Daytona Raises $60 Million Series B at $1 Billion Valuation for AI Agent Cloud

For roughly two decades the cloud computing industry has been organized around a single assumption: the customer on the other end of the API is a human developer who deploys a long running application and expects it to stay up. That assumption shaped everything from pricing to the way virtual machines are provisioned, and it helps explain why the biggest platforms still measure the time to a usable machine in tens of seconds or more. A new crop of startups now argues the assumption is obsolete, because the fastest growing customer in computing is no longer a person but a software agent that spins up work, fails, retries and shuts down again within seconds.

Agents need somewhere to live. When a model calls a tool, executes Python, opens a browser or clicks through a customer relationship management system, it does so inside an isolated environment that has to be created on demand, kept separate from other tenants and destroyed afterwards. The industry calls these environments sandboxes. The engineering challenge is not raw compute but latency and density: an agent that waits thirty seconds for a machine to boot is an agent that burns money, tokens and user patience in equal measure.

Daytona Platforms Inc. is one of the companies selling that layer. The San Francisco based business was founded in 2023 by Ivan Burazin, Vedran Jukic and Goran Draganic, and it began life automating development environments for human programmers before pivoting to agent scale workloads. It now describes itself as an agentic cloud, a phrase meant to signal that its product is designed for machines rather than people, and that its unit economics are consumption based rather than seat based.

On October 20, 2026, that positioning turned into a valuation milestone. Fortune reported on October 20, 2026 that Daytona raised a $60 million Series B at a $1 billion valuation, a round led by FirstMark that pushes total funding to roughly $91 million. The deal makes Daytona a unicorn less than four years after it was founded, and only eight months after the $24 million Series A that FirstMark also led.

Key Facts

The round is unusually well supported by existing backers. Fortune reported on October 20, 2026 that Chief Executive Ivan Burazin put about $50 million of the $60 million down to investors already on the cap table, with newcomers filling the remainder. Crypto Briefing reported on October 20, 2026 that the funding history runs from a pre-seed of about $2 million in 2023, through a $5 million seed in 2024, to the $24 million Series A in February 2026, meaning the jump from a few million dollars of seed money to a ten figure valuation took only a few years.

The investor list mixes venture firms with well known founder angels. Tech.eu reported on October 20, 2026 that FirstMark led the round with participation from Patrick Collison of Stripe, Adam D'Angelo of Quora, Brendan Foody of Mercor, Tobi Lutke of Shopify, Olivier Pomel of Datadog, Aravind Srinivas of Perplexity and Eric Yuan of Zoom, alongside Alumni Ventures, Evergreen Fund, Okta Ventures, Qualcomm Ventures, Samsung Next and Torch Capital. Existing backers increased their commitments rather than simply holding their positions: Pace Capital doubled its investment, Upfront Ventures tripled its and E2VC quadrupled its, according to the same report.

The product itself is a programmable sandbox reached through a single API. Daytona says it can configure a new sandbox in about 60 milliseconds, compared with 30 seconds to several minutes for Amazon Web Services, Microsoft Azure or Google Cloud, because it pre-distributes snapshots of configured machines onto server drives in advance. The environments support Linux, Windows and macOS, scale compute up or down per task, and let agents pause, resume or explore alternative approaches without losing progress. Daytona added a Windows sandbox that Burazin says it was first to offer and remains alone in providing, plus a GPU sandbox used for 3D rendering and design work, with customers running AutoCAD and Blender.

Traction is spread across the AI industry. Customers named in the coverage include Anthropic, OpenAI, Nvidia, Cursor, Cognition, Meta and Cohere, along with Mercor, Handshake, Surge AI, SambaNova, Postman, Bolt.new, n8n and Clay. Fortune reported on October 20, 2026 that the AI data company Snorkel has provisioned more than one million sandboxes on the platform, and that revenue run rate grew 25-fold between January and September 2026, an average of 50.9% a month. Crypto Briefing reported on October 20, 2026 that Daytona reached $1 million in forward annual recurring revenue in under 50 days after launch and $3 million in ARR within 95 days of shifting to agent native infrastructure, without a dedicated sales team.

Headcount remains tiny relative to the valuation. The Next Web reported on October 20, 2026, in a podcast interview with Burazin, that Daytona has about 30 employees, roughly two thirds of them in Europe, and about 1,100 shared Slack channels with customers. Fortune reported that most of the new money will go to go to market sales staff plus some engineers, and Tech.eu reported that the company plans to expand its agent cloud beyond the initial sandbox product, scale infrastructure and hire in San Francisco, Croatia and Singapore.

