Factory Raises $200 Million as AI Coding Agents Become Enterprise Infrastructure

AI coding startup Factory raised $200 million at a $5 billion valuation as businesses increase spending on software agents that can work across engineering workflows.

AI-assisted coding has moved from autocomplete into a race to build agents that can take ownership of larger engineering tasks.

Reuters reported on September 15 that Factory raised $200 million in a funding round that more than tripled the AI coding company’s valuation to $5 billion.

The investment reflects a major shift in developer tooling: companies are no longer paying only for code suggestions — they are experimenting with AI systems that can plan, modify, test and review software.

Coding is a natural agent market

Software development offers unusually clear feedback. Code can be compiled, tested and reviewed automatically, giving AI systems signals about whether a task succeeded.

That makes coding one of the strongest environments for agentic AI. An agent can inspect a repository, identify relevant files, make changes and run tests before asking a human for review.

Enterprise requirements are different

A coding demo can look impressive while still being unsuitable for a large company.

Enterprises need permission controls, repository isolation, audit trails, predictable costs and integration with existing development systems. They also need confidence that generated code does not introduce security or licensing problems.

The winning enterprise coding agent may not be the one that writes the most code. It may be the one that can be trusted with the largest amount of real engineering work.

A crowded market

Factory competes in a category attracting intense investment. Developers now have AI features inside editors, command-line tools, code review systems and autonomous agents.

That competition benefits engineering teams because products must differentiate on workflow quality rather than raw model access alone.

How teams should evaluate coding agents

Companies should measure completed tasks rather than generated tokens. Useful metrics include acceptance rate, tests passed, review time saved, regressions introduced and cost per successful issue.

The most valuable internal benchmark is often a collection of real historical engineering tasks with known outcomes.

What comes next

Coding agents are likely to become more integrated with issue trackers, CI systems, documentation and deployment workflows.

Human engineers will still need to define architecture, review changes and make product decisions, but the unit of automation is getting larger.

AI coding is evolving from “help me type” toward “help my engineering organization finish work.”