EP37 - Unicorn Files - CAST AI: from Kubernetes automation to GPU marketplace

How CAST AI went from cutting cloud bills by 50% or more to launching OMNI Compute, a multi-cloud GPU marketplace: 2,100 customers, 108 million dollar Series C and valuation over 1 billion dollars. The story of the most interesting cloud infrastructure to AI infrastructure pivot in the European ecosystem.

Founded in 2019 in Miami by a team of ex-Google, Lyft and Booking.com, CAST AI was born as an automatic Kubernetes cost optimization platform: it analyzes clusters in real-time, automatically selects the cheapest spot instances with on-demand fallback, performs resource rightsizing and aggressive autoscaling. The average result for customers is a 50-70% reduction in cloud costs with no manual intervention — ROI so obvious it makes sales almost automatic.

With the explosion in GPU demand for AI, CAST AI saw an adjacent opportunity: apply the same optimization and orchestration logic to GPUs, a dramatically more expensive and scarce resource. OMNI Compute is the marketplace that aggregates GPU capacity from AWS, GCP, Azure, Lambda Labs and specialized datacenters in a single interface, with intelligent scheduling and cost optimization for training and inference. In this episode we analyze the double transition — product and market — and the EU compliance risks related to distributed management of AI models.

• Rightsizing/autoscaling e Spot con fallback: cosa significa davvero ‘automazione’ in Kubernetes
• GPU come risorsa fungibile: impatto GTM, metriche e rischi di compliance UE