Three B2B companies got acquired for around $3 billion each in the space of thirty days. On the surface they have nothing in common — a customer-service AI, a factory maintenance app, an industrial data platform. Look closer and they were all bought for the same reason. Each one owned data the buyer couldn't get anywhere else.
Salesforce bought Fin (formerly Intercom) for $3.6 billion in mid-June — an AI agent resolving over two million customer conversations a week. Autodesk paid $3.575 billion for MaintainX, its largest acquisition ever, for a mobile-first maintenance and work-order tool used by frontline workers. Schneider Electric acquired Cognite for $3.1 billion, a platform that contextualises industrial data for manufacturing and energy.
Different industries, one pattern, well summarised by the SaaStr analysis that surfaced it: "Models are commoditizing on a monthly basis. Proprietary operational data that took years to accumulate is not." The buyers weren't paying for software. They were paying for datasets that can't be reproduced — the accumulated record of how millions of support tickets get resolved, how machines actually fail on real factory floors, how industrial processes behave in the wild.
There's a startling detail in the numbers. Across all three deals, the revenue multiple came out at roughly half the growth rate: MaintainX growing 50% fetched 26x forward revenue, Cognite at 36% growth got 18x, Fin's blended rate landed around 9x. When a pattern is that clean across three unrelated acquirers, it isn't coincidence — it's a market pricing the same underlying asset the same way. And that asset is proprietary, compounding data.
The questions the analysis poses to founders cut straight to it: What proprietary data compounds as customers use your product? Does your data moat strengthen automatically as the product runs? Would a legacy platform pay half your growth rate as a multiple just to avoid rebuilding what you've accumulated?
Strip it down and the message is stark. In a world where models are commoditising monthly, the durable value isn't the model or even the application — it's the data that only exists because your specific product ran for years in a specific domain. Anyone can rent a frontier model. Nobody can rent the five years of operational history that makes your model useful.
Which raises the question every software business should be asking: who owns the data your product generates? Because if the answer is "a platform vendor," then the compounding asset — the thing worth 18 to 26 times revenue — is theirs, not yours. You're accumulating the moat and handing it to your landlord.
This is the structural case for building on infrastructure you own, and it's exactly what VBWD is designed for. VBWD is a self-hosted, source-available platform for stores, subscriptions, marketplaces, and AI apps, and its defining property is simple: the data your product generates lives in your database, on your infrastructure. The compounding asset accrues to you.
That's not an abstract benefit given what these acquisitions just demonstrated. Every customer interaction, transaction, and behavioural signal your VBWD-based product captures is yours to keep, analyse, and build on — the raw material of exactly the data moat that commanded these multiples. VBWD even ships a dataset and data-marketplace layer, so if your accumulated data becomes valuable to others, selling access to it is a native capability rather than a bolt-on. And because the AI features run against a model connection you configure, you can turn that proprietary data into intelligence without shipping it to a third party who might learn from it too.
The contrast is the whole point. Build your product on a closed platform, and you spend years accumulating a data moat that legally belongs to someone else. Build it on VBWD, and the moat is yours — to keep, to compound, and, if it comes to it, to be the reason someone pays you 20x revenue.
Owning your data means owning the work: self-hosting, security, and stewardship are real responsibilities, and for some teams a managed platform is the better trade despite the ownership cost. Data only becomes a moat if your product actually generates something proprietary and compounding — infrastructure can't manufacture that, only preserve it for you. And VBWD is one own-your-stack option among several.
But the lesson of three $3-billion cheques in thirty days is hard to argue with: in the AI era, the durable value is proprietary data, and proprietary data is only yours if you own where it lives. VBWD is free for commercial use below 6.7 BTC a year in attributable sales. Start with the architecture, the plugins, and the source on GitHub.