GPU compute infrastructure for AI training and inference at scale
Fluidstack operates a physical compute infrastructure business deploying GPUs for AI workloads, with active construction and capacity-upgrade projects across multiple data centers. The hiring mix—engineering-heavy with significant ops and security headcount—and the stack (Kubernetes, SLURM, dark fiber, BMS, Siemens controls) reflect the capital-intensive, real-time operational demands of hyperscale facilities. Current pain points center on civil cost risks, permitting delays, and cooling optimization across distributed sites, indicating Fluidstack is scaling beyond single-location deployments.
Notable leadership hires: Tax Director, Treasury Director, Director of FP&A
Fluidstack provides GPU compute infrastructure for AI training and inference, serving AI labs, governments, and enterprises. The company operates multiple data centers and is actively constructing new facilities to expand capacity. Projects span infrastructure design (medium voltage distribution, cooling systems, commissioning), capacity planning, and operational workflows. The organization is building internal systems for hiring, design review, and supplier management—typical infrastructure-as-a-service operational complexity—while addressing challenges around permitting, civil construction costs, and multi-site resource optimization.
Fluidstack runs Kubernetes and SLURM for workload orchestration, with physical infrastructure managed via BMS (building management systems), Siemens and ABI controls, dark fiber networking, and Cat6a cabling. AWS, GCP, Azure, and OCI are also in use.
Active projects include new data center construction, capacity upgrades for hyperscale facilities, medium voltage distribution design, cooling system scaling, hyperscale commissioning, and internal playbooks for design, engineering, and sourcing workflows.
Fluidstack's technology stack, projects, and hiring signals are inferred from public hiring and company data — career pages, public listings, and company web presence — then clustered and de-duplicated. Figures are estimates that refresh over time. Read our full methodology →
This is not an official vendor or customer list. It is a technology-adoption signal inferred from public data, intended for B2B research.