Analysis

What this really means is that Daytona has been priced as a category leader in a market that barely existed eighteen months ago, and the number reflects growth rate far more than it reflects durable margin. A 25-fold increase in run rate over nine months is the kind of curve that makes investors overlook a 30 person team, a self serve sales motion and a single core product. It is also the kind of curve that can flatten quickly if the largest customers decide to build their own provisioning layer, which is a live risk when those customers are Anthropic, OpenAI, Meta and Nvidia.

The revenue mix adds a second layer of concentration. The Next Web reported on October 20, 2026 that Burazin estimates about 70% of revenue comes from post-training work, the "school" market where AI labs and data companies run agents inside Daytona sandboxes to train and evaluate models, with the remaining 30% coming from "work" deployments where agents act on behalf of employees. Training budgets are the most cyclical line item in the AI economy, so a valuation built on lab spending is a valuation exposed to every pause in a frontier model program. The company's argument is that the two markets reinforce each other, since the same tooling that evaluates a model also runs it in production.

Burazin's most interesting claim is about silicon, not software. He argues that agent sandboxes are mostly Linux machines with a CPU and no GPU, and he predicts a CPU crunch that will be bigger than the GPU crunch. Fortune reported on October 20, 2026 that AMD raised its server CPU market growth forecast from about 18% to more than 35% a year, reaching over $120 billion by 2030, and that Arm estimates agentic AI data centers could need 120 million CPU cores per gigawatt. Burazin also estimated that Meta's Muse agent would need $2.8 billion worth of sandboxes to serve 100 million users, a figure that hints at the scale of consumption he is selling to investors.

The go to market story is the part that the new money is meant to fix. Burazin told The Next Web that Daytona grew just shy of 51% month on month for 18 months with no salespeople, that events including a 1,200 person conference on the Chase Center basketball court in San Francisco did most of the selling, and that a tweet seen 2 million times led to a partnership with Nvidia to test its CPU. He also said the company had just agreed to hire its first head of sales. The bigger picture here is a company trying to convert a product led growth engine into an enterprise sales machine before the growth rate decays, and that transition is where most infrastructure startups either compound or stall.

Why It Matters

The sandbox layer is quietly becoming a dependency for the entire agent economy. If the models from Anthropic, OpenAI, Meta, Cohere, Cursor and Cognition all execute tool calls inside rented isolated machines, then whoever supplies those machines sits on a choke point that is invisible to end users but decisive for latency, cost and reliability. Daytona's 60 millisecond provisioning claim against the 30 seconds to several minutes of the hyperscalers is the whole pitch in a single number, because agents that run thousands of short jobs per hour cannot absorb boot times designed for servers that run for months.

Security is the second reason the category matters. Burazin's framing is that companies will soon manage thousands of digital employees, and each one needs a safe place to fail so that a single rule breaking agent does not spill into company wide chaos. That argument turns sandboxes from a developer convenience into a governance control, the place where permissions are enforced and blast radius is limited. It also explains why Okta Ventures, Samsung Next and Qualcomm Ventures appear on the cap table alongside the founder angels.

Geography and talent round out the picture. Two thirds of the team is currently in Europe, and Burazin expects most of it to be in the United States next year, with hiring planned across San Francisco, Croatia and Singapore. For a Croatian founded company selling American AI labs, that migration is a bet that proximity to customers outweighs the cost advantage of a European engineering base.

Next Up

The immediate task is expansion beyond the sandbox itself. Tech.eu reported on October 20, 2026 that Daytona intends to build out a broader agent cloud, scale infrastructure and add staff in three regions, while Fortune reported that the bulk of the fresh capital goes to go to market hiring with some engineering attached. The first head of sales, promised in the TNW interview, will be the clearest signal of whether the product led model survives contact with enterprise procurement.

Further out, the company's fortunes are tied to two variables it does not control: whether agent workloads keep multiplying inside the labs and data companies that supply about 70% of revenue, and whether the CPU capacity Burazin predicts will be scarce actually arrives at a price that keeps consumption based sandboxes cheap. If both hold, the $1 billion mark is a starting point rather than a ceiling. If either slips, Daytona will be the first agent infrastructure company to test how forgiving the market is at that valuation.