Utilidata: Bringing AI Power Control to NexGen Data Centres

Utilidata and NexGen Cloud have announced the deployment of the Karman AI power control platform across NexGen Cloud’s data centres.

It is a power-efficient move which is focusing on improving the way electrical capacity supports AI workloads, allowing facilities to extract more computing power from existing infrastructure.

The companies say the platform is increasing available AI compute capacity within current grid constraints.

Through intelligent GPU power management and the reuse of stranded energy, the system targets up to 50% additional usable capacity.

Karman acts as an intelligence layer within data centre power systems, and it runs on a custom NVIDIA module.

The module balances performance and efficiency across AI workloads while monitoring power use in real time.

This approach essentially places optimisation directly at the rack level, where servers and graphics processing units operate.

AI power limits inside data centres

The partnership addresses a core challenge facing AI infrastructure: electricity availability.

Data centres that use AI for their operations require large numbers of GPUs.

However, grid connections and facility power systems limit how many processors operators can install.

Even when capacity exists on paper, much of it remains unused because operators must maintain safety margins to avoid power spikes or instability.

Josh Brumberger, CEO of Utilidata, says: “Power is the single biggest constraint for AI growth today. Billions of dollars of existing capacity are underutilized in data centres worldwide.

“Karman unlocks that trapped capacity immediately while ensuring new data centres are optimized from day one.

“We’re excited to partner with NexGen Cloud, a neocloud that understands the urgency of maximizing today’s infrastructure and the enormous opportunity in building tomorrow’s most effectively.”

NexGen Cloud runs its cloud platform with Hyperstack across Europe and North America.

The infrastructure supporting Hyperstack operates on 100% renewable energy across its European and Canadian data centres.

The parent company is also an NVIDIA Cloud Partner, which links cloud providers with NVIDIA’s ecosystem of AI hardware and software.

This relationship allows NexGen Cloud to integrate new compute platforms while expanding its AI infrastructure footprint.

Rack–level AI power optimisation

Karman is implementing AI directly into the electrical layer of a data centre. 

Sampling electrical signals more than one million times per second, the module reacts with submillisecond latency.

This speed allows the system to observe power changes as quickly as the servers themselves.

This essentially means the platform monitors fluctuations created by GPUs when workloads scale up or down.

It then adjusts distribution and control systems to keep power use within safe limits.

These data-driven responses allow operators to install more GPUs within the same electrical envelope.

Rather than expanding substations or waiting for grid upgrades, facilities extract additional compute from equipment already in place.

The architecture introduces optimisation across several layers of a facility.

GPUs continue to manage compute workloads at the server level while Karman manages electrical delivery across racks, rows and entire buildings.

Expanding Hyperstack capacity

The deployment of NexGen Cloud connects directly with the expansion of Hyperstack’s secure private cloud platform.

The company plans AI Factory deployments across North America and the Nordic region, which are designed specifically for training and running AI models.

Within these environments efficient power use determines how much compute capacity operators can offer customers.

By integrating Karman into its infrastructure, NexGen Cloud expects to improve energy performance while increasing available GPU resources.

Chris Starkey, CEO of NexGen Cloud, says: “We now live in a world where 100MW+ projects are becoming commonplace.

“For infrastructure providers like NexGen Cloud servicing such projects, the next critical factor is energy management – specifically, maximising the capacity already on-site.

“Karman gives us a significant commercial advantage by unlocking that trapped capacity – proving this level of optimisation is a benefit we can then carry over directly to our end customers.”

The same optimisation also extends to Hyperstack’s on demand compute services.

These platforms support AI training, machine learning development and visualisation workloads that rely on GPU acceleration.

The first deployment is taking place in a showcase data centre facility in Montreal.

After this installation the companies plan a broader rollout across NexGen Cloud’s European operations.