Ir para o conteúdo principal

Managed AI Factory

Overview​

Equinix Managed AI Factory is an AI infrastructure solution co-developed by Equinix and NVIDIA. It deploys NVIDIA’s Enterprise AI Factory reference architecture in select Equinix IBX data centers.

The solution supports a distributed AI architecture in which each workload type runs in the location that best fits its requirements:

  • Training - Runs where sufficient compute capacity is available.
  • Inference - Runs close to the source of the data, which reduces latency.
  • Sensitive workloads - Run in sovereign environments that meet data residency and regulatory requirements.

What's Included​

Colocation & Physical Infrastructure

Your AI Factory systems are hosted in Equinix IBX data centers, with dedicated power (30-240 kVA per rack), in-row liquid cooling, and structured cable management.

Connectivity Services - Ordered Separately

Equinix Fabric, Equinix Internet Access, and cross-connects are ordered separately as part of your solution configuration. The connectivity services, bandwidth, redundancy, and applicable charges are specified in your quote.

Managed Services

Equinix manages your AI Factory hardware (compute, storage, network switches) and software up to the OS level. This includes event, incident, and change management, security, and break-fix support.

Monitoring & Telemetry

Equinix Smart View is integrated with the management software to provide real-time power and environmental monitoring, with alerting to keep the managed services team proactively informed.

On-Site Support

Equinix Smart Hands offers 24/7 remote access, custom installations, and hardware troubleshooting at any IBX location.

Technical Support

You can work with NVIDIA directly for data and application layer support, while Equinix handles everything from the management software down to the OS in the IT stack and all power and cooling support in the data center.

Additional services available include: Managed Private Storage, Managed Backup, and Managed Firewall.

How it Works​

Architecture​

Managed AI Factory supports air-cooled and liquid-cooled AI infrastructure. The deployed architecture determines the applicable compute, networking, storage, management, power, and cooling components. Hybrid cooling architectures are supported where compatible with the platform, chipset generation, and infrastructure design.

A deployment can include NVIDIA DGX or Dell AI servers, network switches, storage and management infrastructure, cluster networking, Base Command Manager/Mission Control, Unified Fabric Manager, and the associated data center infrastructure.

Example architecture diagram

Equinix performs data center design, cage construction, rack installation, cabling, infrastructure installation, deployment, and testing as part of the applicable service scope. Day 1 includes fulfillment, activation and commissioning. Day 2 includes steady-state product, management, and support.

Managed Services Components​

Equinix Managed Solutions (EMS) provides infrastructure management for the DGX systems, DGX cluster, DGX-validated storage, InfiniBand networking, network switches, Equinix Fabric, and the software elements up to and including the operating system of the solution. Operations are delivered through Digital Infrastructure Management (DIM) and cover five areas:

  • Incident Management - Event and incident handling across the solution, plus break-fix services.

  • Monitoring, Performance, and Operational management - Monitoring, utilization, and service reporting across the infrastructure and software stack.

  • Change Management - Controlled changes to the infrastructure and software elements within the managed scope.

  • Security and Business Continuity - Security and continuity practices applied across the managed stack.

  • Governance, Intelligence, and Control - Oversight of the service, including service level agreements and reporting.

Responsibility Model​

Equinix is responsible for:

  • Hardware support and maintenance coordination
  • Monitoring
  • Reporting
  • Event, Incident, Problem and Change Management
  • Network management
  • Storage management
  • Operating system administration and patching
  • Firmware update execution
  • Physical access management
  • Capacity management
  • Security role administration.

You are responsible for:

  • Hardware ownership
  • Hardware procurement
  • Licensing
  • Job scheduling
  • AI orchestration
  • End-user support
  • Password and access management
  • Infrastructure scaling decisions

Use Cases​

Managed AI Factory is for organizations that want dedicated AI infrastructure that Equinix deploys and operates.

These use cases illustrate where Managed AI Factory can provide an infrastructure foundation.

  • LLM Training and Fine-Tuning: Managed AI Factory supports the development and customization of large language models (LLMs) across a range of deployment sizes. Organizations can use dedicated AI infrastructure to train models from scratch, fine-tune existing foundation models, or adapt models for specific business domains and applications.

    Example workloads include domain-specific model development, model training, model optimization, and enterprise LLMs.

  • Real-Time and Batch Inference: Managed AI Factory provides dedicated infrastructure for deploying trained AI models for both real-time and batch inference workloads. Organizations can use GPU resources to process high volumes of requests while maintaining control over the underlying infrastructure and data.

    Example workloads include batch processing, fraud detection, document intelligence, computer vision, conversational AI, and chatbots.

The same infrastructure can support development, testing, and production inference environments, depending on your organization's deployment architecture and resource requirements.

  • HPC and Simulation Workloads: Managed AI Factory supports GPU-accelerated scientific, engineering, and simulation workloads. Organizations can use the infrastructure for computationally intensive workloads that require significant parallel processing capacity.

    Example workloads include digital twins, scientific computing, engineering simulation and research and development.

AI and HPC workloads can share an infrastructure platform where appropriate, while providing the dedicated compute resources required for demanding workloads.

  • Agentic AI Workloads: Managed AI Factory supports agentic AI applications that use multiple AI models, tools, and services to execute complex tasks. These workloads can require substantial compute resources as agents reason over information, invoke models repeatedly, interact with enterprise systems, and coordinate with other agents.

    Example workloads include multiple agent orchestration, autonomous enterprise workflows, AI-powered software development, enterprise knowledge agents, and tool-using AI applications.

Dedicated infrastructure provides the compute capacity required when agentic applications generate substantially more model inference and orchestration activity than a conventional single-request AI application.

  • Secure and Sovereign AI: Managed AI Factory provides a dedicated infrastructure foundation for organizations with stringent security, privacy, data-residency, or regulatory requirements. Deploying AI infrastructure within an Equinix IBX data center can help organizations maintain greater control over where AI workloads and data are processed.

    Potential applications include financial services, healthcare, government, and regulated enterprise applications.

The specific compliance, sovereignty, and security capabilities available for a deployment depend on the selected Equinix location, infrastructure configuration, customer requirements, and applicable regulations.

  • Hybrid Multicloud AI: Managed AI Factory serves as a dedicated private AI infrastructure layer while connecting AI workloads to public cloud services through Equinix Fabric. This architecture allows organizations to combine dedicated AI compute with cloud-based applications, data, and services.

    Example architectures include private model training, cloud bursting, distributed data and AI workflows, multicloud inference, and enterprise hybrid AI.

Equinix Fabric provides the connectivity layer between the private AI infrastructure and connected cloud and network environments, helping organizations integrate Managed AI Factory into existing hybrid and multicloud architectures.

Getting Started​

Term​

The initial term is three years. After that, you can renew for up to two successive one-year periods.

Availability and Locations​

Managed AI Factory is available within select Equinix IBX data centers.

Global map demonstrating IBX locations

For more information about Managed AI Factory solutions, contact Equinix.

Esta página foi útil?