# Three $3B B2B Acquisitions in 30 Days: Intercom/Fin, Cognite, and MaintainX. They All Bought the Same Thing: Data for AI > Source: > Published: 2026-07-16 14:10:45+00:00 ### Three $3B B2B Acquisitions in 30 Days: Intercom/Fin, Cognite, and MaintainX In about a month, three B2B companies recently were acquired for right around $3 billion each. On the surface they have nothing to do with each other: **Salesforce bought Fin**(the company formerly known as Intercom) for**$3.6B** on June 15.**Autodesk bought MaintainX** for**$3.6B** on May 28.**Schneider Electric bought Cognite** for**$3.1B** on June 30. One does customer service. One does facility and equipment maintenance. One does industrial data plumbing for oil rigs and power plants. Different buyers, different categories, different continents. But in many ways, as deals … they are the same deal. **Each one is a legacy platform paying a giant premium to own the proprietary data that makes AI actually work in its domai**n. Not the model. The data. And when you line up the revenue and growth numbers, a second pattern shows up that’s even more useful: the price each buyer paid tracked one thing above all else, and it wasn’t revenue. Here’s all three in detail, then what they share, then what it means if you’re building. ## Deal 1: Salesforce buys Fin (Intercom) for $3.6B This is the most interesting of the three once you look under the ARR headline, so it’s worth getting the numbers exactly right. Fin started life as Intercom in 2011, one of the original darlings of the B2B messaging era. By 2022 it was in trouble: five straight quarters of declining net new ARR, growth approaching zero. Six weeks after ChatGPT shipped, the team had a working prototype of an AI support agent. They bet the company on it, turned over about 40% of staff, and rebuilt around that agent. In May 2026 they renamed the whole company Fin, after the agent. Five weeks later Salesforce bought it. The number that gets quoted is **$400M+ in total ARR**. That number hides the actual story. Break it apart: - The **Fin AI agent line crossed $100M ARR** and is growing**350% year over year (3.5x)**. - That AI line is roughly **a quarter of total ARR** and, per Fin’s own disclosures,**virtually all of the company’s net new growth**. - The other ~$300M, the legacy Intercom seat-based messaging and support business, is close to flat. The whole company reaccelerated from near-zero growth only to the mid-teens, and the AI line is why. - Outcome-based pricing on Fin pushed **net revenue retention from 112% to 146%**. - Fin resolves **2M+ conversations a week** across**8,000 customers on the agent**, inside** 30,000+ customers**on Intercom products overall. Named accounts include Anthropic, DoorDash, Mercury, Asana, and Riot Games. So the real shape of the deal: Salesforce paid **9x on the blended $400M**, but the thing it was actually buying is a **$100M line growing 350%**, and a $100M line growing 350% is worth 30x or more on its own. The blended multiple looks cheap only because half the revenue is flat legacy that AI is busy commoditizing. The strategic logic is the model plus the customers. Fin runs on **Apex**, a model it post-trained specifically for support, which it claims beats frontier models from OpenAI and Anthropic on resolution. Fin resolves about **76% of support volume with no human**, against roughly **62%** for Salesforce’s own Agentforce Help Agent. Salesforce has Agentforce growing 205% to $1.2B ARR and still chose to buy the category-definer rather than out-build it. It also removes a competitor and adds 30,000 businesses whose support data now flows through a Salesforce-owned model. This is the biggest exit ever for an Irish-founded tech company, Salesforce’s fifth acquisition of 2026, and its third in June alone after M3ter and Contentful. ## Deal 2: Autodesk buys MaintainX for $3.6B MaintainX is the youngest of the three and got the richest multiple. Chris Turlica and his cofounders started it in 2018 with a boring, correct observation: the software given to frontline and maintenance workers was stuck in 1995. They built a mobile-first work-order and maintenance app as simple as Slack, aimed at the factory floor. The numbers: **$3.575B** headline (Autodesk’s largest acquisition ever), all cash plus new debt, plus**$150M in retention RSUs** for the team.- Entered 2026 at roughly **$115M ARR**, guided to**$135M+ for calendar 2026**, growing** 50%+**. - That’s about **26x forward ARR**. - Raised a **$150M Series D at a $2.5B valuation** in July 2025; total funding was around $243M. **500,000+ frontline workers** and**11M+ assets** under management across roughly 8,000 to 11,000 companies.- Targeted close as early as August 3, 2026. MaintainX didn’t stay a simple CMMS. It kept absorbing adjacent workflow: inspections, safety, asset intelligence, predictive parts, compliance, enterprise asset management. That’s the vertical software playbook run correctly. Start with a stupidly simple wedge, absorb the next workflow, then the data, then the decisioning, and eventually you stop being a tool and become the operating layer. Autodesk folded it into a new division, Autodesk Operations Solutions, next to its Tandem digital twin and simulation tools. The framing in Autodesk’s own deck is blunt: this adds roughly **$40B of Operations TAM** on top of Autodesk’s existing **$78B Design & Make TAM**, and it extends Autodesk’s relationship with a physical asset from a few years of design work to the **50 to 60+ years** that asset actually lives. CEO Andrew Anagnost said it hjimself: AI is only as good as the data it ingests, and MaintainX captures how assets behave under real conditions, which is the context that makes AI accurate and actionable. Design software knows how an asset was supposed to work. MaintainX knows how it actually performs in the field. Nobody at Autodesk could generate that telemetry from design seats. At 26x, they weren’t pricing revenue. They were pricing a data stream they couldn’t build, growing 50%. ## Deal 3: Schneider Electric buys Cognite for $3.1B Cognite is the one most B2B founders have never heard of, and it’s the cleanest illustration of the thesis. Founded in Oslo in 2017 as a spinout of Norwegian industrial conglomerate Aker, it spent nearly a decade on a problem industrial operators rarely talk about: the majority of the data pouring out of factories, grids, and energy assets sits in silos, poorly labeled, and useless to AI. The numbers: **$3.1B**, all cash.