resource usage.
Ram Velaga, President of Broadcom Infrastructure Software Group, described the platform as a convergence of private cloud and private AI infrastructure. It is designed to support production inference and agentic workloads without sacrificing data sovereignty, compliance, or predictable operational costs.
VCF Optimizes AI Infrastructure and Token Costs
VMware Private AI Cloud targets three key pain points in enterprise AI adoption: hardware capital costs, operational complexity, and token consumption—all of which scale rapidly as AI moves from small pilots to production workloads with larger models, higher traffic, and greater GPU demand.
VCF 9 introduces efficiency-focused features to boost resource utilization. NVMe memory tiering leverages high-speed NVMe storage to expand effective memory capacity, while cluster-wide deduplication reduces redundant data across the environment, optimizing resource usage for memory-heavy AI workloads.
The platform supports heterogeneous multi-vendor infrastructure, including CPUs, GPUs, and other accelerators, alongside mainstream OEM/ODM server hardware. This flexibility lets enterprises select hardware based on workload needs, availability, and budget rather than rigid, single-vendor configurations.
New AI resource management tools enhance observability and control. Token monitoring tracks model usage and inference activity; multi-tenant model sharing enables shared access to common AI services; refined GPU and vGPU tracking delivers granular accelerator allocation and utilization data. A centralized AI metrics dashboard consolidates all workload performance and consumption insights.
VMware AI Factory Accelerates AI Production Deployment
VMware AI Factory serves as the software-defined foundation of Private AI Cloud, streamlining AI-ready infrastructure deployment and cutting the time required to transition from hardware setup to production model endpoints.
It automates end-to-end AI operations, including infrastructure provisioning, model service deployment, and post-launch Day 2 workload management, while improving visibility into token economics and resource consumption as AI usage scales. Built on vLLM, the VCF model runtime provides a unified serving layer supporting over 150 open-source and commercial models, enabling enterprises to select models based on language support, context window size, multimodal features, reasoning capability, and hardware compatibility.
Validated Partner Models Streamline Private AI Deployment
Broadcom has certified models from Google, NVIDIA, NEC, Alibaba Cloud, and Z.ai for VCF, offering enterprises a validated path for on-premises model deployment and internal Model-as-a-Service delivery. The key validated models include:
NVIDIA Nemotron 3: Hybrid Mamba-Transformer MoE multimodal models with up to 1 million-token context windows,
built for long-running enterprise agentic workflows.
Google DeepMind Gemma 4: Open-weight multimodal models optimized for local deployment and autonomous agent
development on private infrastructure.
NEC cotomi: Japanese-language optimized model with curated training datasets and 40% improved token efficiency.
Alibaba Cloud Qwen 3.7-Max: Proprietary multimodal model with 1 million-token context windows, tailored for agentic r
easoning workloads.
Z.ai GLM 5.2: Open general language model supporting local coding, multi-step reasoning, and private autonomous workflows.
This validation program reduces deployment uncertainty for on-premises AI, with partners working to deliver certified models via integrated VCF services. Broadcom’s 2026 Private Cloud Outlook found 56% of enterprises are running or planning production AI inference on private clouds, and VCF supports these deployments alongside existing VM, Kubernetes, and container workloads. Independent MLPerf Inference v5.1 testing confirmed VCF delivers bare-metal-equivalent performance, eliminating the need for separate dedicated AI infrastructure stacks.
Defense-in-Depth Security for AI Workloads
VMware Private AI Cloud adopts a NIST Cybersecurity Framework 2.0-aligned defense-in-depth strategy, combining infrastructure controls, network segmentation, application protection, supply-chain security, and continuous compliance.
VMware vDefend provides hypervisor-level microsegmentation and lateral movement protection. Virtual patching mitigates known vulnerabilities without application changes or infrastructure downtime, while traffic monitoring detects anomalous cross-workload communication and unauthorized shadow AI deployments. Updated vDefend capabilities now monitor agentic AI behavior, identify unauthorized AI system usage, and deploy AI-generated IPS virtual patches.
VMware Avi Load Balancer adds WAF and API security for AI workloads, restricting agent access to unauthorized tools, detecting zero-day anomalous behavior, and enforcing data protection rules to prevent sensitive data exfiltration.
TrueSource Secures AI Software and Data Supply Chains
Broadcom’s new TrueSource capabilities secure private AI software and data service supply chains. TrueSource Trusted Artifacts applies clean-room build processes to Java, Python, Node.js ecosystems and Bitnami Secure Images, with curated Spring Enterprise releases featuring human-verified, model-scanned security patches across release lines.
TrueSource Data Services extends this security framework to core data infrastructure, supporting PostgreSQL, RabbitMQ, MySQL, and Valkey to deliver secure, controlled foundational components for AI and agentic workflows.
Tanzu Platform Delivers Autonomous Agent Controls
VMware Tanzu Platform acts as the application and agent runtime layer for Private AI Cloud, with new AI-ready data foundations and agent governance tools. Autonomous agents pose unique risks including unauthorized data access, over-scoped tool usage, data exposure, and unplanned cloud costs, which Tanzu addresses with a deny-by-default runtime model.
Agents require explicit approval for all API, network, server, and internet access. An isolated credential store prevents credential theft, misuse, and prompt injection attacks. Prebuilt developer tooling includes approved skill sets, workflow buildpacks, human-in-the-loop oversight, and integrated memory services to enable secure agent development and execution.
AgentMinder Enables Centralized AI Agent Governance
Broadcom’s new AgentMinder serves as a central control plane for autonomous agents, assigning each agent a unique enterprise identity tied to defined missions, approved tools, and authorized resources. It enforces least-privilege runtime policies for tool invocation, delivers comprehensive agent activity audit logs, and provides full compliance visibility for multi-agent enterprise deployments.
AI-Ready Data Foundations Improve Context and Efficiency
Tanzu’s AI-ready data foundations process structured and unstructured enterprise data entirely within private infrastructure, generating curated, context-rich data products for AI agents without exfiltrating source data. These governed data products are publishable to the Tanzu marketplace for authorized developer and agent consumption, with data owners retaining full control over data preparation and exposure.
This approach boosts agent context quality, reduces reliance on uncurated raw data, lowers token consumption with more relevant model inputs, and keeps all sensitive data, metadata, and compute resources within the enterprise security perimeter.
Availability
All new VMware Tanzu Platform AI capabilities announced at VMware Explore will reach general availability in Fall 2026.
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!