Modal Labs has closed a $750M Series D, valuing the AI infrastructure provider at $15.75B, more than three times what it was worth after a round in May. Accel led the financing, joined by Kleiner Perkins and a set of returning investors.
The top line kept pace. Modal says its annualized revenue has more than tripled since May, now above $1B.
Sandboxes account for nearly a third of that total. The service spins up hosted containers configured for AI workloads, and developers use it for training runs, for agents, and for the code those agents write.
According to the company, Sandboxes can launch tens of thousands of containers each second, coordinated by a custom scheduler that spreads provisioning across servers. In one internal trial, Modal ran a million containers at once.
Several design decisions target agents specifically. Sidecar containers that catch hallucinations or run helper tools reside on the same physical machine as the workloads they serve, which trims latency. Secret Injection, a feature Modal offers, lets agents use application credentials without ever seeing them, guarding against rogue agents and stolen credentials.
Rather than owning data centers, Modal rents capacity from over twenty cloud providers, allowing workloads to move when a vendor falters. Its inference offering, Endpoints, places customer models on hosted infrastructure, and it also supplies a library of open-source algorithms for enterprises that prefer not to train their own.