**$170M+ revenue** in 2025, with**36% growth in ARR bookings** and fast adoption of its Atlas AI platform. (Worth noting the disclosed growth figure is bookings, a forward indicator, not trailing revenue growth.)- That’s roughly **18x revenue**. **800+ employees**; total funding around $225M, including a $150M Series B from TCV in 2021 at a $1.6B valuation.- Aker walks away with about **$1.48B in cash**, near a 20x return, the largest software exit in Norway’s history. Other backers included Saudi Aramco, Accel, and TCV. Cognite’s two products tell you what Schneider paid for. **Data Fusion** is the contextualization layer, a unified industrial data model plus a knowledge graph that connects decades of messy operational, engineering, and sensor data. **Atlas AI** sits on top with generative and agentic capabilities that act on that contextualized data. Schneider is going to fold Cognite into AVEVA, its industrial software arm (itself built on the 2021 OSIsoft/PI acquisition). The logic is the one every hardware-plus-software incumbent is running into: Schneider sells the switchgear, sensors, and automation, but without owning the data contextualization layer, it depends on a third party every time a customer wants to build an AI workflow on its hardware. That dependency is tolerable when AI is a feature. It’s fatal when AI becomes the product. The knowledge graph is the moat. A frontier model with no access to a refinery’s actual sensor history is useless to a plant manager. Cognite makes decades of siloed operational data queryable and AI-ready, and that’s not something Schneider could rebuild quickly. The 18x reflects scarcity, not current revenue. For context on how hot this corner is: PwC pegged industrial manufacturing M&A at $173B over the past year, up 28%, with mega-deals above $5B now making up 56% of deal value. CB Insights counted 266 AI M&A deals in Q1 2026 alone, up 90% year over year. Everyone is buying the picks-and-shovels data layer instead of competing on model benchmarks. ## The price tracked the growth, at almost exactly half Put the revenue, growth, and multiple side by side and two things jump out. **First, the revenue multiple landed at roughly half the growth rate in all three deals**. MaintainX grew 50% and went for 26x. Cognite’s bookings grew 36% and it went for 18x. Fin’s blended business grew mid-teens and went for 9x. Across three unrelated buyers on three continents, the multiple came out near 0.5x the growth number. That’s not a coincidence. It’s what buyers pay for durable, data-backed growth right now. **Second, Fin proves the rule rather than breaking it**. Its 9x looks low only because the blended growth is dragged down by ~$300M of roughly flat legacy revenue. Strip that away and the thing Salesforce actually bought, a $100M line growing 350%, prices right back up around 36x, in line with MaintainX. The buyers weren’t paying for revenue. They were paying for the rate of change and the data underneath it. ## Data is the acquisition premium now Three different buyers, three different industries, one identical thesis. Each is an incumbent that owns distribution and, in two cases, the physical hardware, but not the data and AI layer that’s about to matter most. Each concluded it couldn’t generate that data internally fast enough. So each paid roughly $3B to own it outright. Notice what none of them paid a premium for: the model. Fin built Apex specifically to cut its dependence on OpenAI and Anthropic, and Apex is a real asset, but it’s not the $3.6B asset. The 30,000 customers and their support data are. Models are commoditizing on a monthly basis. Proprietary operational data that took years to accumulate is not. If you’re building in B2B + AI, the questions these deals should put in front of you: **What proprietary data do you accumulate that a foundation model can’t get anywhere else?** If the honest answer is “nothing,” your moat is thin no matter how good the current product feels.**Does that data compound?** MaintainX gets more valuable every work order. Cognite’s graph gets denser every asset. Fin’s model gets better every resolved ticket. The best data moats improve on their own as customers use the product, with no extra engineering.**Is your growth concentrated in the AI-native line?** Fin’s whole valuation rests on the $100M piece growing 350%, not the $400M headline. If your fast line is real, separate it out and show it, because that’s what a buyer prices.**Would a legacy platform pay half your growth rate as a multiple to avoid rebuilding what you have?** That’s the exit math this cycle. Generic B2B multiples are compressing. Data-moat companies growing 40% to 50% are clearing 20x+, because the buyer’s alternative is years of building they don’t have time for. In every one of these deals, the AI-native company is the one being bought, and the multi-decade incumbent is the one doing the bolting-on. Salesforce, Autodesk, and Schneider are all excellent companies. In this cycle, they’re the buyers, not the built. The companies that owned the data, and put their growth in the AI-native line, got the check.