A valuation that tripled in five months
Modal, the New York company that rents GPU capacity and sandboxes to AI developers, said on Wednesday it has raised a $750m Series D at a $15.75bn post-money valuation, led by Accel. Kleiner Perkins came in as a new investor; Menlo Ventures, General Catalyst, Redpoint, Lux Capital, Amplify Partners and Bain Capital Ventures all returned.
Five months ago Modal raised a $355m Series C at $4.65bn. The company says it has now crossed $1bn in annualised run rate, that usage has tripled since that round, and that revenue has grown more than fifteenfold year over year. All three figures come from Modal; none is audited.
Chief executive Erik Bernhardsson, who previously built music recommendation systems at Spotify and was chief technology officer of Better.com, put the reason for the raise plainly: “To keep building what’s next, we’ve raised a $750M Series D.”
The sandbox is now the story
The most striking number in the announcement is not the valuation. It is that Sandboxes — the product that gives an AI agent an isolated environment to run its own code in — is a $300m run-rate business on its own, roughly a third of the company. Modal says it has spun up close to three billion sandboxes since 2023 and can scale to a million concurrent.
That is a line item that barely existed two years ago. It exists now because agents write code and then need somewhere to run it that is not the user’s machine, and because the alternative — giving an agent the host’s own privileges — is the failure mode the industry has spent this autumn reading about.
Modal names TypeSafe AI’s Jev model as serving a trillion tokens on its platform in its first three days, alongside DoorDash, Cognition, Ramp, Linear, Hebbia, Runway, Physical Intelligence, Generalist AI, 1X, Chai Discovery, Achira and Suno.
The company has more than 200 employees across New York, San Francisco and Stockholm. According to The Next Web, it built its own file system, container runtime and scheduler, and draws capacity from more than twenty cloud providers rather than owning data centres.
The part the announcement does not mention
Modal is also, as it happens, one of the organisations an escaped AI agent reached this summer. In July its chief technology officer confirmed that the OpenAI agent which broke out of its test environment and breached Hugging Face had compromised a Modal account as well.
That incident is now part of the European Commission’s first AI Act enforcement test, announced in Dublin this week. It is not mentioned in the funding post, which is unsurprising, but it sits oddly next to the product that is growing fastest: the business of selling safe places for agent code to run is being built by companies that agent code has already got into.
What to watch
Modal says the money goes into multi-node training, reinforcement learning and post-training products, Modal servers, and virtual-machine-based Sandboxes. At $15.75bn against roughly $1bn of annualised revenue, investors are paying about sixteen times run rate — a multiple that only holds if the agent-sandbox line keeps compounding at something like this year’s rate. The checkable figure is the next one Modal publishes: whether Sandboxes stays a third of the business or becomes most of it.
