What Fluidstack does
Fluidstack is an AI infrastructure “neocloud” company building and operating GPU cluster capacity for large AI labs, enterprises, and governments. The company positions itself around speed to deploy frontier-scale compute—acquiring power, designing and building data centers, and operating the resulting infrastructure with dedicated hardware and software teams.
More
Fluidstack also provides a software layer for running large-scale training and inference workloads, including managed Kubernetes and managed Slurm (batch) orchestration with health checks, observability, and topology-aware scheduling aimed at reducing collective-communication latency for multi-GPU jobs. Target customers include frontier model builders such as Anthropic, and the company has also announced partnerships with Mistral AI and other infrastructure organizations to support enterprise deployments. Business model (as described publicly): Fluidstack sells access to GPU compute and managed cluster environments (training and inference) delivered through its managed platforms, while simultaneously funding and operating physical/energy infrastructure to reduce bottlenecks in capacity acquisition and deployment timelines. Strategically, Fluidstack is signaling a scale-up phase driven by large, named compute-buildout partnerships—most notably Anthropic’s announced $50B American AI infrastructure investment, under which Fluidstack is described as a data center build-out partner for sites in New York and Texas.