Bending Spoons

A Milan operator that most enterprise buyers had never heard of in 2024 now owns software that more than 750,000 organisations use. It buys mature application software at two to three times revenue, runs it with a much smaller team, and never sells. The prices it pays are the clearest available measure of what the market now thinks an installed base is worth.

Section one

What does Bending Spoons actually buy

It buys finished application software with a long-tenured customer base, and it buys all of it. Every transaction in the record below is a full acquisition rather than a stake, and the company states in its prospectus dated 30 June 2026 that it has never sold a material business. Across the portfolio as of the first quarter of 2026, the revenue-weighted average subscriber tenure was 8.0 years and 48 per cent of subscription revenue came from customers of five years or more.

The layers it holds are consumer internet, media and video, and work software. There is no position in models, compute or data tooling in any of the eighteen acquisitions with a sourced completion date between June 2021 and 11 September 2026.

Scroll the table sideways to see every column.

Source: Bending Spoons prospectus, Form F-1, 30 June 2026, acquisition track record; Q2 2026 results release, 13 August 2026; Bending Spoons deal announcements for Brightcove, Vimeo, Eventbrite, Airtable and Miro; Wall Street Journal reporting via AOL.com for the AOL price, which the company has not disclosed. Excludes the unnamed assets acquired from IAC Inc. in 2024 and Splice, whose acquisition dates are [to confirm]. Miro was agreed on 10 September 2026 and had not completed as of 11 September 2026.

Scroll the table sideways to see every column.

A dot is absent. Take-private means a listed company delisted on completion. Carve-out means a business bought from a corporate parent. Source: as the ledger above, Global AI Forum classification, 11 September 2026.

Section two

Who funds it, and how much is left to spend

There is no fund and no fund life. Bending Spoons buys from its own balance sheet, which it refills from operating earnings, bank debt, private equity rounds and, since 1 July 2026, the public market. It sold 57,971,015 shares at 29 dollars in its Nasdaq listing on 1 July 2026, raising 1.68 billion dollars gross, of which about 954 million dollars before expenses went to the company.

The prospectus dated 30 June 2026 discloses the rule the company applies when it sizes a transaction: acquisition debt is modelled at the lower of 85 per cent of enterprise value, or the amount the target's projected free cash flow can repay within five years. It applied internal rate of return hurdles of 65 per cent levered and 25 per cent unlevered to nearly every acquisition completed from 2023 through the first quarter of 2026.

Source: Bending Spoons Q2 2026 results release, 13 August 2026, balance sheet and income statement; prospectus, Form F-1, 30 June 2026, share capital.

Source: Bending Spoons Q2 2026 results release, 13 August 2026; deal announcements for Airtable, 4 September 2026, and Miro, 10 September 2026. Global AI Forum calculation, 11 September 2026, using enterprise value as a proxy for the cash outlay because each target's own net cash is acquired with it.

Section three

Why a profitable company sells at ten cents on its own dollar

Because the last valuation stopped being reachable by any other route. Miro was reported as profitable with about 435 million dollars of net cash at the time of the agreement of 10 September 2026, and Airtable's equity value of about 2.25 billion dollars implies about 965 million dollars of net cash. Neither sale was a rescue. What had gone was the path back to a 2021 mark, after the software sector's forward price to earnings multiple fell from 84.1 times at the 2020 to 2022 peak to 22.7 times in the first quarter of 2026.

On revenue the prices are consistent and low. Against the revenue base named in each announcement, Eventbrite cost about 1.6 times its 2024 revenue of 325.1 million dollars, Miro about 2.3 times annual recurring revenue of about 600 million dollars, Airtable about 2.7 times annual recurring revenue of about 480 million dollars in June 2026, and AOL about 3.0 times reported revenue above 500 million dollars. The Vimeo multiple of about 3.3 times 2024 revenue rests on a revenue figure that is [to confirm] against a primary source.

Section four

What happens to a business in the first four quarters

The pattern is consistent enough to plan against. The organisation gets much smaller, the product is moved onto the acquirer's own platform, packaging and prices are redesigned, and the roadmap narrows. Evernote went from 341 staff at acquisition in January 2023 to about 60, and StreamYard from 154 at acquisition in April 2024 to 44, both as read from the prospectus dated 30 June 2026. About 75 per cent of more than 350 WeTransfer staff were cut after the July 2024 acquisition, as reported by TechCrunch in September 2024.

