Daytona’s $1 Billion Valuation and the Base Behind 25× Growth

On October 20, 2026, San Francisco-based Daytona raised a $60 million Series B led by FirstMark at a $1 billion valuation, taking total funding to about $91 million; about $50 million came from existing investors, and CEO Ivan Burazin said revenue run rate grew 25-fold from January to September, according to Fortune’s account of the round. The growth multiple has no disclosed starting revenue, current dollar run rate or margin attached to it, so the private valuation cannot be translated into a sales multiple.

The company sells isolated computing environments, or sandboxes, in which AI agents can run code and use software tools. In Tech.eu’s report, Burazin says each agent needs “a safe place to fail”; the service launches environments through an API and supports Linux, Windows and macOS. That is the product behind the financing: a managed place for an agent to do work, separate from the model that decides what work to do.

Who backed the round and where the money goes

The investor group includes Qualcomm Ventures, Samsung Next and Okta Ventures, alongside Patrick Collison of Stripe, Tobi Lütke of Shopify, Aravind Srinivas of Perplexity, Eric Yuan of Zoom, Olivier Pomel of Datadog, Adam D’Angelo of Quora and Brendan Foody of Mercor. Existing backers supplied most of the new capital. Their participation gives Daytona funding to expand a service it already sells while building a larger commercial operation.

Most of the proceeds are intended for sales and other go-to-market hiring, with some engineering expansion. That allocation puts a commercial task alongside the technical one: reaching more customers and supporting larger deployments as agent workloads move from experiments into production. Growing usage may attract investors, but the planned hiring shows that Daytona must also turn demand for sandboxes into sustained customer spending.

The financing price is a private transaction valuation. Daytona has not disclosed the share of the company sold in the round, the terms attached to the new capital or the calculation used to arrive at the valuation. The amount raised therefore says little by itself about the cost of delivering each sandbox, and the deal price cannot yet be compared with a disclosed revenue figure.

What Daytona’s agent cloud actually supplies

An AI agent that writes code, browses the web or operates business software needs a computer environment with the relevant tools and permissions. Daytona supplies that environment as an isolated sandbox. The agent can execute its instructions there while the customer controls what the environment can access. Isolation matters when software generated or operated by an agent may fail, change files or interact with outside services.

The company’s design addresses a pattern different from a long-lived application server. An agent may start a short job, pause while another step completes, resume with its state intact and branch into copies to pursue alternative approaches. A sandbox that can be provisioned and removed around that pattern reduces the need for a customer to assemble a separate virtual machine setup for every task. Cloning a working environment also lets an agent explore alternatives without discarding its original state.

Daytona describes a larger agent cloud built around computers, browsers and storage. Its sandbox offering spans multiple operating systems and GPU environments, allowing different workloads to use the same management layer. The distinction is concrete: Daytona sells execution capacity and control over the environments in which an agent acts. The model and application still determine what task the agent attempts and whether its output is useful.

The growth ratio and the missing dollar figures

In Daytona’s October 20 announcement, the company says revenue has increased approximately 25-fold since January, average monthly net revenue retention is about 135%, and customers present six months earlier now spend nearly three times as much in aggregate. The company’s wording refers to revenue growth since January; Burazin’s description to Fortune refers specifically to revenue run rate through September. Those measures and end points should not be treated as identical without a definition of the underlying calculation.

A multiple describes relative change, not the size of the business. If starting revenue were small, a dramatic increase could still leave a modest current business; if the base were substantial, the same ratio would imply a much larger absolute gain. Daytona has not given that starting amount or a current revenue run rate in dollars. The omission prevents a comparison between its private valuation and annualized sales on a common, dated basis.

The retention figure answers a different question. It indicates that existing customers are spending more in aggregate, but it does not reveal revenue from newly signed customers or how spending is distributed across accounts. Daytona names model builders, data and evaluation businesses, and developers of agent-powered products among its customers. Training and evaluation can create bursts of sandbox demand, while production agents may require continuing capacity. Those patterns could produce different revenue durability even if both contribute to growth.

Margins are the other missing part of the economics. Daytona incurs costs to supply compute and operate the environments it sells; rising usage improves the business only to the extent that revenue exceeds those costs and the expense of winning customers. A dated dollar run rate, its comparable starting base and gross margin on sandbox activity would show more clearly whether demand for the service supports the scale of the new private valuation.