Bending Spoons

IThe business

Bending Spoons buys digital businesses that already have customers, rebuilds how they are run, raises what they earn, and uses the cash to buy the next one. It was founded in Copenhagen in 2013 and moved to Milan in 2015, and it has completed more than 50 acquisitions.² Its shares began trading on the Nasdaq on 1 July 2026 under BSP, at an IPO price of $29.² It is a new listing, so there is one annual report with audited numbers behind it, covering three years, and one quarterly report as a public company.

The scale changed fast. Revenue was $387M in 2023, $671M in 2024 and $1.31B in 2025, and the company guides to $2.78B to $2.82B for 2026.²³ It owns AOL, Vimeo, Eventbrite, Evernote, WeTransfer, Brightcove, Harvest, komoot, StreamYard, Remini and, since this autumn, Tractive, Airtable and Miro.³⁸ In March 2026 its products had more than 500M monthly active users and more than 9M monthly paying customers, though the company says its counts include duplicates.² Of revenue in Q1 2026, 84% was subscriptions, 12% advertising and 4% other, and 65% came from North America.² No customer was more than 1% of revenue.²

The three steps

The company describes its playbook in three steps. It acquires a business whose core products are digital, with a large revenue base it can forecast. It transforms the business, which in its words means reorganizing teams, rebuilding the technology, redesigning the product and changing marketing and pricing. Then it reinvests the earnings, plus debt, into the next deal.² It says it has never sold a material business and plans to hold what it buys.²

The transformation is where the money is made, and it is blunt. At Evernote the team shrank from 341 people at the acquisition to 60 by the end of 2024, a cut of 82%. At StreamYard it went from 154 to 44 in under a year, a cut of 71%.² The company says it removed management layers, cut cloud costs by 47% to 65% as a share of revenue, and rebuilt products. It also says Evernote revenue per active user in 2025 was 2.5 times the 2022 level.² The company chose these case studies itself, so I read them as the best examples, not the average.

The same pattern shows at the companies it bought from public markets. Vimeo had 1,102 employees at the end of 2024 and an operating income of $18.7M on revenue of $417M.⁹ Eventbrite had 636 employees at the end of 2025 and an operating loss of $26.1M on revenue of $292M.¹⁰ Brightcove had 671 employees at the end of 2023.¹¹ The group as a whole had 2,284 full-time equivalent team members at the end of Q1 2026, of whom 621 were the company’s own selected core staff, called Spooners.² I do not have the post-deal headcount of each business, which the company does not publish.

How the money moves

The enterprise value of its acquisitions was $194M in 2023, $876M in 2024 and $1.92B in 2025, then $2.01B in the first quarter of 2026 alone.² Tractive added $759M in May.³ Airtable, at $1.285B, and Miro, at $1.355B, add $2.64B more.⁴⁵ That is about $5.4B in ten months of 2026, and $8.4B over four years C (calculated).²³⁴⁵ It paid for them with debt and with equity. The IPO raised $1.10B net after the underwriters’ option was used.¹² Berkshire Hathaway bought $600M at the IPO and $330M more in October.⁶⁷ A year ago it was a private company funded by lenders and a few long-term holders. Now it has a share price, and it will be judged on it.

Bending Spoons is a buyer first and an operator second, and the shares are a claim on how well it keeps doing both at a larger size.

IIThe moat

The products it buys have no strong moat. Evernote, WeTransfer, Vimeo and Eventbrite were all established brands with plenty of rivals, and Remini and Splice are shrinking.³ If the protection were in the products, a buyer of old software would have none. The protection is the buyer’s method, and the question is whether that is durable.

The company claims three things: its people, its internal tools, and its data. It says its team received about 800,000 applications in 2025 and hired 286 people.² It says it ran 3,000 product experiments in 2025 and that more than 90% of its code changes were written or co-written by AI by the end of Q1 2026.² It says revenue per full-time equivalent core staff member was $2.57M in 2025, up from $1.12M in 2023.² That last figure fell to $0.97M in Q1 2026 as the headcount of the acquired companies came in, and the company does not give a comparable full-year number for 2026 C (calculated).²

The evidence for the method is margin. Adjusted Operating Income margin went from 36% in 2023 to 47% in 2025, and 54% in Q2 2026, while the company bought bigger and bigger businesses.²³ It is hard to get that from luck. But it is also a number the company defines, and I show below how much it leaves out.

I give the method some credit, but I put its value in what it earns on money spent, not in the products. A competitor can buy the same product. It cannot easily copy a team that has restructured fifty companies.

