n8n's Valuation Doubled to $5.2B — So Much for 'AI Killed Low-Code'
Column Overview
Every few months, the same verdict resurfaces on tech Twitter and in venture memos: AI can now write code on its own, so drag-and-drop workflow builders are finished. Why would anyone click nodes onto a canvas when a chatbot can generate the same automation script in thirty seconds, no low-code middleman required? It's a tidy story. It also happens to be contradicted by the actual numbers coming out of the two biggest companies in that exact category.
The valuations say otherwise
Start with n8n, the open-source-rooted workflow automation platform that lets users chain together APIs, AI models, and business logic on a visual canvas. In October 2025, it closed a $180 million Series C led by Accel, with Meritech and Deutsche Telekom's venture arm among the participants, pushing its valuation to $2.5 billion. That alone would have been a strong year. Instead, by the second quarter of 2026, SAP made a strategic investment that reset the valuation to $5.2 billion — more than double, in well under twelve months. Ahead of the Series C, n8n's user base had grown sixfold and its revenue tenfold in a single year; the company's annual recurring revenue is now pushing past the $100 million mark, with 2025 revenue landing around $40 million, and it counts more than 230,000 active users across its free and paid tiers.
Dify, the other name that keeps coming up in this conversation, tells a similar story from a different angle. In March 2026 it raised a $30 million Pre-A round led by HSG, with GL Ventures, Alt-Alpha Capital (the new fund spun out of Bessemer Venture Partners), Wu Capital, Mizuho's Leaguer Investment, and NYX Ventures all participating, valuing the company at $180 million. Its GitHub star count crossed 100,000 in June 2025 and has since climbed past 144,000. Since launching in 2023, it has picked up more than 2,000 teams and over 280 enterprise customers.
None of that reads like a category in its death throes. If anything, capital is flowing into these two companies at a pace that outstrips most of the "AI-native" startups that are supposedly making them obsolete. So the interesting question isn't whether workflow platforms are dying — the money says they aren't — it's why the "AI writes code now, so who needs this" argument feels intuitive in the first place, and what that argument actually gets wrong.
What a workflow platform is really selling
The mistake in the "AI killed low-code" thesis is treating these platforms as if their entire value proposition were "you don't have to write code." That was always only one piece of a bundle, and it was arguably never the most defensible piece. Peel back what n8n or Dify actually hands a customer and there are at least four separate things stacked on top of each other.
The first is a runtime — scheduling, retry logic, credential and secrets management, the unglamorous plumbing that means nobody has to stand up and babysit a server just to run an automation on a cron schedule. The second is a connector ecosystem: hundreds of pre-built integrations to third-party APIs, each one having already solved the tedious part of OAuth flows, rate limits, and API quirks that a freshly generated script has to solve from scratch, every single time. The third is observability — a visual trace of what actually happened when a workflow ran, built for the marketing manager or ops lead who will never read a stack trace but needs to know why yesterday's automation silently failed. The fourth is a collaboration surface: the ability for a non-engineer to open a workflow and adjust it without filing a ticket and waiting for an engineering sprint to free up.
Generative AI, for all its genuine progress, has made real inroads on exactly one of those four things: writing the logic itself. It has done essentially nothing to the other three and a half. Code that an AI model generates in a chat window still needs somewhere to run, still needs someone managing its credentials, and still fails in ways that a non-technical stakeholder needs a way to see and understand. Those are precisely the layers n8n and Dify have spent years building — and precisely the layers that don't show up when someone frames the comparison as "chatbot versus canvas."
The platforms are absorbing the thing that was supposed to kill them
What's actually shifting is subtler than a death sentence, and it's visible in how both companies are spending their new capital. Dify has been explicit that its Pre-A funds are going toward folding more advanced agentic capabilities directly into its core product, alongside building out a dedicated enterprise team and hardening performance and compliance. That's not a company bracing for AI code generation to eat its lunch — it's a company treating AI generation as a feature to absorb into the product, the same way earlier waves of automation tooling absorbed cloud APIs and, before that, absorbed scripting languages.
Seen that way, the more useful question shifts from "will workflow platforms survive AI" to "who inside this category doesn't survive it." Industry observers following the space tend to point to a specific tier: smaller vendors whose entire pitch was the drag-and-drop canvas itself — no code required — with no meaningful connector library, no runtime infrastructure at scale, and no enterprise trust layer behind it. For that tier, AI code generation genuinely does erode the core selling point, because "you don't have to write code" was the whole product. Without a moat built from infrastructure, integrations, or accumulated enterprise relationships, that group is the one actually facing the pressure the "workflow tools are dying" headlines describe — just not the household names the headlines usually reach for.
The part AI still hasn't touched
Zoom out and this fits a pattern that shows up every time a new capability triggers a "tool X is now obsolete" reflex. Finishing a task is only half of what any piece of software is actually selling. The other half is making that task happen reliably, repeatably, and in a way that people who aren't engineers can still participate in and trust. Generating a script is the easy half. Keeping it running six months later, when an API changes its auth scheme or a business user needs to tweak a single condition without opening a pull request — that's the half infrastructure companies have always been built around, and it's the half that AI code generation, as capable as it's gotten, still hasn't found a way to replace.
That reframes what n8n and Dify should actually be anxious about. It was never "will an AI make our product unnecessary." It's whether they can absorb AI capability into their platforms faster than competitors do, and faster than a new generation of AI-native tools can build the infrastructure layer these companies already have. On the evidence of the last year — rising valuations, growing revenue, expanding enterprise rosters — both are winning that race so far. Whether that holds once the next funding cycle tests it, and whether the mid-tier "canvas-only" players find a way to build a moat before the runway runs out, is the more interesting question left open.