Modal raises $355M Series C to build agent-focused cloud infrastructure

Modal, a cloud platform, raised a $355M Series C to build infrastructure specifically for AI agents rather than human developers. The company’s CTO, Akshat Bubna, argues that traditional cloud stacks were designed for humans who can read documentation, reason through configuration, and debug failures, but agents lack those capabilities.

Bubna states that agent workloads require tighter integration, faster iteration, and programmatic primitives such as sandboxes, elastic inference, and GPU snapshotting. These features enable agents to spin up isolated environments, run custom models, and debug failures without human intervention.

The company’s infrastructure now includes serverless functions, decorator-based configuration, elastic inference for custom models, GPU snapshotting, speculative decoding, Auto Endpoints, persistent storage, networked containers, private IPv6, RDMA, multi-node training, and a capacity pool spanning 17 cloud providers.

Bubna notes that production agents need specialized sandboxes and hard guardrails, and that observability is critical when agents are writing their own code. He also highlights the need for RL rollouts that can require up to 100,000 sandboxes, reflecting the scale of agent-driven experimentation.

Modal’s shift from developer experience to agent experience reflects a broader industry trend: AI infrastructure must evolve to support bursty, compute-intensive, and programmatic workloads that agents impose, rather than the steady-state workloads of traditional web applications.