Modal crossed $1 billion in run rate. Nearly a third comes from sandboxes for agents

Modal says it has crossed $1 billion in annualized run rate only five months after announcing a funding round at a $4.65 billion valuation. The company has now raised $750 million in a Series D at a $15.75 billion post-money valuation. Its announcement says usage tripled since the earlier round and grew more than fifteenfold year over year.[1][3]

The most revealing number is inside the total. Modal says Sandboxes, its isolated virtual-computer product, has reached a $300 million run rate after close to 3 billion sandbox launches. That means roughly 30% of the stated company run rate comes from a building block that lets software—and increasingly AI agents—execute code away from the user’s main system.[1][2]

Why agents need disposable computers

A chatbot can answer in text without touching a machine. A coding agent may need to install packages, run tests, open a browser, compile software or execute an untrusted script. Giving it the developer’s laptop or a shared production server would mix experimental actions with valuable data and credentials. A sandbox creates a temporary boundary: the agent receives a controlled environment, performs the job and can be discarded afterwards.

This is broader than “AI coding.” Modal lists background agents, reinforcement learning, prompt-driven development platforms and other workloads. The common requirement is bursty, programmatic compute that can start quickly and be isolated by default. The platform becomes valuable when thousands of short-lived jobs are easier to launch through code than through a conventional cloud console.[1]

A billion-dollar pace contains several products

Sandboxes are not Modal’s only business. The company also supplies GPU inference, training, multi-node clusters, web endpoints and other serverless infrastructure. It says TypeSafe AI’s Jev served one trillion tokens on Modal in its first three days. That example illustrates the platform’s appeal: a product team can consume large amounts of compute without first building an infrastructure organisation. It does not reveal the economics of that workload or whether the launch rate persisted.[1]

The financing history shows how quickly expectations have changed. Reuters reported in May that Modal was raising at a $4.65 billion valuation and had reached an annualized revenue pace of roughly $300 million, up from about $60 million the previous September. The new post-money valuation is about 3.4 times the May valuation. The reported company run rate has also more than tripled, but valuation, capital raised and revenue pace are different quantities and should not be blended into one growth percentage.[3][1]

The unanswered question is the margin behind the convenience

Infrastructure growth can be spectacular while consuming spectacular amounts of capital. Modal rents or procures access to expensive accelerators, competes with hyperscalers and specialist clouds, and must keep enough capacity available for workloads that arrive unpredictably. A million concurrent sandboxes is useful only if utilisation, pricing and orchestration turn that technical capability into durable gross profit.

The $300 million sandbox figure nevertheless identifies a real market. AI agents are not only a software-interface story; they create demand for places to act. Every time an agent opens a clean environment, runs a tool and disappears, the transaction resembles a tiny cloud workload. Modal’s bet is that those tiny workloads will become one of the largest new categories in infrastructure—and that developers will pay to avoid managing the machines underneath them.

Report an error or suggest a correction