Crusoe has introduced Crusoe Spark Liquid, the sixth generation of its modular AI factory platform, announced on October 21, 2026. The new system is engineered for ultra-high-density rack-scale systems, including NVIDIA GB300 NVL72 and NVIDIA Vera Rubin NVL72, and supports up to 3.5MW of compute per unit. Units are designed to be deployed in clusters to power AI training and inference at any scale.

Manufacturing Set for Colorado Facility in 2027

The units will be manufactured in Colorado, with Crusoe Spark Liquid joining existing Crusoe Spark production at the company's Brighton factory in 2027.

Crusoe Spark is a compact modular AI factory data center built in the company's factory and shipped ready to deploy.

The company describes it as a key growth engine for Crusoe Cloud, scaling capacity rapidly and predictably and bringing it online at lower cost than conventional construction.

Designed for the latest silicon and networking systems, Crusoe Spark shortens data center construction timelines in the field from years to months.

Density Gains and Closed-Loop Cooling Architecture

Crusoe Spark Liquid provides a 5x increase in compute density over previous generations. The system uses a closed-loop, non-evaporative direct liquid cooling (DLC) architecture that delivers full thermal liquid cooling benefits without consuming or evaporating water. Crusoe says this ensures AI companies and enterprise workloads can scale high-density infrastructure responsibly in water-stressed regions and local communities.

The company identified three key use cases for the platform.

The first is capacity expansion, rapidly adding high-density AI compute for Crusoe Cloud to new or existing sites with the flexibility of a smaller footprint.

The second is edge deployments, bringing compact, purpose-built AI capacity closer to end-users to support latency-sensitive inference workloads. The third is sovereign AI, enhancing existing sovereign infrastructure in targeted international geographies to maintain local data privacy regulations.

Vertically Integrated From Factory to Cloud

Built for Crusoe Cloud, Crusoe Spark Liquid units move compute capacity closer to localized demand, providing the low latency essential to real-time inference and agentic AI applications. Crusoe Spark Liquid is a single, vertically integrated system that the company builds and operates end to end, uniting its modular AI factories and cloud.

According to the announcement, it delivers NVIDIA Exemplar-level performance for training along with the data privacy that inference and sovereign AI deployments require, with efficiencies passed through to customers as lower costs and faster deployment.

"Spark Liquid is Crusoe's sixth generation of modular data centers, taking the best of everything we've learned across hundreds of past modular deployments to define a new standard of excellence for liquid-cooled modular computing infrastructure," said Cully Cavness, Co-founder, President, and Chief Strategy Officer at Crusoe.

"Crusoe has vertically integrated the design, engineering, manufacturing, deployment and operation of Crusoe Spark in order to drive down the cost and timeline of AI compute infrastructure for our customers. The Spark platform is designed to empower AI innovators with cutting-edge GPU capacity on a timeline and scale that matches their ambitions."

Ollama Runs Production Workloads on Spark Infrastructure

Crusoe Spark is already deployed for Crusoe Cloud customers in production, including Ollama, described as the world's most popular way to run open source models privately and locally. Ollama serves more than nine million developers and 85% of Fortune 500 enterprises.

The company uses Crusoe Spark infrastructure in the United States to host select models on Ollama's cloud.

In addition to running open models on Crusoe's managed inference service, Ollama runs dedicated workloads on Crusoe Spark modules powered by renewable solar energy and second-life battery storage, enabling it to bypass traditional data center supply chain bottlenecks.

"Millions of developers around the world count on Ollama for every frontier open model on day zero, with leading performance and zero data retention," said Michael Meltzer, Chief Business Officer at Ollama. "Crusoe is a key partner in how we deliver that. They own the full stack, from power to GPUs, across data centers in the U.S. and Europe, and we look forward to continuing to scale with them on Crusoe Spark."

Software Additions Enter Private Preview

Crusoe Spark Liquid expands the infrastructure behind Crusoe Cloud, and the company continues to build out the platform on top of it.

Alongside the hardware announcement, Crusoe also introduced Crusoe Managed RL (reinforcement learning) and Crusoe Managed MLflow, both now in private preview.

These offerings give customers a single platform to customize frontier open models for their business and prove they perform in production.

Company Background

Crusoe describes itself as the AI factory company, on a mission to accelerate the abundance of energy and intelligence. The company provides a reliable, scalable, cost-effective, energy-first solution for AI infrastructure.

By harnessing large-scale energy sources, building AI-optimized data centers, and delivering an AI cloud platform, Crusoe aims to empower its customers.