The Disciplines of M&A
The first AI-era acquirer with audited public accounts: the record, the production process, the economics as filed, and the two places the evidence is contested.
For a decade the argument about AI-era acquirers ran on private numbers and management claims. On July 1, 2026, it stopped: Bending Spoons listed on Nasdaq, and the most-discussed acquisition model of the cycle acquired audited accounts. That is what earns it a case study on this shelf. Not because the model is admirable or replicable — both are open questions — but because for the first time the claims can be checked, and checking them teaches more about value creation in the AI age than any thesis deck.
The model itself takes one sentence. Buy mature digital revenue at a depressed price, replace the acquired cost base with one shared operating platform, reprice the installed base, and never sell.
Milan, founded 2013, a consumer-app developer that spent its first decade building a subscription and monetization engine for its own products. The acquisition program that made it consequential starts in 2023 and has not paused since: Evernote (closed January 2023, undisclosed), Meetup (January 2024, undisclosed), WeTransfer (July 2024, undisclosed — a business that had targeted roughly €716M in its abandoned 2022 IPO), StreamYard and Issuu, Brightcove ($233M all-cash, February 2025), komoot (March 2025, undisclosed), Vimeo ($1.38B all-cash at $7.85 a share, a 91% premium to the 60-day volume-weighted average, November 2025), AOL from Yahoo (roughly $1.4–1.5B, January 2026, on a $2.8B term loan and revolver package from a twelve-bank group), Eventbrite (roughly $500M at $4.50 a share, March 2026), Miro ($1.355B enterprise value, all-cash, announced September 2026), Airtable ($1.285B enterprise value, about $2.25B of equity value including the target's net cash, closed September 2026).
Airtable is the one to sit with. Venture investors last priced it at $11B for the growth story. Bending Spoons paid about 2.7× ARR — roughly $480M of it, still growing north of 20% — for the revenue. Same company. The distance between those two prices is the entry-multiple arbitrage, expressed in a single line item, and it says as much about what venture paid as about what the acquirer got.
The capital kept pace with the record: a ~$700M private round at $11B in late 2025, then the IPO at $29 a share — $1.68B raised, roughly $18.4B on the screen. Forbes put the sharpest frame on it: the AOL, Vimeo and Eventbrite purchases cost about $3.3B and came attached to businesses earning pro forma 2025 net profit of $22M. That is not an embarrassment for the model. That is the model. The profit is not bought; it is manufactured after closing.
Call it what it is. Across a dozen acquisitions the sequence has been consistent enough to be a production process, not a strategy.
Within months of closing, 70–90% of the acquired headcount goes — Evernote from 341 people to about 60, WeTransfer down roughly 75%, komoot reported near 85%. The product migrates onto the group's central platform: one subscription engine, one user-acquisition machine, one data and machine-learning stack, serving every property in the portfolio. Then price moves against the base that stayed. Evernote's personal plan had already gone from roughly $37 to $80; it then went to $130, another 63%. Vimeo users saw 34–180% increases with forced plan migration inside a few months. Nothing has ever been divested.
The design choice underneath is the interesting part. Constellation Software, the reference compounder of the prior era, runs hundreds of autonomous units and treats decentralization as the moat. Bending Spoons is its deliberate negation — one platform, no autonomy, every acquired product rebuilt on shared rails — and the shared rails are where the AI leverage lives. AI-authored pull requests went from under 10% of the company's code in early 2025 to over 90% by the first quarter of 2026. Revenue per employee went from $1.12M in 2023 to $2.57M in 2025. The strongest public software companies do not get near that second number.
Revenue of $1.31B in 2025, up 95%. First-quarter 2026 revenue of $601M, up 132%. Almost all of that growth is acquired rather than organic — which is not a flaw in the model, it is the model's definition. Management guides to roughly $1.4B of adjusted EBITDA for 2026 and presents margins near 50%.
Now read it from the credit seat, because the balance sheet is the half of this story the admirers skip. Roughly $4.4B of debt against $1.06B of equity at March 2026. Interest expense of $142.6M in 2025. About 80% of first-quarter 2026 acquisition spend was debt-funded, against a stated 25%-plus return hurdle. Strip the technology story away and a lender is looking at a leveraged serial acquirer running four turns of debt against earnings that are still being argued about — priced, correctly, as exactly that.
Retention. The arithmetic of the whole model rests on repriced users staying, and management calls retention on repriced bases remarkably stable. Independent teardowns of Evernote's subscriber behavior say otherwise. The company discloses group-level rather than cohort-level figures, so the dispute cannot be settled from the filings — which is itself worth noting. A 63% price increase against a base that holds is a pricing-power business. The same increase against a base that quietly halves over three years is a harvest. From outside, the two look identical for about two years.
The earnings. Adjusted EBITDA add-backs ran $335M in 2025 — more than the $278M of reported operating income. Restructuring is treated as one-time, but in a company whose production process is restructuring, the cost recurs by design. S&P models 2026 margins near 30% against the roughly 50% presented. Both descriptions are internally consistent. They disagree about one thing only: whether the cost of running the playbook belongs inside the playbook's margins. A buyer of the stock is taking a side in that argument whether they know it or not.
Three levers did the work here: entry-multiple arbitrage on neglected assets, cost-base replacement through the shared platform, and repricing power over captive consumer bases — users with high switching friction and no procurement department. The first two travel. The third mostly does not. An enterprise customer employs a person whose job is to refuse the Evernote treatment, in a market where the largest platforms bundle adjacent capability at zero marginal price. Anyone carrying this playbook into enterprise software or services is running a two-lever version of a three-lever trade, and should underwrite the returns that way rather than borrowing the consumer math.
The second lesson is the one the market has already voted on. Through 2025 and 2026 it marked the decentralized compounders down and funded the centralized-platform acquirers up — the clearest evidence available that autonomy versus platform has stopped being a culture preference and become the load-bearing design decision in serial acquisition.
And the caution, which the shelf exists to record: this is the best-evidenced version of the AI-first acquirer there is — listed currency, audited accounts, a cost structure nothing else matches — and it still carries four turns of leverage against contested earnings and a retention question the disclosures cannot answer. The category is real. The benchmark is real. Proven is a different word, and it has not been earned yet.
Sources: Bending Spoons F-1 and IPO filings (2026); company and target announcements for Brightcove (2024), Vimeo (2025), AOL (2025), Eventbrite (2025), Miro (2026) and Airtable (2026); Forbes (2026) on aggregate outlay versus pro forma profit; TechCrunch (2024) on the WeTransfer reduction; S&P commentary on margin modeling (2026). Values as announced or filed; ranges where sources conflict.