Key Takeaways

  • Factory secured $150 million at a $1.5 billion valuation to expand its AI coding agents for enterprise teams.
  • Khosla Ventures led the round, with managing director Keith Rabois joining Factory’s board.
  • The startup promotes model switching across Claude, DeepSeek, and others as its main differentiator.

AI-assisted coding keeps pulling in both customers and capital, even as the broader generative AI landscape becomes crowded. Factory, a startup developing AI agents for enterprise engineering teams, has entered the upper tier of contenders after announcing a $150 million fundraise on Wednesday at a $1.5 billion valuation. Investors are clearly convinced that there is still plenty of room in the market, despite established competition from Anthropic, Cursor, and Cognition.

For all the noise around AI in design, marketing, or general productivity, coding remains the place where engineering leaders see immediate economic payoff. Faster iteration cycles, fewer bottlenecks, and cleaner handoffs between junior and senior developers offer tangible value. It is a type of work where AI does not have to simulate creativity in the abstract; it can simply help teams produce and maintain more software.

The new funding round was led by Khosla Ventures, with Sequoia Capital, Insight Partners, and Blackstone also participating. Keith Rabois, a managing director at Khosla Ventures, has joined Factory’s board. His involvement adds significant weight, as he has been publicly bullish on AI automation inside enterprises for more than a decade.

Factory founder Matan Grinberg told the Wall Street Journal that the company stands apart due to its system's ability to shift across different foundation models, such as Anthropic’s Claude or Chinese AI startup DeepSeek. Many engineering leaders have been asking for this type of flexibility so they can avoid vendor lock-in while still getting access to the best available model for a particular task. Cursor has already positioned itself similarly on this point, which raises a natural question regarding how much value model switching itself provides versus the overall workflow design.

Factory has already landed significant enterprise deployments. Engineering teams at Morgan Stanley, Ernst & Young, and Palo Alto Networks are using the platform. These are not small pilot customers; they tend to run software stacks filled with legacy systems, internal tools, and strict compliance constraints. If Factory can succeed in those environments, it is usually a strong indicator that the approach can scale to other large organizations.

The startup’s origin story is an outlier compared to the typical Silicon Valley founding path. Grinberg was still a PhD student at UC Berkeley in 2023 when he cold-emailed Sequoia partner Shaun Maguire. The two connected over shared academic interests. Maguire had earned his PhD at Caltech in the same physics field that Grinberg was studying, and that shared background ended up shaping the company’s trajectory. Within a short period, Maguire persuaded Grinberg to leave his program and launch Factory, with Sequoia providing backing at the seed stage.

Some investors argue that the market for coding agents is expanding so quickly that companies like Factory, Cursor, and Cognition will likely carve out distinct niches. Anthropic continues to build ambitious enterprise tools around Claude Code, but AI-assisted coding remains fragmented enough that new entrants with specialized tooling can gain real traction. Engineering managers increasingly want tools capable of handling everything from lightweight code suggestions to the orchestration of multi-step changes, and Factory is attempting to lean heavily into that complex orchestration category.

Enterprises rarely adopt generative AI tools simply because they generate code. What they usually want is predictable repeatability. They require systems that can scaffold code changes, run tests, write documentation, and interact with existing CI pipelines. When viewed from that perspective, the focus on model switching is only part of Factory’s strategy. It hints at a deeper goal: routing each type of development task to the most competent model rather than asking one model to handle everything. It is a tougher engineering challenge, but potentially a more durable one.

There is still an open debate about whether model diversity is a long-term differentiator. Frontier models keep improving rapidly, and industry executives wonder if a single high-performing model will eventually dominate most use cases. If that happens, the companies built around model orchestration will need other advantages, such as superior workflow depth or tighter integration with enterprise security frameworks.

For now, investors are betting that Factory can build momentum before any such consolidation arrives. Large enterprises are racing to integrate AI into their engineering organizations, and many teams are willing to run multiple tools in parallel. AI agents are becoming another standard layer in the development stack, much like the adoption of version control platforms or cloud-based build systems a decade ago. The next year will show whether Factory’s approach resonates beyond early adopters, and whether its model-switching capability truly gives it a lasting edge.