Some measures move the other way, and all of them are the company's own. Bending Spoons reported in its Q2 2026 results release of 13 August 2026 that system stability incidents at Vimeo fell by about 90 per cent against pre-acquisition levels and that uploads and search were at least 30 per cent faster. It reported Evernote net revenue retention of 99 per cent, the highest in the portfolio, on the average from the first quarter of 2023 to the first quarter of 2026. No independent measurement supports any of these.

Section five

Which layer the portfolio does not hold

The operating advantage the company describes is increasingly an AI advantage. The share of its code changes authored or co-authored by AI rose from under 10 per cent in the first quarter of 2025 to more than 90 per cent by the end of the first quarter of 2026, with about 70 per cent written by AI alone. Revenue per full-time employee recruited through its own process rose from 1.12 million dollars in 2023 to 2.57 million dollars in 2025.

All of that runs on models the company buys or hosts rather than owns. The prospectus dated 30 June 2026 names Anthropic, Google and OpenAI as providers, alongside open-source models and narrow models built in-house. The productivity gain that justifies the prices is therefore rented, and the portfolio holds no position in the layer that sets the rent.

The AI Scissor

The condition in which AI lowers the price a mature software business can fetch from its own market faster than it lowers that software's usefulness, so that value moves to whoever can operate the software with an AI-era cost base. Defined by Global AI Forum, September 2026.

The stranded mark

The condition of a company whose last priced valuation sits above any value its own market will now pay, so that the preference stack rather than the product decides the exit. A company can be profitable and cash-rich and still be stranded. Defined by Global AI Forum, September 2026.

Section six

What an enterprise should secure before its next renewal

Ownership is now a category of vendor risk rather than a footnote in a procurement file. Miro states that it is used by more than 250,000 organisations and that more than 750 customers pay over 100,000 dollars a year. Airtable states that it is used by more than 500,000 organisations, including 80 per cent of the Fortune 100. For those buyers the window is now, before the Miro transaction completes in the fourth quarter of 2026 and before the first renewal after each reset.

• Establish which of your Airtable bases and Miro boards are business-critical, who owns each one, and what they connect to. On Airtable the dependency usually sits in what your own teams built, not in the licence.
• Review renewal dates, change-of-control clauses and any cap on price increases, and consider extending on current terms before plans are redesigned.
• Obtain written confirmation of how your content may be used to train or improve models, in language that survives a later change to the vendor's standard terms.
• Test a full export of critical data and logic, and cost one credible alternative per critical workflow.
• Ask which committed features and integrations survive the transition. A vendor in the middle of a reorganisation generally cannot answer this, and an unanswerable question is itself the finding.

Section seven

Which software companies are already on the list

The company published its own screen. Its prospectus dated 30 June 2026 reports a study identifying 1,031 digital businesses it expects could be attractive acquisition targets over the next few years, with nearly 400 billion dollars of combined estimated 2025 revenue. Of those, 791 are headquartered in North America with 315 billion dollars of revenue and 240 are in Europe with 77 billion dollars.

Three criteria are stated as filters: estimated revenue between 50 million and 5 billion dollars, headquarters in Europe or North America, and a self-serve subscription, sales-led subscription or advertising model rather than a services model. A fourth is stated as a preference and does most of the work: a trajectory the company can forecast several years out with the risk of disruption from advances in AI factored in. Applied in 2026, that preference excludes the model layer.

Section eight

What breaks first

Three things are open, and each is measurable. The playbook has never been run at enterprise scale: Brightcove, with about 1,700 paying customers as of March 2026, is the only enterprise-led position before this year, and Airtable and Miro together will be close to 28 per cent of pro forma revenue on a Global AI Forum estimate using combined annual recurring revenue of about 1.08 billion dollars against 2026 revenue guidance of 2.78 to 2.82 billion dollars.

Growth is the second. Bending Spoons reported organic growth of 3 per cent in the second quarter of 2026 against 6 per cent in the first quarter, and the shares fell about 18 per cent on results that beat forecasts. A model that must keep buying in order to grow carries its cost of capital inside its product.

Leverage is the third. Net debt of 4.09 billion dollars as of 30 June 2026 sits at 2.4 times an EBITDA measure that includes savings not yet achieved, and 794 million dollars of that debt falls due within twelve months.

Global AI Forum · GAIF-CP-003 · v1.0 Research current as of 11 September 2026 gaiforum.com