Neuralwatt Increases AI Inference Throughput by 33% on Crusoe Cloud

July 01, 2025

Neuralwatt x Crusoe Cloud

Neuralwatt x Crusoe Cloud

By Marisa Krystian, Content Marketing at Crusoe

Founded by industry veterans with deep expertise in energy-efficient computing, Neuralwatt is on a mission to solve one of the AI industry's biggest challenges: reducing the environmental impacts of its staggering power needs. 

The startup provides intelligent AI power management software that allows customers to dramatically reduce their energy consumption and associated greenhouse gas emissions. How? By increasing their AI server density and efficiency without sacrificing performance.

The Challenge

As a new, bootstrapped startup with a visionary software solution, Neuralwatt faced a classic chicken-and-egg problem. They needed to create a compelling, data-driven demo in order to prove its value and attract early design partners and investors. This would require significant GPU compute resources, a major expense for an early-stage company. As co-founder Scott Chamberlin noted, for a startup at their stage, the cost of compute is a huge consideration, making the choice of a cloud partner a potential make-or-break decision.

Beyond cost, their software also needed deep, unrestricted access to GPU hardware-level telemetry to work its magic — a level of control that many cloud providers limit or do not offer. And as a company committed to the climate, they also wanted an AI cloud partner that was not only cost-effective but also deeply aligned with their values.

The Solution

Luckily, Neuralwatt was selected as one of nine companies in the highly competitive Climate Collective’s AI x Climate accelerator program. To provide access to advanced computing resources for AI companies developing innovative climate-positive solutions, Crusoe supported the accelerator program by giving participating companies Crusoe Cloud credits. This gave Neuralwatt access to the GPUs they needed to run their demo and pre-train their power optimization models. They were immediately able to spin up NVIDIA H100 clusters on Crusoe Cloud in a data center in Iceland that ran on 100% geothermal and hydro power.

“On Crusoe, we’ve had access to everything we needed," says Chad Gibson, co-founder of Neuralwatt. "We have access to all of the NVIDIA System Management Interface (SMI)... nothing has been restricted or unavailable to us, which isn't the case with a lot of the other providers we've tried. The ease of access to the NVIDIA hardware has been instrumental.”

This combination of cost-effective access, climate alignment, and technical freedom enabled Neuralwatt to quickly build a demo that brought their vision to life, turning abstract ideas into tangible proof of performance.

The Results

The impact of Neuralwatt’s software, unlocked by Crusoe Cloud, was immediate and dramatic. Their demo showcased a monumental leap in AI efficiency. By running their software on a DGX H100 system on Crusoe Cloud, Neuralwatt was able to:

  • Increase inference throughput by 33% while consuming less peak power.

  • Improve GPU density by 33%, making it possible to run eight GPUs in the same power envelope that would ordinarily only support six GPUs with stock operation.

  • Reduce idle GPU power draw by over 40%, from 125 watts down to 73 watts.

Crusoe’s support was a critical turning point for Neuralwatt. They were able to show concrete, game-changing performance gains. As co-founder Scott Chamberlin noted, "It was the credits that really enabled us to do that... it's made a lot of our conversations with investors and customers so much easier."

Looking Ahead: The Future of Efficient AI

With a powerful demo and proven results, Neuralwatt is now focused on scaling its impact. Their vision is to optimize thousands of compute units across the AI industry.

“At some point, the models that were trained in Crusoe will be responsible for gigawatts of power savings down the road,” Gibson explains.

Jack Policar of the Climate Collective agrees. "Seeing a company like Neuralwatt create this kind of breakthrough is exactly why we put the accelerator together," he says. "This is the story we want to tell: innovative founders using resources like Crusoe Cloud to tackle the massive challenge of AI's energy footprint. It’s the future of sustainable AI."

The partnership between Neuralwatt and Crusoe is a perfect example of how an AI-native infrastructure company can empower a visionary startup. By providing a reliable, cost-effective, and technically open platform, Crusoe enabled Neuralwatt to build the software that will help the entire AI ecosystem run more efficiently and unlock maximal GPU performance per watt of power — a win for innovation, economics, and the climate.


Liked what you just read? Share it

Ready To Build Something Amazing?