Emergent Runs Over 30,000 AI Coding Environments Using Kubernetes Pods

AI-powered software development platform Emergent said it runs its agent infrastructure on Kubernetes-based ephemeral environments to support more than 30,000 concurrent development workspaces, enabling AI agents to build and deploy applications in full Linux environments.

Kubernetes ephemeral environments are temporary, isolated setups that allow developers to test new features, fixes, or staging workloads without impacting production.

The company disclosed details of the system in a technical blog post outlining how its infrastructure evolved as usage grew rapidly following its launch. 

According to Emergent, the platform assigns an AI agent to a dedicated development environment whenever a user describes a product they want to build.

“You describe the product you want to build—a SaaS dashboard, a REST API, a data pipeline—and our agent takes it from there: planning the work, writing the code, installing dependencies, running tests, and deploying a live, working application,” the company wrote in the blog post.

Emergent stated that each user prompt triggers the creation of an isolated development environment where the agent writes code, installs packages, runs builds, and starts application services. The goal is to reduce the time between an idea and a deployed product from days to minutes.

According to Emergent, the platform’s infrastructure had to evolve quickly as demand grew from a few hundred concurrent environments to over 30,000 active environments. The challenge was not only provisioning environments at scale but also ensuring they start quickly, persist user work reliably, and release resources cleanly when sessions end.

Emergent evaluated multiple approaches for running these environments, including virtual machines and lightweight sandbox runtimes. 

The company stated that virtual machines provide complete operating systems but take 30–120 seconds to provision, while lightweight sandboxes start quickly but often lack support for system packages, databases, or background services.

“AI coding agents need real infrastructure, not lightweight sandboxes dressed up as infrastructure,” the company wrote.

Emergent ultimately chose Kubernetes pods as the foundation for its agent environments. The company said this approach allows each environment to run a full Linux system where the agent can install packages, run tests, start services, and serve application previews through a reverse-proxied URL.

The architecture uses a multi-stage pod lifecycle that restores user state from backups, runs the development environment, and saves incremental changes before termination. The company said this design allows sessions to resume even if infrastructure failures occur.

“For Emergent, because we want to control a tight feedback loop between agents and the underlying dev environment, we wanted to own our infra end to end,” the company wrote.

Also, Kubernetes scheduling, bin-packing, and autoscaling allow the platform to run large numbers of environments efficiently while maintaining isolation between sessions.

The company said the next phase of its infrastructure work will focus on scaling mechanisms such as pre-warmed environment pools, multi-cluster orchestration, and new approaches to state persistence.

Recently, Emergent achieved an annual run rate (ARR) of $100 million, just eight months after its launch, making it one of the fastest AI-native platforms to reach this milestone.

In January 2026, it secured $70 million in a Series B funding round led by Khosla Ventures and the SoftBank Vision Fund 2, with contributions from Prosus, Lightspeed, Together, and Y Combinator, bringing its total capital raised to $100 million to date.