Cast AI has raised $108 million to help businesses reduce cloud waste and run AI workloads more efficiently. The Miami-based startup is building tools that cut through the growing chaos of cloud costs, especially as AI adoption explodes.
With this new funding, Cast AI plans to invest heavily in product development and expand its presence across the U.S. and Europe. The round pushes its valuation close to $900 million, according to insiders.
CEO Yuri Frayman says the company’s mission is simple: boost the efficiency of GPU and compute usage. With more companies struggling to train AI models due to rising infrastructure demands, Cast AI steps in to unlock idle resources and stretch every dollar.
The timing couldn’t be better. Today, most enterprises waste huge amounts of cloud power. Cast AI’s internal research shows that only 10% of CPUs and 23% of memory are used effectively. It’s a gap they aim to close with automation and smart workload allocation.
From Cloud Waste to Cloud Wins
Since its founding in 2019, Cast AI has grown rapidly. It now serves over 2,100 customers, including names like BMW, HuggingFace, Akamai, Swisscom, and FICO. The platform connects with all major cloud providers and works across hybrid environments too.
The latest investment round was co-led by SoftBank Vision Fund 2 and G2 Venture Partners. Other big names joined in, such as Aglaé Ventures, Hedosophia, Creandum, Vintage Investment Partners, and Cota Capital.
Beyond raising funds, Cast AI is building deep partnerships. It works closely with Crusoe Energy and SoftBank to help optimize AI data centers. The company also plays a role in OpenAI and SoftBank’s efforts to scale services in Japan.
While AI is now at the center of its story, Cast AI began with a broader focus. The team first tackled cloud cost issues during their previous startup, Zenedge, which was later acquired by Oracle. That experience sparked the idea to streamline Kubernetes workloads with automation.
The founding trio—Frayman, Leon Kuperman, and Laurent Gil—have long been at the edge of machine learning. Back in 2006, they built Viewdle, one of the first startups to use Nvidia GPUs for training image classifiers. Google eventually acquired that company.
Today, they’re applying that same forward-thinking to a bigger problem: how to make AI and cloud systems faster, smarter, and more affordable.
