Infrastructure Software Watch: VMware AI Factory, AI SRE Agents Race, Kubernetes & Data Infrastrucure
Infrastructure Software Watch · November 4, 2026
Cloud, Kubernetes, AIOps & Enterprise Platforms — Why the Software Control Layer Is Becoming the Real Infrastructure Battleground
Who Controls the Software Layer That Operates Modern Enterprise Infrastructure?
The enterprise infrastructure conversation has shifted. For a decade, the question was whether to run workloads in public cloud, private cloud or on-premises. In 2026, the question is who supplies the software that controls, automates, secures and operates those environments as AI workloads converge with traditional IT. Broadcom is repositioning VMware Cloud Foundation as the operating layer for enterprise AI, with VMware AI Factory automating the path from bare metal to first model deployment and AgentMinder providing a governance plane for autonomous agents. Red Hat shipped OpenShift 5.0 security updates on November 3 and is preparing the OpenShift 5 platform launch later in 2026. Pulumi made its Terraform state backend generally available on November 2, absorbing the incumbent IaC ecosystem rather than competing with it head-on. Dynatrace, Datadog and New Relic have all put AI SRE agents into market within months of each other, turning observability into an autonomous operations category. And VMware Data Services Manager 9.1.1 now runs as a native VCF service with automated PostgreSQL, MySQL and SQL Server lifecycle management.
The through-line: the software control layer is consolidating around a small number of platforms, and each is trying to become the single plane through which enterprises govern AI infrastructure, traditional workloads and the agents that operate both. The winners will not necessarily be the vendors with the best individual tools. They will be the ones that own the control plane — identity, policy, observability and automation — across the entire estate.
Editorial Note: This edition examines verified developments from October 31–November 4, 2026, alongside earlier 2026 announcements that remain strategically relevant. Each item states its original date. This is not a list of announcements made on November 4 itself.
This Edition's Stories
VMware AI Factory Aims to Automate the Path From Bare Metal to First Model Deployment
Announced August 31, 2026; VMware Technology Briefing scheduled November 4, 2026.
What happened: At VMware Explore 2026, Broadcom introduced the VMware Private AI Cloud, a single private platform for running inference workloads, agentic applications, containers and traditional virtual machines. The centerpiece is VMware AI Factory, a software-defined foundation that automates infrastructure provisioning and lifecycle operations. Broadcom says VCF can reduce the path from bare metal to first model deployment from weeks to hours, and a MetalSoft integration extends that automation to heterogeneous physical hardware with the goal of reducing bare-metal provisioning from weeks to minutes. The stack includes VCF 9, which reduces hardware costs through NVMe memory tiering and cluster-wide storage deduplication; Validated AI Models for VCF, offering more than 150 open-source and commercial models on-premises; AgentMinder, a central control plane that binds agent authority to specific missions and enforces least-privileged access; and agentic security features in vDefend and Avi Load Balancer. Broadcom is holding a VMware Technology Briefing on November 4 covering Private AI capabilities, Multi-tenant Models-as-a-Service and AI Factory.
Key numbers: VMware AI Factory · VCF 9 · bare metal to model in hours vs weeks · bare metal provisioning in minutes · 150+ validated models · AgentMinder · vDefend · Avi Load Balancer · NVMe memory tiering.
Why it matters: Broadcom is making a deliberate strategic bet: private infrastructure is not a legacy alternative to public cloud, but a deliberate architecture for enterprise AI. That argument rests on three claims — data sovereignty, predictable token economics and Day 2 operational simplicity. The AI Factory product is the execution mechanism, because the hardest part of private AI is not acquiring GPUs but standing up the stack around them. If Broadcom can genuinely collapse weeks of integration into hours, it changes the calculus for enterprises that have been defaulting to public cloud for AI workloads. The risk is that the abstraction layer becomes another proprietary control point, locking customers into VCF as tightly as they are locked into any hyperscaler. Connect to: Cloud Computing Watch · Enterprise AI Intelligence · AI Infrastructure Watch
Datadog, Dynatrace and New Relic All Shipped AI SRE Agents Within Months of Each Other
Vendor releases throughout 2026; comparative analysis published October 2026.
