Factory raised $200M at a $5B valuation and its droids now run inside Nvidia, RBC, Adobe, and T-Mobile. A funding round is not a safety audit.
A chipmaker, a bank, a design-software giant, and a telecom carrier are now running autonomous coding agents in production, and the company behind them just raised $200 million to expand. Factory closed the round at a $5 billion valuation on September 15, 2026, more than tripling the $1.5 billion valuation it held just five months earlier. Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, and Evantic Capital led it, and Factory disclosed Nvidia, Royal Bank of Canada, Adobe, and T-Mobile as customers, alongside Blackstone itself and Palo Alto Networks. Total funding now tops $400 million.
Sources: Factory: $200M at $5B valuation
The valuation jump is the headline. The more revealing detail is who is deploying the product: a chipmaker whose design tools underwrite most of the industry's semiconductors, a systemically important bank, a design-software company that holds the working files of a huge share of the world's designers and editors, and a telecom carrier running network and billing infrastructure that tens of millions of people depend on daily. Those are not companies that experiment casually with a codebase. Enterprise buyers at that scale putting real budget behind autonomous coding agents, at this pace, says as much about the gap between adoption speed and governance maturity in AI coding as it does about Factory's product.
How much did Factory raise, and what changed since April?
Factory's Series C closed in April 2026: $150 million led by Khosla Ventures at a post-money valuation of $1.5 billion, itself about seven months after a $50 million Series B in September 2025. The round announced September 15 more than tripled that valuation to $5 billion on $200 million in new capital, led by Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, and Evantic Capital, with Sound Ventures, NEA, Mantis VC, and Clearlake also participating. Individual investors Marc Benioff, Brad Gerstner, and former Formula One driver Nico Rosberg joined as well. Factory has now raised more than $400 million in total.
Sources: TNW: Factory raises $200M at $5B valuation
Three priced rounds in twelve months, each one arriving faster and richer than the last, is a funding cadence, not a governance cadence. A Series B closes, a company hires and signs pilot customers, a Series C prices that traction seven months later, and five months after that a growth round more than triples the number again. Revenue and logos can move on that timeline because a signed enterprise contract is a single decision by a handful of executives. A review process that can be trusted with thousands of autonomous changes a week moves on a different, slower timeline, because it has to survive contact with every team, every codebase, and every edge case those teams find.
What is Factory, and what do droids actually do?
Factory describes itself as "an agent-native software development platform" whose droids, in Factory's own words, "plan, build, review, test, and ship software across the tools engineering teams already use." Its agents, called droids, work across the command line, desktop, web, mobile, Slack, Microsoft Teams, Jira, Linear, and CI, and are built to sit alongside a team's existing tools rather than replace them.
Sources: Factory product overview
- Backlog work: picking up tickets and shipping the change
- Pull request review and test generation
- Incident investigation and root-cause analysis
- Documentation maintenance across a codebase
- Recurring delivery workflows triggered by events or schedules
Factory's own framing splits engineering work into an outer loop and an inner loop. The outer loop, reasoning about requirements, talking to colleagues and customers, and making architectural calls, stays with people. The inner loop, writing code, running tests, building, and reviewing, is where droids operate. Factory 2.0, launched in April 2026, extended this from individual coding agents to what CEO and co-founder Matan Grinberg calls "software factories that serve as the core foundation from which an entire software company operates." The platform also ships Factory Router, which Factory says cuts token spend by more than 60 percent, and offers managed cloud, on-premise, or air-gapped deployment for regulated sectors.
Factory ships three ways to run a droid, and the differences matter for how much autonomy a team is actually granting. The Droid CLI is interactive: a person is in the loop turn by turn, the way a pair-programming session works. Missions hand off a multi-phase project and let a droid work through it with less turn-by-turn steering. Automations run entirely on triggers, an event in Jira, a schedule, a webhook from CI, with no person prompting the session at all. Factory also describes persistent "Droid Computers" that keep state across a task instead of starting cold each time. Moving from CLI to Missions to Automations is a straight line toward less supervision per unit of code shipped, and it is the Automations tier, running unattended on a schedule inside a bank's or a chipmaker's CI pipeline, where a governance gap costs the most.
That range of integrations is also an access-control question, not just a workflow one. A droid that can act in Slack, Teams, Jira, Linear, and CI needs credentials scoped into each of those systems, and a token broad enough to file a PR is often broad enough to read a repository's secrets or trigger a deploy. Least-privilege access for AI coding agents covers how to scope those tokens and deploy keys so a compromised or misdirected droid session cannot reach further than the one task it was given, a control that matters more, not less, as a platform adds more surfaces for an agent to act on.
Why do Nvidia, RBC, Adobe, and T-Mobile matter more than the valuation?
Each named customer sits in a codebase where a bad autonomous change is expensive in a specific way. Nvidia's software stack underwrites chip design and drivers the rest of the industry builds on. Royal Bank of Canada operates inside banking regulation, where a change to a payments or risk system carries audit and compliance obligations that outlast any sprint. Adobe's products hold the working files of a huge share of the world's designers and editors. T-Mobile runs network and billing infrastructure that tens of millions of subscribers depend on every day. None of these are greenfield side projects. They are the kind of codebase where a mistake propagates to customers, regulators, or both.
