Karman Platform Fits 16 NVIDIA B300 Servers Into 10-Server Power Envelope at Montreal Data Center

NexGen Cloud, a European AI cloud provider, has completed a deployment with embedded AI company Karman in which the Karman power orchestration platform allowed 16 NVIDIA B300 servers to run within a power budget originally provisioned for 10 servers.

The deployment, carried out in a commercial data center cluster in Montreal managed by NexGen Cloud, delivered 60% more nameplate compute capacity and token throughput within the same provisioned power budget, according to Karman, the company formerly known as Utilidata and based in Ann Arbor, Michigan. Following the Montreal results, the two companies said they are extending their partnership for a broader rollout across NexGen Cloud's European operations.

Measured Results From Varied Workloads

The Montreal deployment tested Karman across a range of power profiles, from steady, moderate loads such as LLM conversations to sustained high-power loads such as LLM summarization.

Karman reported three outcomes from those scenarios. First, the platform enabled 60% more nameplate compute capacity, with 16 NVIDIA B300 servers operating within the power budget set aside for 10. Second, token throughput rose 60% over the 10-server baseline while remaining inside the same power budget. Third, the company reported no measurable reduction in inference quality, including token latency, during sustained conversational workloads.

Karman said it held this performance under a variety of workloads, including scenarios with sustained high-power demand. Earlier this year, the two companies had announced plans to unlock more AI compute from existing power capacity and strengthen infrastructure resilience by introducing real-time, AI-driven power orchestration at the rack level.

Tapping Power Held in Reserve

Karman CEO Josh Brumberger framed the outcome around capacity that data centers keep on standby. "The fastest, low-cost way to scale AI today is by getting more compute out of the infrastructure already built," Brumberger said.

"Data centers hold power in reserve for worst-case scenarios that rarely happen. Karman turns that stranded capacity into usable compute. We're excited to bring this value to NexGen Cloud and their customers."

The company described the results as a scalable answer to what it called the biggest constraint facing the AI and energy sectors. As demand for AI compute climbs, data center operators face a mismatch between available AI infrastructure and the power capacity needed to run it, Karman noted. Adding new electrical capacity can require significant time and investment, whereas the company positions its platform as a way to address that constraint today by unlocking trapped capacity.

Efficiency as a Customer Value Driver

NexGen Cloud CEO Chris Starkey said efficiency is becoming as important as capacity as demand for AI compute accelerates. "The results we've achieved with Karman demonstrate that we can deliver significantly more compute from our existing infrastructure without requiring additional provisioned power," Starkey said.

He added that the outcome creates an opportunity to improve infrastructure utilization, accelerate deployment and ultimately deliver greater value to customers. "We're excited to build on these results as we expand our partnership across Europe," he said.

Millisecond Control Built on Custom NVIDIA Hardware

The Karman platform is powered by a custom NVIDIA module and provides the visibility, controls, and safety mechanisms that data center operators need to fully use their provisioned power capacity. According to the company, its software-defined controls respond within 20 milliseconds to continuously optimize power, and act within 2 milliseconds to limit power when demand threatens to exceed the budget.

Karman describes itself as providing dynamic power orchestration for AI data centers, enabling operators to better utilize the power already available to them. By embedding real-time visibility and millisecond control directly into power distribution equipment, the company says it increases tokens per watt by at least 50% by unlocking stranded power.

The platform is built on a custom NVIDIA Jetson Orin Nano and was previously embedded into smart meters to bring real-time visibility and power-flow control to the electric grid.

The Companies Behind the Rollout

NexGen Cloud describes itself as the foremost provider of sovereign-by-design, full-stack AI cloud environments for the global market. Its offerings include Hyperstack Secure Private Cloud for single-tenant enterprise deployments and Hyperstack On-Demand, a full-stack AI cloud platform offering GPU-as-a-service.

As an NVIDIA Cloud Partner, the company hosts its managed infrastructure in Tier 3 data centres across Europe and North America. The Montreal announcement follows another recent Karman partnership. On October 6, 2026, Karman announced a new partnership with Legrand, a global leader in physical infrastructure, aimed at increasing data center compute capacity.