News · Tech
AI infrastructure company Modal has raised a $750M Series D at a $15.75B post-money valuation after crossing $1B in annualised run rate, with its Sandboxes product alone at $300M. Here is what the round says about where AI compute spending is going.
Modal, the serverless compute platform for AI workloads, has raised a $750 million Series D at a $15.75 billion post-money valuation after crossing $1 billion in annualised run rate[^1]. Accel led the round[^2].
01 / 04
CEO Erik Bernhardsson wrote that usage has tripled in the five months since Modal's Series C, and that the company has grown more than 15 times year over year[^1]. Accel says Modal crossed the $1 billion revenue run rate during September 2026[^2]. Kleiner Perkins joined the round, and every existing investor participated, including Menlo Ventures, General Catalyst, Redpoint, Lux Capital, Amplify Partners and Bain Capital Ventures[^1]. Axios Pro, which had the news first, frames it as part of a jump in valuations for AI-specific serverless clouds such as Modal and Baseten as companies race to secure compute[^3].
02 / 04
The most telling figure is Sandboxes. Modal launched them in 2023 after noticing customers were using LLMs to generate code that needed somewhere isolated to run[^1]. Three years later Modal says Sandboxes are a $300 million run-rate product on their own, with close to 3 billion spun up and the ability to scale to a million running at once[^1]. Background agents, reinforcement learning and vibe-coding platforms are the main uses[^1].
Modal says it rebuilt most of the stack to get there: its own file system, container runtime, scheduler and networking, running on a capacity pool that spans more than 20 clouds[^1].
03 / 04
Modal lists DoorDash's AI platform team, background-agent builders Cognition, Ramp, Linear and Hebbia, video generation at Runway, robot training at Physical Intelligence, Generalist AI and 1X, and drug discovery at Chai Discovery and Achira[^1]. It also says TypeSafe AI's Jev served 1 trillion tokens on Modal in its first three days[^1].
New products include multi-node training with no capacity commitment, upcoming reinforcement learning and post-training tools, Modal servers for HTTP workloads, and virtual machines inside Sandboxes, now generally available[^1].
04 / 04
A $300 million business built only on running untrusted, model-written code is the clearest sign yet that agents are moving from demos into production workloads that need real isolation. If your product has an agent that writes or runs code, the sandbox is now a line item worth designing for early: how long sessions live, what they can reach on the network, and what each run costs. Rent it rather than build it until the volume proves otherwise, and keep the interface thin enough that you can switch providers as this market keeps moving.
FAQ
Modal raised $750 million in a Series D at a $15.75 billion post-money valuation, announced on 14 October 2026 (US time). Accel led the round, Kleiner Perkins joined, and all existing investors took part, including Menlo Ventures, General Catalyst, Redpoint, Lux Capital, Amplify Partners and Bain Capital Ventures.
Modal says it has crossed $1 billion in annualised run rate, and Accel says that happened during September 2026. Modal also says usage has tripled in the five months since its Series C and grown more than 15 times year over year.
Sandboxes are isolated environments where code, often written by an AI model, can run safely away from everything else. Modal launched them in 2023, says they are now a $300 million run-rate product on their own, and reports close to 3 billion sandboxes spun up, scaling to a million running at once. Uses include background agents, reinforcement learning and vibe-coding platforms.
Modal names DoorDash's AI platform team, Cognition, Ramp, Linear and Hebbia building background agents, Runway generating video, Physical Intelligence, Generalist AI and 1X training robots, and Chai Discovery and Achira running drug discovery. It also says TypeSafe AI's Jev served 1 trillion tokens on Modal in its first three days.
Modal points to multi-node training that gives you an RDMA-connected cluster from one decorator with no capacity commitment, new reinforcement learning and post-training products, Modal servers for HTTP workloads, Endpoint Candidates for LLM serving, and virtual machines in Sandboxes, which are now generally available.
That’s the story
Spectrum Web Co ships SI and Web3 products, places talent and builds social narratives as monthly services.