What happened: The three major observability vendors converged on autonomous operations in 2026. Dynatrace launched domain-specific agents in January and added an Autonomous SRE Agent and Cloud SRE Agent in July, with the former triggering autonomously on newly detected problems and the latter coordinating remediation across AWS, Azure and Google Cloud. Datadog released Bits AI SRE, an autonomous agent that investigates alerts, and acquired Adaptive ML in June. New Relic shipped an SRE Agent, an agentic platform for building custom observability agents, AI Evaluation for production safety in October, and OpenTelemetry-based AI monitoring with prompt and completion capture off by default. AWS's DevOps Agent reached general availability in April 2026. Early AWS preview customers reported roughly 75% lower mean time to resolution and 94% root cause accuracy.
Key numbers: Dynatrace Autonomous SRE Agent · Datadog Bits AI SRE · New Relic SRE Agent · AWS DevOps Agent GA April 2026 · ~75% lower MTTR · ~94% root cause accuracy · prompt capture off by default.
Why it matters: Observability is no longer just about dashboards and alerts. The vendors are competing to become the operational control layer for AI-driven enterprises — the system that not only explains what went wrong but proposes and increasingly executes the fix. That is a significant expansion of scope, and it raises the same governance question as every other agentic system: what is the agent allowed to do without human approval? Policy-as-code frameworks like Open Policy Agent are gaining traction as the governance layer between an agent's decision and its execution. The practical distinction for buyers is between AI that explains and recommends, which is low-risk and worth adopting broadly, and AI that acts autonomously, which needs an explicit policy layer and an audit trail. Connect to: Enterprise AI Intelligence · Cybersecurity Watch · Enterprise Software Intelligence
Red Hat Ships OpenShift 5.0 Security Updates as It Prepares the Platform's Next Major Version
OpenShift 5.0 security advisory November 3, 2026; OpenShift 5 roadmap disclosed October 2026.
What happened: Red Hat published security advisories for OpenShift Container Platform 5.0 on November 3, confirming that the next major version is now in customer hands, even as the company prepares a broader OpenShift 5 launch later in 2026. Red Hat's published roadmap says OpenShift 5 will embed applied AI experiences to simplify operations, enhance troubleshooting, support post-quantum cryptography, and offer extended lifecycles and in-place upgrades from OpenShift 4. OpenShift 4.22, released in July 2026, already added security, virtualization and AI capabilities, including OpenShift Sandboxed Containers 1.12. OpenShift Virtualization has grown 417% in virtual machines for the 2025/2026 period, and Azure Red Hat OpenShift 5.8 now unifies containers, VMs and AI governance.
Key numbers: OpenShift 5.0 · security advisory November 3 · applied AI · post-quantum cryptography · in-place upgrade from OpenShift 4 · OpenShift Virtualization +417% · Azure Red Hat OpenShift 5.8.
Why it matters: OpenShift is Red Hat's bid to be the enterprise application platform for both containers and virtual machines, and the 417% growth in OpenShift Virtualization shows that customers are using it to consolidate VM estates rather than just containerize new applications. That matters strategically because it puts Red Hat in direct competition with VMware for the virtualization layer. OpenShift 5 adds a second dimension: applied AI embedded in the platform to simplify operations and troubleshooting, which is the same control-layer thesis Broadcom and the observability vendors are pursuing. Whoever owns the platform that runs the workloads also controls the policy, identity and automation around them. Connect to: Cloud Computing Watch · Enterprise Software Intelligence · Hardware Watch
Automation and Data Infrastructure
Pulumi Makes Terraform State Backend Generally Available, Letting Teams Keep Their HCL While Migrating Control
Generally available November 2, 2026.
What happened: Pulumi made its Terraform state backend generally available on November 2, allowing organizations to store and manage Terraform state in Pulumi Cloud alongside their Pulumi stacks. Teams can migrate from S3, Azure Blob, GCS or local state by updating their backend block and running terraform init -migrate-state. Pulumi Cloud adds update history, state locking, role-based access control, audit policies and unified resource visibility on top of the Terraform state. The company had already made HCL a first-class Pulumi language in August 2026, alongside TypeScript, Python, Go, .NET, Java and Pulumi YAML, and added full support for Terraform modules and OpenTofu compatibility.
