Can Your Data Centre Handle Its First AI Rack? Three Checks Before You Buy GPU Servers
Introduction
Looking to install GPU servers within your proposed AI rack? Before you do, it’s important to review the electrical infrastructure already in place and see if it can handle the increased load, as AI applications and hardware are typically power hungry.
It’s also vital to establish your existing cooling capacity and check whether your UPS and generator systems can provide adequate autonomy during a mains failure or planned maintenance.
Our blog will look further into these checks and provide you with tips and recommendations on how to establish whether your data centre is ready for AI implementation.
Can your electrical infrastructure deliver enough power to the AI Rack?
To begin with, it is crucial to establish the total capacity in kilowatts (kW) across all racks and the design capacity of the data centre. Your building may have supplementary electrical capacity available, but a correctly designed data centre will use dedicated rated local distribution boards for in rack power distribution units (PDUs) and separately for mission critical systems such as air conditioning, fire suppression, lighting, UPS and generator systems.
Once these figures have been confirmed, you will need to ask your hardware supplier for the minimum and maximum power draw in kilowatts (kW) of the AI hardware you’re looking to install (including any supporting systems such as networking equipment).
It is vital to check the backup supply too. If the existing set up utilises two independent power feeds (typically Feed A & B), establish what happens when one is unavailable. Can the remaining feed and its uninterruptible power supply (UPS) carry the intended load alongside the equipment you already run?

Reviewing peak and minimum IT loads on your existing UPS systems can help establish current critical IT demand and inform your plans for AI implementation.

Checking the available capacity of your distribution boards and the circuits supplying each rack helps establish whether your existing electrical infrastructure can support the additional AI hardware.
Is there enough cooling capacity for the new AI hardware?
Start by confirming how much cooling capacity your data centre has at peak load and what redundancy is available. Reviewing EMS readings and existing operating conditions alongside the manufacturers’ manuals will help confirm the cooling capacity of your existing systems.
Next look at the rack itself. Even if the room has adequate cooling, cold supply air must reach the front of the cabinet by checking the temperature at the inlet. Additionally, see where the hot air from the exhaust goes, if that hot air comes back into the rack or flows into a nearby cabinet the servers could overheat even if the cold aisle temperature sensor says the temperature is optimal. Using thermal imaging can reveal existing hot spots and airflow modelling can also show how the new AI loaded rack might affect equipment.

Blanking panels help prevent cooled air from escaping through unused rack spaces, improving airflow control and reducing unnecessary cooling demand.

Thermal imaging can help identify hot spots, airflow issues and temperature variations across racks before planning AI implementation in your data centre.
Can existing services stay online during maintenance or power outages?
Start with planned maintenance. If one cooling unit is taken offline during a planned service, can the remaining air conditioning systems keep the AI rack and surrounding servers within their operating temperature limits? The same question should be applied to the UPS systems should you have redundancy in place such as N+1.
it is critical to confirm which cooling systems the generator supplies and how quickly those cooling systems restart during the brief outage whilst the critical hardware remains online via the UPS system. If your site solely relies on UPS, check the UPS battery runtime at the proposed load and allow enough time for a controlled graceful shutdown. In either case, establish how long the supported racks can operate safely when cooling systems go offline.
Review AI implementation with data centre experts like us!
Before you approve the GPU server order, contact us today with the proposed rack specification, data centre design specification and any recent power or temperature readings. Our Data Centre Audit & Risk Review assesses electrical distribution, UPS provision and cooling, with findings and recommended improvements in a written report. That gives you a basis for deciding which upgrades to fund before installation. Book a Data Centre Audit & Risk Review today.
Contact us today on 0844 225 0094 to book an exploratory meeting or click below to fill in our contact request form:
Discover what our data centre consultancy services can offer.
Frequently asked questions about data centre AI implementation
What information do i require before deciding?
Review all proposed AI hardware specifications, rack layouts, electrical drawings with recent power and temperature readings. Maintenance records and UPS battery test results will also help identify where further checks are required.
Can I upgrade my data centre while it remains operational?
This depends on the work required and the redundancy available in the data centre. Some upgrades can be phased around live equipment, while others require planned shutdown. A site audit with one of our data centre experts will help establish a suitable installation sequence.
Can I start with a small AI deployment and expand later?
Yes, provided you account for expansion in the initial design. Share your expected growth with us today so we can assess future power, cooling and space requirements alongside the first installation.
What should I include in the budget beyond AI hardware?
Allow for any respective electrical and cooling upgrades, installation, commissioning and monitoring identified during the AI implementation assessment. Include ongoing energy and maintenance costs (including any work required to migrate existing equipment).
What should be checked before the new AI rack goes live?
Commissioning should confirm that the installation supports the agreed load and operating conditions. This includes checking power distribution, cooling performance, monitoring alarms and the response to a power failure through an agreed design.
How can Infiniti IT help prepare my data centre for AI?
Infiniti IT’s Data Centre Audit & Risk Review can assess your existing infrastructure against the proposed installation. The findings provide a basis for prioritising improvements, budgeting for upgrades and planning the work before your GPU servers arrive.








