Factory closed a $150M Series C in April 2026 at a reported $1.5B post-money valuation. Khosla Ventures led the round, with participation from Sequoia Capital, Blackstone, Insight Partners, Evantic Capital, 20VC, NEA, Mantis VC, Lux Capital, J.P. Morgan, NVIDIA, Abstract Ventures, SV Angel, and BoxGroup.

Founded in 2023 by Matan Grinberg and Eno Reyes, Factory raised a $5M seed round that November. In June 2024, the company raised a $15M Series A at a $120M valuation. Factory then raised a $50M Series B in September 2025 at a reported $300M valuation, with NEA, Sequoia, NVIDIA, and J.P. Morgan among the investors.

Factory has raised $220M in total disclosed funding across four rounds.

Factory builds autonomous AI agents called Droids that handle multi-step software engineering tasks rather than only suggesting code snippets. A developer specifies an outcome, such as fixing a bug, migrating a service, or reviewing a pull request. Droid then explores the repository, creates a plan, edits files, runs commands and tests, inspects failures, revises its work, and submits a diff or pull request for human approval.

The product uses a shared agent runtime across four interfaces. The desktop and web app lets developers connect a codebase, select a model, and monitor Droid's work, while the CLI runs the agent in terminal environments. A TypeScript and Python SDK embeds Droid into internal tools, ticket-to-code pipelines, and CI/CD workflows, and managed connectors integrate with systems including Slack, Linear, GitHub, Sentry, Datadog, Salesforce, and Figma.

For larger projects, Missions provides a multi-agent orchestration layer. Users define goals, milestones, and success criteria, while Mission Control coordinates orchestrator, worker, and validator agents. Missions can run for hours or days on persistent Droid Computers that retain installed packages, repositories, and running services between sessions.

The Software Factory layer applies these capabilities to repeatable workflows across the development lifecycle, including automated pull request review, security review with STRIDE and OWASP coverage, browser- and terminal-based QA, incident response triggered by Slack alerts, and documentation generation. A dashboard tracks automation coverage by repository, workflow health, tickets triaged, and PRs merged.

Factory supports models from Anthropic, OpenAI, Google, xAI, and open-weight alternatives. Customers can provide their own API keys, use Factory Router to optimize for cost, speed, and performance, or run open models locally in air-gapped environments. Enterprise deployment options range from standard SaaS to fully on-premises infrastructure with customer-managed encryption keys and SOC 2 Type II, ISO 27001, and ISO 42001 compliance.

Factory combines product-led SaaS with enterprise platform software, selling B2C to individual developers and B2B to engineering organizations. Individual developers adopt Droid through self-serve plans starting at $20 per month, creating internal champions who bring the product into team and enterprise procurement.

Pricing combines subscriptions with usage-based consumption. Each plan includes a pool of credits governed by rolling five-hour, seven-day, and 30-day limits, with models consuming credits at different multipliers based on inference cost. When included usage runs out, customers can switch to cheaper open-weight models or purchase prepaid credit packs. Enterprise contracts add seats, shared consumption commitments, dedicated compute, deployment complexity, and support tiers.

Cost to serve varies by task: a quick code question might cost pennies, while a multi-day Mission running frontier reasoning models across several repositories can consume substantial inference and compute. Rolling limits, model multipliers, BYOK options, and routing optimization distribute that cost risk more directly than flat per-seat pricing.

Gross margins face the AI tax common across the category. Sacra has previously estimated margins of 23% for Replit and 35% for Lovable, well below traditional SaaS norms. Factory manages margin through its orchestration layer, routing easy work to cheaper models, allowing customers to bring their own keys or run open models locally, and pricing long-running autonomous work through consumption rather than unlimited seat fees.

Expansion can occur as a single developer adds a team, one repository becomes a portfolio, coding extends into review, QA, security, and incidents, and assisted sessions shift to fully automated workflows. The August 2026 launch of a $100M partner network with systems integrators like Wipro extends distribution into large global enterprises without requiring Factory to build its own professional services organization.

Factory competes in a converging market as AI IDEs, autonomous coding agents, model providers, source-control platforms, and DevOps vendors expand toward an enterprise control plane for autonomous software production.

Anthropic, through Claude Code, and OpenAI, through Codex, are structural competitors because they control the underlying model economics and can integrate vertically into developer workflows. OpenAI Codex exceeded five million weekly users by June 2026 and is bundled into existing ChatGPT enterprise relationships rather than requiring separate procurement. Claude Code has bottom-up developer adoption through terminal-native workflows and can run against models in existing Amazon Bedrock or Google Vertex AI environments.

Both vendors are adding enterprise governance features, including audit logs, role-based controls, and sandboxing, that were previously a Factory differentiator. Factory's counter is model neutrality: customers seeking to avoid strategic dependency on a single provider can use Factory as an orchestration layer across providers.