Key numbers: Pulumi · Terraform state backend GA November 2 · HCL first-class since August 2026 · Terraform and OpenTofu compatible · state locking · RBAC · audit policies.
Why it matters: This is a clever competitive move. Rather than trying to displace Terraform, Pulumi is making itself the control plane for the Terraform estate that already exists. Enterprises do not have to rewrite their infrastructure code; they can keep their HCL and still gain Pulumi's platform capabilities. The strategic risk for HashiCorp, now part of IBM, is that the value of Terraform shifts from the tool to the state management and governance layer — exactly where Pulumi is positioning. For platform engineering teams, it means the choice between IaC tools is becoming less about syntax and more about which control plane they want to standardize on. Connect to: Enterprise Software Intelligence · Cloud Computing Watch · Tech M&A Watch
VMware Data Services Manager 9.1.1 Becomes a Native VCF Service With Automated PostgreSQL and MySQL Lifecycle Management
Released September 3, 2026; included in VCF 9.1.1.
What happened: VMware Data Services Manager 9.1.1 is now a native service within VMware Cloud Foundation, meaning its lifecycle, upgrades, rollbacks and recovery are managed as part of VCF rather than as a separate tool. The release adds PostgreSQL 18 support, automated major and minor upgrades for MySQL, SQL Server 2025 with cumulative update support, and API-driven on-demand backups. It also adds PostgreSQL read replicas and support for VMware vSphere Kubernetes Service 3.7 and multi-cluster Kubernetes, abstracting the Kubernetes complexity that has historically kept database teams away from modern data services. The earlier 9.1 release had already introduced production-ready PostgreSQL with native pgvectorcapabilities for RAG, semantic search and embeddings.
Key numbers: Data Services Manager 9.1.1 · native VCF service · PostgreSQL 18 · MySQL automated upgrades · SQL Server 2025 CU support · PostgreSQL read replicas · VKS 3.7 · pgvector.
Why it matters: Databases are where infrastructure software meets the data that AI systems depend on, and they are the part of the stack that infrastructure teams most often hand off to a separate DBA organization. By making DSM a native VCF service with automated lifecycle management, Broadcom is trying to bring database operations into the same control plane as the rest of the private cloud. The pgvector support is particularly important: it means enterprises can run vector search and RAG workloads on PostgreSQL inside their own VCF environment rather than exporting data to a separate vector database. That is the private AI argument applied to the data layer.
Connect to: Enterprise AI Intelligence · Cloud Computing Watch · Enterprise Software Intelligence
Why It Matters
The infrastructure software market is consolidating around control planes, not tools. Broadcom wants VCF to be the plane that governs AI workloads. Red Hat wants OpenShift to be the plane that runs containers and VMs. Pulumi wants Pulumi Cloud to be the plane that manages infrastructure state, even when the infrastructure code is Terraform. The observability vendors want their agents to be the plane that investigates and remediates incidents. Each is expanding from its original category into the layer above it.
AI is the forcing function, not the product. Every vendor in this edition is using AI as the justification for expanding its scope. Broadcom's AI Factory automates model deployment. Red Hat's OpenShift 5 embeds applied AI to simplify operations. Dynatrace, Datadog and New Relic use AI agents to investigate incidents. Pulumi frames its roadmap as bringing Terraform estates into the "agentic era." The AI narrative is doing a lot of commercial work here, and buyers should evaluate the underlying automation and governance capabilities separately from the AI branding.
The governance question is the same at every layer. Whether the agent is deploying a model, remediating an incident or provisioning infrastructure, the question is what it is allowed to do without human approval. Policy-as-code frameworks like Open Policy Agent are emerging as the neutral layer between the agent's decision and the action. Enterprises that standardize on a policy layer early will have more freedom to adopt agents from multiple vendors without losing control.
Private infrastructure is getting a second life. Broadcom's core argument — that enterprises should bring models to their data rather than exporting data to external AI services — is resonating because production AI introduces sovereignty, compliance and token-economics concerns that experimentation did not. That does not mean enterprises will abandon public cloud. It means the default architectural decision is shifting from "public cloud unless there is a reason not to" to "where should this workload run, based on data sensitivity, cost and governance."
