Modal

Modal is a cloud infrastructure company for teams that build and run AI applications. Developers write code, and Modal runs it on managed CPU and GPU capacity that scales up and back down. On 14 October 2026 it announced a $750 million Series D led by Accel at a $15.75 billion post-money valuation, with Kleiner Perkins joining. The company puts its annualised run rate at $1 billion, a figure it supplied itself.

What it does

Modal offers several products on one platform. Inference serves language, image, video and audio models, with OpenAI-compatible endpoints and room for custom serving stacks. Training supports fine-tuning methods such as supervised and reinforcement learning on single or multi-node GPU runs. Sandboxes are isolated environments for coding agents, reinforcement learning rollouts and untrusted code, with snapshot and restore for fast starts. Functions run arbitrary code as serverless jobs, batch work or HTTPS endpoints. Underneath sit a custom container runtime, a distributed filesystem for weights and datasets, and capacity drawn from more than 20 clouds. Pricing is usage-based, with GPU time billed by the second and a monthly free compute allowance advertised on the site.

Who it serves

The audience is AI engineering teams, from startups to large firms. Testimonials on the site come from DoorDash, Cognition, TypeSafe AI and Ramp. For enterprises it lists role-based access, audit logs, data residency controls and SOC 2 Type II compliance.

Key facts

Headquarters: New York, with staff also in San Francisco and Stockholm People: Erik Bernhardsson, co-founder and CEO; more than 200 staff Funding: $750M Series D, October 2026, led by Accel at $15.75B post-money; Kleiner Perkins participated Run rate: $1B annualised, company figure Products: Inference, Training, Sandboxes, Functions, Notebooks, Batch