Modal

Serverless GPU infrastructure for AI

Modal is the infrastructure foundation for AI teams, providing instant GPU access, sub-second container startups and native storage so that training models, running batch jobs and serving low-latency inference all happen in one system rather than three. Sub-second cold starts are the load-bearing detail: they are what make serverless economics work for GPU workloads that would otherwise sit idle and billed. Thousands of customers run production workloads on it, including Lovable, Scale AI, Substack and Suno. The team spans New York, San Francisco and Stockholm, and the company has passed nine-figure ARR.