The database layer is the quiet consolidation story. DSM 9.1.1's native VCF integration and pgvector support mean that the private cloud control plane now extends to the data layer. That is where AI workloads actually live, and it is the layer that has historically been the hardest to standardize across infrastructure and database teams.
What to Watch
- VMware Technology Briefing (November 4): What Broadcom discloses about Multi-tenant Models-as-a-Service, AI Factory and AgentMinder beyond the Explore announcements.
- OpenShift 5 general availability: Whether the platform launch lands in 2026 and how the applied AI and post-quantum cryptography features perform in production.
- Pulumi Terraform adoption: Whether enterprises actually migrate Terraform state to Pulumi Cloud, and how HashiCorp and IBM respond.
- AI SRE agent governance: Whether Policy-as-Code adoption accelerates as agents move from recommendation to autonomous action.
- DSM pgvector workloads: Whether enterprises run RAG and semantic search on PostgreSQL in VCF rather than adopting a separate vector database.
- Broadcom AI Factory customer wins: Named enterprise deployments and whether the bare-metal-to-model timeline claim holds in practice.
- Observability vendor consolidation: Whether Datadog, Dynatrace or New Relic acquires another capability to close a gap in their agentic operations story.
- Kubernetes platform competition: How SUSE, Mirantis and VMware respond to Red Hat's OpenShift 5 roadmap.
- Private AI token economics: Whether enterprises can demonstrate measurable cost advantages from running inference on their own infrastructure versus public AI services.
- Policy-as-Code standards: Whether Open Policy Agent or a competing framework becomes the default governance layer for agentic infrastructure operations.
The software control layer is where the infrastructure battle will be decided. The vendors that own the plane through which enterprises govern identity, policy, observability and automation will capture the durable economics of enterprise infrastructure, regardless of which cloud or chip wins.
Broadcom's VMware AI Factory is the most ambitious version of this thesis. It is not trying to sell AI features; it is trying to make VCF the operating layer for enterprise AI. If Broadcom can genuinely reduce the path from bare metal to first model from weeks to hours, it changes the economics of private AI deployment. The risk is that the abstraction layer becomes another lock-in, replacing one form of dependency with another.
Pulumi's Terraform state backend is the smartest competitive move in the edition. Instead of fighting Terraform, Pulumi is absorbing its state layer and making itself the control plane for the infrastructure code that already exists. That is a much faster path to enterprise adoption than asking customers to rewrite their HCL.
The AI SRE agent race is real, but the governance is not keeping up. Datadog, Dynatrace and New Relic have all shipped autonomous investigation and remediation agents within months of each other. The vendors that separate "explains and recommends" from "acts autonomously," and provide the policy layer and audit trail to govern the latter, will win the enterprise deals.
Red Hat's OpenShift 5 and VMware's DSM 9.1.1 show that the platform and data layers are converging. The control plane is extending downward into databases and upward into AI operations. The companies that can deliver a single governed plane across compute, data and agents will define the next generation of enterprise infrastructure software.
The CODEW Stat
417% — the growth in virtual machines running on Red Hat OpenShift Virtualization for the 2025/2026 period. The number matters because it shows that OpenShift is no longer just a container platform. Enterprises are using it to consolidate their VM estates, which puts Red Hat in direct competition with VMware for the virtualization layer. That competition is the subtext of this entire edition: Broadcom is repositioning VMware as the AI control plane, Red Hat is repositioning OpenShift as the unified platform for containers and VMs, and both are betting that the customer who consolidates on their platform will also buy their automation, observability and AI operations layers.
Sources: Broadcom, VMware Explore 2026, The Cube Research, Pulumi, Red Hat, Red Hat Security Advisories, Windows News, heise, Dev Community, VantraOps, New Relic, Dynatrace, IT Brief, VMware Blogs, Broadcom TechDocs, CormacHogan.com, and other English-language industry publications covering August–November 2026. Vendor-reported figures and product claims are labeled as such. This briefing does not provide investment advice.
Reported but unconfirmed developments are labeled accordingly. Product availability, pricing and roadmap dates are company guidance and may change. This edition examines verified developments from October 31–November 4, 2026, alongside earlier 2026 announcements that remain strategically relevant. It is not a list of announcements made on November 4 itself.
Reviewed by Erwin Castro
on
Wednesday, November 04, 2026
Rating:
No comments: