_edited.png)
Case Study:
Delivering 1,200 Intelligent PDUs for a 500-Rack NVIDIA Blackwell AI Cluster in Just 8 Weeks
Project Overview
A leading European neo-cloud operator selected Link Power to deliver 1,200 intelligent PDUs for a 500-rack NVIDIA Blackwell AI deployment, supporting approximately 75kW per rack across 37.5MW of AI compute capacity.
The project required a fully managed power distribution platform with per-outlet monitoring, remote switching and dual-PDU redundancy, all delivered within a fixed eight-week programme.
Link Power successfully delivered the complete deployment on schedule, enabling the customer to achieve its planned go-live date with zero PDU-related programme delays.
Introduction
A leading European neo-cloud operator selected Link Power to supply 1,200 intelligent PDUs for one of Europe's largest NVIDIA Blackwell AI deployments.
The project required power distribution capable of supporting approximately 75kW per rack across 500 AI racks, with every unit manufactured, factory tested and delivered within an uncompromising eight-week programme.
With GPU hardware committed and a fixed go-live date, failure was not an option. Link Power delivered every unit on schedule.
GPU Platform
NVIDIA B300 Blackwell Ultra
Rack Density
75kW per rack
Solution
125A Three-Phase Fully Managed PDU
Industry
AI / HPC Data Centre
Delivery
1,200 Intelligent PDUs
Programme
Completed in 8 Weeks
The Challenge
Supporting Europe's Next Generation AI Infrastructure
The customer was deploying a 500-rack NVIDIA Blackwell AI cluster, designed for high-performance AI training and inference.
Each rack housed fully populated 8-GPU servers delivering sustained power densities of approximately 75kW, creating one of the most demanding power environments in Europe.
At this scale, power distribution becomes mission critical.
The customer required:
-
Complete remote monitoring
-
Per-outlet switching.
-
High-density power distribution
-
Dual PDU redundancy
-
Guaranteed delivery within eight weeks
Failure to deliver on time would have delayed the entire programme.
Why Link Power Was Selected
Quality
Purpose-built intelligent PDUs engineered specifically for high-density AI workloads.
Every unit was factory tested before shipment and verified to IEC billing-grade accuracy standards.
The customer also completed a factory audit, providing complete confidence in Link Power's manufacturing processes and quality controls.
Delivery
With GPU hardware already committed, delivery dates could not move.
Rather than providing a simple quotation, Link Power produced a fully transparent manufacturing schedule before production began.
This gave the customer complete confidence in programme delivery.
Technical Capability
The deployment required a platform capable of supporting approximately 75kW per rack, intelligent monitoring, outlet-level switching and seamless integration into the customer's DCIM environment.
Link Power already had a platform designed specifically for these requirements.
Quality
Purpose-built intelligent PDUs engineered specifically for high-density AI workloads.
Every unit was factory tested before shipment and verified to IEC billing-grade accuracy standards.
The customer also completed a factory audit, providing complete confidence in Link Power's manufacturing processes and quality controls.
Delivery
With GPU hardware already committed, delivery dates could not move.
Rather than providing a simple quotation, Link Power produced a fully transparent manufacturing schedule before
production began.
This gave the customer complete confidence in programme delivery.
Technical Capability
The deployment required a platform capable of supporting approximately 75kW per rack, intelligent monitoring, outlet-level switching and seamless integration into the customer's DCIM environment.
Link Power already had a platform designed specifically for these requirements.
Programme Timeline
Weeks 1–2
Final specification approval
Production scheduling
Engineering review
Weeks 3-6
Manufacturing
Factory Testing
First Shipment Dispatched
Weeks 6-8
Phased Deliveries
Customer Installation
Full Programme Completed
The Technical Solution
High-Density Intelligent Power Distribution
Link Power supplied a custom 125A three-phase fully managed PDU designed specifically for AI and GPU infrastructure.
Every rack received dual intelligent PDUs providing full redundancy together with per-outlet monitoring and switching.
Why Intelligent Outlet-Level Management Matters
High-density AI environments require far more than simple power distribution.
Link Power's intelligent PDUs provide:
-
Real-time outlet level monitoring
-
Individual outlet reboot capability.
-
Threshold alerts before circuit limits are reached.
-
Controlled power sequencing
-
Native DCIM integration
-
Improved energy efficiency through detailed outlet-level data
These capabilities reduce operational risk while giving infrastructure teams complete visibility across every rack.
Results
1,200
Intelligent
PDUs delivered
500
High-density AI racks powered
37.5MW
Total AI
compute capacity
8-Weeks
Complete programme
delivery
Zero
PDU-related delays
Project Outcomes
-
1,200 fully managed PDUs delivered on programme
-
First shipment completed after six weeks
-
Full deployment completed in eight weeks
-
500 racks commissioned with dual-PDU redundancy
-
Zero programme delays
-
Factory quality audit successfully completed
-
Customer engineering teams supported from specification through commissioning
Customer Feedback
"Link Power were transparent every step of the way, they already had PDU designs tailored for high-density AI stacks and, most importantly, provided a detailed proposal and schedule to deliver within our requirements. Delays would have prevented us from going live — but Link Power went the extra mile and delivered."
Head of Data Centre Infrastructure
Leading European Neo-Cloud Operator
Looking to Build High-Density AI Infrastructure?
Whether you're designing a new AI cluster or upgrading an existing facility, Link Power provides intelligent power distribution engineered specifically for high-density GPU and HPC environments.