A bank and a telecom carrier in particular raise the bar further: audit and accountability requirements do not relax because the code shipped faster. Shipping AI code in a regulated codebase covers what SOC 2, HIPAA, and GDPR environments specifically require of AI-generated changes, and the same logic applies to any bank or carrier deploying droids at scale, whether or not the regulation is named on the label.
Does raising $200 million prove Factory's droids are safe?
It does not, and Factory itself does not claim otherwise. A growth-stage funding round is underwritten on revenue trajectory, market position, and competitive moat, not on a security or reliability audit of the code an agent ships. Factory's public materials do not disclose a defect rate, an incident count, or an independent audit of droid-authored changes at any of its named customers. A $5 billion valuation says investors believe enterprises will keep paying for this category. It says nothing about how many of those enterprises have a review gate strong enough to catch a bad autonomous change before it reaches production.
Revenue and valuation growth measure market demand, not code quality. Treating the two as the same thing is exactly how a governance gap goes unnoticed until an incident forces the question.
Factory's own product pages are specific about what ships today and what does not. Laid out plainly, the split looks like this:
| Capability | What Factory provides | What the enterprise still has to define |
|---|---|---|
| Human approval checkpoints | Configurable per workflow | Which changes require one, and who is accountable for the decision |
| Deployment isolation | Managed cloud, on-premise, or air-gapped | Data classification and access boundaries around each option |
| Model routing and cost | Factory Router, claimed 60%+ token reduction | Whether cheaper inference changes how deeply a change gets reviewed |
| Usage and spend visibility | Agent Effectiveness analytics | Whether higher droid usage correlates with more rework or incidents |
| Merge or ship gate enforcement | Not part of the platform | The actual blocking mechanism that stops an unresolved issue from shipping |
What does enterprise adoption at this speed mean for AI coding governance?
Speed of adoption and maturity of governance are not the same curve, and they rarely move together. A vendor can roll out an autonomous coding agent to hundreds of thousands of developers faster than any single customer can build the review discipline, ownership model, and audit trail that make autonomous changes safe to trust at that volume. Rolling out AI coding tools across a team without chaos lays out what that discipline actually requires: shared conventions, one review gate everyone goes through, CI guardrails, and metrics that surface rework instead of hiding it. None of that scales automatically just because a vendor's revenue does.
The gap tends to show up first as policy without a mechanism. A company can publish a rule that every droid-authored change needs a human sign-off, the same way Factory's approval checkpoints are configurable per workflow, and still ship an incident, because a checkpoint nobody is required to use is not a control. Your AI coding policy template enforces nothing makes the same point about a different vendor entirely: a rule with no mechanism and no owner is not a policy, it is a hope written down.
What should a regulated-adjacent team check before deploying droids at this scale?
- Is the human approval checkpoint mandatory for high-risk changes, or an opt-in a team can quietly skip under deadline pressure?
- Does the platform produce an audit trail that ties a shipped change back to the spec, the review, and evidence that tests actually passed?
- Who owns the decision to block a merge: a person who can be overruled, or an enforced mechanism that cannot?
- Is usage volume measured against defect or incident rate, not just tickets closed or tokens saved?
- Is the air-gapped or on-premise option actually deployed for the workloads that need it, or is it a line in the sales conversation?
Ask a vendor for the incident postmortem process for an autonomous change gone wrong before signing, not after. If the answer is vague, the governance program is too.
Factory's platform already gives enterprise buyers real building blocks: configurable approval checkpoints, deployment isolation, and spend visibility. What it does not claim to be is a merge gate. TLM Forge is built around that specific gap. Reviewers work with an adversarial mandate, a threat-modeler at design time, a red-team pass on the finished diff, instead of assuming the agent's own work is basically right, and whether a change ships comes down to a scored gate, not a checkbox someone can leave unchecked under deadline pressure. A bank's compliance team or a chipmaker's release process can point to a mechanism that ran, not a policy that was merely available.
Frequently asked questions
01How much did Factory raise and at what valuation?
Factory raised $200 million on September 15, 2026, at a $5 billion valuation, more than tripling the $1.5 billion valuation it held after its April 2026 Series C. Total funding is now over $400 million, led by Blackstone, Khosla Ventures, Sequoia Capital, Insight Partners, and Evantic Capital.
02What are Factory's "droids"?
Droids are Factory's AI agents for software engineering. They work across the command line, desktop, web, Slack, Teams, Jira, Linear, and CI to handle backlog work, pull request review, test generation, incident investigation, and documentation, while engineers focus on requirements and architecture.
03Which companies use Factory's droids?
Factory names Nvidia, Royal Bank of Canada, Adobe, T-Mobile, Palo Alto Networks, and Blackstone, which is also an investor in this round, as customers. Factory says hundreds of thousands of developers use the platform.
04Does Factory's funding round prove its code is safe?
No. A funding round is underwritten on revenue and market position, not a security or reliability audit. Factory has not published a defect rate or independent audit of droid-authored changes at its named customers, so the valuation is not evidence of code safety.
05Does Factory offer human review of autonomous changes?
Factory's platform includes configurable human approval checkpoints per workflow. Configurable is not mandatory: whether a checkpoint is required for high-risk changes, and who enforces it, is a decision each enterprise customer still has to make itself.