What Should Be Checked Before Adding AI or GPU Equipment to an Existing Data Center?

Before adding AI or GPU equipment to an existing data center, the facility should review whether its current power infrastructure has the capacity and configuration needed to support the planned computing load. That review should include the UPS system, battery capacity, backup generators, automatic transfer switches, electrical distribution and required redundancy.

For many facilities, an AI or GPU expansion is also an opportunity to add UPS capacity, upgrade existing equipment or expand backup generation around the data center’s next stage of growth.

Critical Infrastructure Group supplies UPS systems, commercial and industrial generators, automatic transfer switches and other critical-power equipment for data center expansion and infrastructure projects.

Start With the Power Requirements of the New AI or GPU Equipment

The first step is establishing the electrical requirements of the equipment being added.

AI and GPU systems are available in many configurations, so the facility should use the specifications for the actual servers, racks or integrated systems being deployed rather than applying a generic power assumption.

Useful information can include:

• Power requirement per server or system
• Planned number of systems
• Expected rack density
• Required voltage and phase
• Power-supply configuration
• Required number of power feeds
• Manufacturer redundancy requirements
• Expected future expansion

Once the new computing load is defined, it can be compared with the capacity available from the existing data center infrastructure.

Review Available UPS Capacity

The UPS system is one of the first areas to evaluate because it provides immediate power protection for the critical IT load.

The review should compare the existing UPS load with the additional AI or GPU capacity being planned. It should also account for the amount of capacity the data center wants to preserve for redundancy and future expansion.

If more capacity is appropriate, CI Group offers UPS systems across a broad range of capacities and manufacturers, including 1201+kVA UPS equipment for large critical-power applications.

For additional sizing information, see What Size UPS Does a Data Center Need?

Confirm the Power Architecture Required by the GPU Platform

Manufacturer requirements should be reviewed as part of the electrical design because different GPU platforms can have different power and redundancy configurations.

For example, NVIDIA’s DGX H100 data center design documentation includes specific requirements for rack power distribution, multiple power sources and facility-level UPS and generator backup. Those requirements apply to that particular platform and should not automatically be applied to every GPU deployment.

Review NVIDIA’s DGX H100 electrical design guidance.

For more on NVIDIA-focused redundancy planning, see Why Redundant Backup Power Matters for NVIDIA GPU Data Centers.

Evaluate Redundancy Along With Capacity

Adding enough capacity for the new equipment is only one part of the project. The facility should also determine how the additional load fits into its existing redundancy strategy.

If the data center uses N+1 or another redundant configuration, the expanded system should be planned so that the intended redundancy remains available at the new load level.

That may create an opportunity to add another UPS module, install additional UPS capacity or redesign portions of the critical-power architecture around the planned expansion.

See What Is N+1 Redundancy in a Data Center Power System? for more information.

Check UPS Battery Runtime at the New Load

If the UPS load increases, the battery configuration should also be reviewed against the data center’s desired runtime.

A facility may choose to expand compatible battery capacity, replace an existing battery system or incorporate new batteries into a larger UPS upgrade.

CI Group provides UPS Battery Replacement Services for existing critical-power systems.

For runtime planning, see How Much UPS Battery Runtime Does a Data Center Need?

Review Backup Generator Capacity

Additional IT and supporting facility load can also change the amount of capacity required from the backup generator system.

The generator review should consider the complete load that will be supported during backup operation, including the new computing environment and other designated critical facility loads.

For larger projects, this may create an opportunity to add generator capacity, expand a multi-generator configuration or replace equipment with higher-capacity units.

CI Group supplies 2001+kW generators for large data center and mission-critical projects.

See How Do You Size a Backup Generator for a Data Center? for more information about generator capacity planning.

Include Automatic Transfer Switches in the Review

The automatic transfer switch connects the facility’s normal and backup power sources, so ATS capacity and configuration should be considered as the data center load changes.

The project team should confirm that the transfer equipment matches the voltage, amperage, switching arrangement and backup-power configuration planned for the expanded facility.

CI Group supplies Automatic Transfer Switch equipment and provides ATS Service & Maintenance for existing installations.

AI Workloads Can Change the Electrical Load Profile

In addition to total capacity, the engineering team should consider how the planned computing equipment uses power.

Uptime Institute notes that large AI computing clusters can have electrical load characteristics that differ from more conventional server environments, particularly during large training workloads. The amount and behavior of that load depend on the equipment and configuration being deployed.

Review Uptime Institute’s electrical considerations for large AI compute.

This makes manufacturer specifications and project-specific electrical planning particularly useful when selecting UPS, generator and distribution capacity.

Inspect Existing Critical-Power Equipment Before Expansion

An expansion project is also a practical time to review the equipment already supporting the facility.

UPS inspections, battery testing and preventative maintenance can help document the current system before additional capacity is added. The resulting information can then be used to plan maintenance, component replacement or new equipment as part of the expansion.

CI Group provides UPS Preventative Maintenance & Inspection and broader UPS Maintenance Services.

Temporary UPS Equipment Can Support the Upgrade

If permanent UPS equipment needs to be replaced or expanded during the AI infrastructure project, temporary UPS capacity can be incorporated into the project plan.

CI Group offers UPS Rental Systems for equipment replacement, capacity upgrades, maintenance and other transitional power requirements.

See Can a Rental UPS Be Used During a Data Center Upgrade or Replacement?

What Should Be Reviewed Before Adding AI or GPU Equipment?

A power-infrastructure review for an existing data center should include:

• New AI or GPU equipment load
• Existing UPS capacity and current load
• Required UPS expansion or replacement capacity
• UPS redundancy configuration
• Battery runtime at the planned load
• Backup generator capacity
• Automatic transfer switch capacity and configuration
• Power-distribution requirements
• Manufacturer power-feed requirements
• Existing equipment condition and maintenance needs
• Temporary UPS requirements during the upgrade
• Future AI or computing expansion plans

With those requirements established, the facility can determine which portions of the existing infrastructure can support the expansion and where additional critical-power equipment would best support the project.

Build the Power Infrastructure Around the AI Expansion

Adding AI or GPU equipment can be an opportunity to increase data center power capacity while preparing the facility for future growth. UPS systems, batteries, generators, ATS equipment and redundancy can all be reviewed together so the permanent infrastructure is selected around the planned computing environment.

For a broader look at these requirements, see Power Infrastructure Requirements for AI Data Centers.

Critical Infrastructure Group supports these projects with UPS equipment, large-capacity generators, automatic transfer switches, temporary UPS rentals, battery replacement, preventative maintenance and nationwide Data Center Power Infrastructure solutions.