Factory’s $5B AI Coding Round Has a Rotten Problem: Its Investor Backed Both Sides
A startup worth $5 billion just got a public vote of confidence after its own investor called it second-rate. That is not venture capital. That is a governance failure with better branding.
Factory did not need another investor this week. It needed a grown-up in the room.
The AI coding startup raised $200 million at a $5 billion valuation on September 15. Then, barely two weeks later, its CEO publicly accused a board adviser of compromising trust before that adviser joined Factory’s larger rival, Cognition. Vinod Khosla — whose firm backs both companies — then publicly called Factory a “struggling second tier competitor.”
On October 5, Menlo Ventures announced it had joined Factory’s latest financing round. That is more than a cheque. It is a very public signal that somebody with money, reputation and a functioning risk department thinks Factory is worth backing despite the circus.
But don’t miss the bigger point: the startup world has become far too casual about investors owning slices of direct competitors, sitting near sensitive information, and pretending that a polite memo about conflicts solves everything.
It doesn’t.
The $5 billion company that got publicly kneecapped
Factory was founded in 2023 by Matan Grinberg and Eno Reyes. It sells AI software-development tools to enterprises — not just a chatbot that writes a few lines of code, but a platform designed to help companies plan, build, test, secure and maintain software through AI agents.
On September 15, Factory announced a $200 million round at a $5 billion valuation. The company said the financing brought its total funding to more than $400 million. Its named backers included Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, Evantic Capital, Sound Ventures, NEA, Mantis VC and Clearlake.
That valuation was more than triple the $1.5 billion Factory claimed when it raised $150 million in April. That is an absurd jump in five months by normal standards. In AI coding, apparently it is Tuesday.
Factory says customers including Nvidia, Blackstone, RBC, Palo Alto Networks, Adobe and T-Mobile are using its platform. The pitch is straightforward: individual coding copilots are useful, but enterprises want an operating system for AI-assisted software creation — one that gives them governance, visibility and control over which models touch their code.
That last bit matters. Big companies do not simply want more code generated. They want fewer security disasters, fewer broken releases, lower costs and someone accountable when the machine gets ambitious. The winner in this category will not be the firm with the flashiest demo. It will be the one that becomes trusted enough to sit inside a serious company’s software-production line.
Then Factory’s funding victory became a boardroom mess played out in public.
The Chris Degnan dispute is the real story
On September 30, Grinberg said Factory had removed Chris Degnan, a partner at RPT Partners and a Factory board adviser, over what he described as unethical conduct involving Cognition. Degnan then announced he was joining Cognition — maker of the Devin coding agent — as chief revenue officer.
Grinberg alleged that Degnan had been talking with Cognition while attending Factory board meetings and advising its leadership. Degnan disputed the account. He said he had resigned from Factory before joining Cognition, had not shared confidential information, and had declined a full-time role offered by Factory.
Cognition’s chief executive, Scott Wu, also said Cognition had no interest in Factory’s confidential information. As of now, the public accounts establish a nasty dispute, not proof that Factory information changed hands. That distinction matters. Founders should not turn allegations into facts just because they are furious.
But the governance issue remains even if every claim of leaked information proves false.
A board observer or adviser with access to roadmaps, commercial plans, customer details and internal debates cannot casually wander into a direct rival’s executive team without blowing up trust. Whether the formal paperwork says “board member,” “observer” or “adviser” is beside the point. If you are in the room when the confidential stuff is discussed, you carry a duty that lasts longer than the meeting.
This is why the modern VC line — “we back competitors all the time” — deserves far more scrutiny than it gets.
Khosla Ventures invested in both Factory and Cognition. Cognition had just raised $2 billion at a $48 billion valuation, according to TechCrunch. Khosla then entered the public argument and attacked Factory’s CEO and business position. A Khosla Ventures partner, Keith Rabois, separately defended Factory’s view that interviewing with a competitor while retaining board-level access is unacceptable.
That is not a clean governance story. It is an internal contradiction exposed in daylight.
Why Menlo’s move matters
Menlo Ventures did not disclose the size of its Factory investment. But its October 5 announcement was unusually emphatic. Menlo described Factory as a system designed to automate the full software-development lifecycle, emphasising enterprise governance, transparency, sovereign deployment and a model-agnostic approach.
Translation: Menlo is betting Factory is not merely another coding-agent wrapper hoping to surf a hot market. It is betting that enterprises will pay for the layer that controls agents, workflows, models, permissions and measurement across the whole engineering organisation.
That is a credible investment thesis. It is also a bloody hard business to build.
The coding-assistant market has moved from autocomplete to agents, and from agents toward more autonomous systems. Menlo argues that coding is AI’s first killer use case and says the category has grown from $550 million to more than $30 billion in two years. That is Menlo’s own market framing, so treat it as investor advocacy rather than holy scripture. But the underlying commercial logic is hard to dispute: software is a huge cost base, engineers are expensive, and businesses will spend aggressively if AI can safely reduce cycle time.
Factory’s own claim that its routing product can cut token spend by more than 60% while maintaining frontier-model performance is exactly the sort of promise enterprise buyers should test brutally. Not applaud. Test.
Show me the baseline. Show me the workloads. Show me the security exceptions. Show me the rework created by bad output. Show me whether deployment time actually fell. Then we can talk about savings.
The overlooked angle: the real moat is trust, not code generation
Everyone is obsessed with whether AI will replace developers. That is the lazy question.
The better question is: who gets trusted to run the workflow around the developers?
Code generation is becoming a feature. Every major model provider, cloud giant and well-funded startup wants in. The economic prize will sit higher up the stack: governance, integration, auditability, proprietary organisational context, security controls and a verified record of what the agents did.
That is why Factory’s positioning makes sense. It is also why this board dispute is so damaging.
You cannot sell a “trusted software factory” while the market watches a fight about who knew what, who sat in which meeting, and whether a senior adviser was negotiating with the competition. The product may be secure. The governance optics were not.
Founders often underestimate this. They think governance is a legal cost imposed after success. Wrong. Governance is commercial infrastructure. It becomes visible when the business gets valuable enough that people start fighting over the upside.
A clean cap table and disciplined information rights are not boring administration. They are what stop your next funding round, acquisition process or board meeting becoming a public embarrassment.
What this means for you
If you are a founder, use this tomorrow:
1. Map your conflicts before the next round, not after it. Ask every proposed investor which direct and adjacent competitors they back, what information rights they expect, and whether anyone from their firm has an operating role with a rival.
2. Separate access from status. A person called an “adviser” can be just as dangerous as a director if they attend board meetings or see board materials. Give access on a need-to-know basis. Sentiment is not a control system.
3. Put transition rules in writing. If a board member, observer or adviser explores a role at a competitor, require immediate disclosure, recusal from sensitive discussions and a clear exit protocol. Do this before anyone is tempted.
4. Do not confuse a giant valuation with safety. Factory’s rise from a $1.5 billion valuation in April to $5 billion in September looks brilliant. It also raises the cost of every operational mistake. The more money and attention you attract, the less room you have for sloppy governance.
5. For buyers: buy proof, not AI theatre. If a vendor claims autonomous engineering, ask for measurable evidence: deployment frequency, incident rates, review burden, security outcomes, token costs and the human-hours genuinely removed. A slick demo is not an operating model.
Menlo’s investment says Factory still has heavyweight believers. Fair enough. The company may well build something enormous.
But the lesson is bigger than Factory: in the AI gold rush, capital is everywhere and confidentiality is suddenly treated as optional. Smart founders will treat trust as a product feature, a board responsibility and a competitive advantage.
Everyone else will learn why the hard way.