The takeaway
The gap between a working agent demo and a production agent is mostly execution infrastructure: isolation, state, and cleanup. Daytona's round signals that investors now treat that layer as its own market, with n8n among the named customers.
Why it matters for builders
If you run agents in production, sandboxed execution with state snapshot and fork is becoming a bought-not-built primitive. Evaluate whether your current container setup can pause, clone and discard an agent environment in under a second; if not, that is where your reliability budget is going.
Daytona Raises $60M to Put a Computer Behind Every Agent
Infrastructure for AI agents just got its first billion-dollar bet. Daytona Platforms announced on Friday that it raised a $60 million Series B at a $1 billion valuation led by FirstMark Growth, bringing total funding to $91 million, according to TechCrunch.
The round pulled in an unusual roster of angels: Stripe's Patrick Collison, Shopify's Tobi Lütke, Perplexity's Aravind Srinivas, Zoom's Eric Yuan, Datadog's Olivier Pomel, Quora's Adam D'Angelo and Mercor's Brendan Foody. Qualcomm Ventures, Samsung Next and Okta Ventures joined as strategic investors.
What Daytona actually sells
Daytona is not an inference provider or an observability dashboard. It sells the execution layer: a sandboxed computer for every agent, where that agent can write code, use Git, drive a browser or open a desktop app. The same platform offers Linux, Windows, macOS and GPU environments.
The pitch, laid out in CEO Ivan Burazin's announcement post, is that agents break the assumptions baked into ordinary cloud infrastructure. A normal instance takes minutes to boot and runs until you kill it. An agent's computer has to start in milliseconds, pause with its state intact, fork into parallel copies to try several approaches at once, and vanish when the job ends.
The growth numbers
Daytona says revenue has grown roughly 25x since January, with average month-over-month net revenue retention of about 135 percent - meaning the customers it had six months ago now spend nearly three times as much in aggregate.
The customer list is a who's who of the agent stack: model builders Meta and Cohere, data and environment teams Handshake, Snorkel, Surge AI, Turing and AfterQuery, and product companies Writer, Clay, n8n, Postman, Mintlify and bolt.new.
Writer is the clearest case study. The company ran its own sandbox platform until it crossed 20,000 sandboxes a day and keeping it alive became a full-time job. On Daytona it provisioned 300,000 sandboxes in a single month. AfterQuery swapped out its in-house GKE cluster in a day.
Why it matters
Every team shipping agents eventually hits the same wall: the model works, but the thing the agent does has to run somewhere safe, isolated and disposable. Most builders either wire up containers and spend months on isolation, cleanup and state, or hand agents credentials they should never touch.
A funded execution layer turns that into an API call. For anyone running multi-agent workflows, the fork-and-rewind primitive is the interesting part - it makes parallel exploration and rollback native to the infrastructure instead of something you bolt on in application code. That is the difference between an agent that retries and an agent that can afford to be wrong.
The real test is whether agent-native sandboxes stay a distinct category or get absorbed into the hyperscalers. With $91 million and a customer base that already includes n8n, Daytona has a head start.
The Automation Brief
Read 5 AI stories instead of 50.
The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.
No noise. Unsubscribe anytime.
Editorial notes
Stefan Trbojevic
n8n Lab Editorial
23 October 2026
23 October 2026
Sources
AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.