Cursor disclosed over $500M in ARR and adoption by 53% of the Fortune 1000. It began as an AI-assisted editor but has expanded into background agents, API-driven automation, and enterprise analytics, bringing it into direct competition with Factory. Cognition combines its Devin autonomous agent with the Windsurf IDE and proprietary SWE models in an integrated model-to-IDE stack. Cognition reportedly raised roughly $1B at about a $25B valuation in May 2026 and was discussing another round at $40B or more.

Cursor and Cognition own the developer interface, which Factory lacks. Factory differentiates through broader lifecycle coverage, sovereign deployment, and compatibility across IDEs and terminals rather than requiring developers to switch editors.

GitHub Copilot is a distribution threat because GitHub owns the repository, issue, pull request, review, Actions, and enterprise policy surfaces where autonomous work is initiated and approved. GitHub has developed an agent marketplace and governance layer that allows Claude, Codex, Copilot, and custom agents to run with shared context under a unified Agent Control Plane, reducing differentiation among standalone vendors offering model neutrality. Organization plans start at $19–$39 per user per month, below Factory's apparent price.

AWS Kiro integrates with Bedrock, DevOps Agent, and Security Agent while holding FedRAMP High and DoD IL4/5 authorization, giving it a procurement advantage in government and defense. Google Jules is exploring proactive agents that continuously monitor codebases rather than waiting for user-initiated tasks. Factory's response is cloud and platform neutrality, with support for GitLab alongside GitHub, multiple clouds, and fully air-gapped environments that cloud-first competitors cannot easily match.

Higher code volume from autonomous agents shifts bottlenecks downstream into review, testing, security, and deployment. Specialized vendors such as CodeRabbit, Greptile, Snyk, Semgrep, and Sonar address narrower portions of Factory's workflow, with potentially stronger accuracy in their respective domains.

Factory has added automated PR review, STRIDE-based security analysis, Droid Shield secret detection, and QA workflows that operate applications through browsers and terminals. The risk is that these specialists add code generation and remediation capabilities. AWS Security Agent already validates vulnerabilities by simulating exploits and can trigger fixes through Kiro and Claude Code.

Factory's initial wedge is autonomous coding, but the company is expanding across the software development lifecycle, into other enterprise functions, and into regulated and sovereign markets that cannot use conventional cloud-based agents.

Factory has moved from code generation into an operating layer spanning triage, planning, implementation, testing, code review, security, release, documentation, and incident response. Each lifecycle stage addresses budgets previously spread across separate QA, DevSecOps, observability, and knowledge management tools. The same codebase context, organizational policies, model router, and agent memory can be reused across stages: an incident can be connected to the responsible pull request, a security finding can inform code review, and a release can trigger updated tests and documentation.

Legacy modernization adds work across refactors, migrations, testing, documentation, and environment remediation, areas historically served by labor-intensive consulting engagements. Factory's partner program explicitly targets mainframes, data platforms, and technical debt as migration factory opportunities.

Factory launched Personas and managed Connectors on August 25, 2026, extending Droids beyond engineering into Product, Design, Finance, Marketing, Sales, and Operations. With governed access to systems such as Salesforce, Stripe, Notion, Amplitude, and PagerDuty, Droids can synthesize customer feedback, reconcile financial forecasts, analyze sales pipelines, and automate cross-application workflows.

The agent runtime already has file-system access, computer use, multi-step reasoning, and governed write capabilities, so expansion primarily requires domain-specific onboarding, permissions, and connectors rather than a different agent architecture. OpenAI reported that non-developers represented roughly 20% of weekly Codex users by June 2026 and were growing more than three times as fast as developers, indicating demand beyond engineering teams.

Factory's deployment architecture spans SaaS, hybrid, dedicated compute, on-premises, EU-specific data residency, and fully air-gapped environments with local open-weight model execution. These options address financial services, government, defense, healthcare, and critical infrastructure, sectors where source code and telemetry cannot leave the customer's network.

Factory opened a New York office partly to serve financial institutions and disclosed work with two of the world's largest global banks. The $100M partner network commitment targets consulting firms and systems integrators that can package Droids into repeatable migration, compliance, and managed software-factory offerings across the United States, EMEA, and Asia-Pacific.

Platform compression: OpenAI, Anthropic, GitHub, and AWS are each expanding from models, repositories, and cloud infrastructure into full-lifecycle agent orchestration with enterprise governance, potentially bundling Factory's core capabilities into broader subscriptions and reducing demand for a standalone agent platform.

Inference economics: Factory's gross margins are structurally exposed to the AI tax because longer, more autonomous agent tasks consume proportionally more third-party model inference and compute, while competitive pressure to offer generous bundled usage could compress margins before routing efficiency and model cost declines catch up.

Autonomy liability: Droids with access to repositories, terminals, cloud infrastructure, and business systems can cause more damage than conventional developer tools through prompt injection, credential exposure, destructive commands, or cascading errors in long-running Missions, creating a tension between the autonomy required to deliver ROI and the authority enterprises are willing to grant.