IIIThe competitive advantage

The company competes at two ends. At the product end it faces AI tools that do what Evernote, Vimeo and Miro do. At the deal end it competes with private equity funds and strategic buyers for the same assets. The advantage is at the deal end. It wins by being able to pay a fair price and still earn well, because it expects to cut costs further than a financial buyer would, and because it plans to hold forever, which lets it ignore exit value.²

Look at what it paid. Airtable was $1.285B of enterprise value for about $480M of annual recurring revenue growing more than 20%, or 2.7 times.⁴ Miro was $1.355B for about $600M, or 2.3 times.⁵ Both come with large cash balances, which is why the equity values are $2.25B and $1.79B. Vimeo made $18.7M of operating income the year before it was bought.⁹ If the company can run these at its own margin, around half of revenue before intangible amortization, the price is about five times the profit it expects to make. That is how I get the 5.5 times I use in the valuation E (estimate).

Who else does this

Private equity funds buy software at similar prices and cut costs too. They also use debt, but they sell within five to seven years, and they face a clock. Bending Spoons faces none, says it never sells, and is now public. That is a different cost of capital, and it matters. A buyout fund wants 20% a year over five years. Bending Spoons reports an internal hurdle of 25% unlevered and 65% levered on its deals since 2023, and I read these as targets, not results.²

On this site, Berkshire Hathaway is the nearest comparison for the idea of buying a business and keeping it. Berkshire’s own page has bear, base and bull values of $394, $467 and $561 against a price of $542.19, so the market pays a premium for the owner there as well. Prosus is the other holding company on this site, with a base value of $49 against a price of $38.18.

What Bending Spoons has that those two do not is growth: revenue rose 84% a year from 2023 to 2025. What it lacks is a long record of public accounts, and a long record of holding old software through the AI shift E (estimate). The first point supports the price. The second argues against paying for it before the proof arrives.

The edge is in the price paid and the costs removed. It lasts as long as sellers keep accepting those prices and as long as debt stays open.

IVManagement

Four founders run it and control it. Luca Ferrari is chief executive and chair. Francesco Patarnello leads acquisitions. Matteo Danieli sits on the board. Luca Querella holds shares but is not on the board or in the officer list.² Together they own all 310.2M class A shares, which carry five votes each. That is 47% of the shares and about 82.5% of the votes C (calculated).²⁷ A class A share turns into an ordinary share if it passes to anyone else, if its owner dies, or if the owner falls below 37.5M class A shares.² So control is personal and ends with each founder, but it will last as long as they do.

How they are paid

The founders are paid very little in cash. In 2025 all executive officers and directors together received $10.8M, of which $2.2M was cash.² Ferrari and Patarnello have no employment agreements.² Their wealth is in the shares. At $45.52, Ferrari’s class A stake is worth about $3.7B C (calculated).¹² That pays them to make the share price go up, and it also pays them to keep control, since the stock gives them 82% of the votes with 47% of the shares. A private-equity style carry would pay them for exits. They have none, and say they never sell a material business.

The wider staff are paid in stock options. As of 15 June there were 36.3M options outstanding, at an average price of $0.28, which is 5.5% of the shares, plus a new plan that reserves 51M shares more.² At that strike price, the options are the same as shares. I count them in full.

What they did with capital

The company has lent the founders money. In 2024 Ferrari, Patarnello and Danieli each received a two-day loan of about €1.1M to cover tax due on shares they sold, and each loan was repaid.² I mention it because a public company cannot do that, and the rules will now make it ask. The IPO did not sell founder shares: the 23.6M shares offered by holders came from Galileo, Baillie Gifford, StarTIP and others.²

Over a full cycle, the record is of growth bought with debt and shares. Lenders have kept funding it, and Moody’s and S&P raised its ratings in September.¹³ But the cycle has been one of cheap money and rising markets. I do not know how it behaves when a large deal goes wrong, because none has yet.

New shares have cost existing holders. Between the IPO, the Miro holders’ reinvestment and Berkshire’s placement, the share count rose about 9.3% in four months C (calculated).¹²⁸⁷ Each was reasonable on its own terms. Together they say that equity is a tool the founders will use when it is cheap, and a share price of $45.52 makes it cheap for them.

What worries me about control

Governance is thinner than the control suggests. The company relies on home-country exemptions from some Nasdaq rules as a foreign private issuer, and it told investors it has material weaknesses in internal control over financial reporting.² One director, Leah Schwartz, is a managing director at Allen & Company, an underwriter of the IPO, which the filing flags as a conflict.² Most of the old shareholders can sell in late December 2026, when the 180-day lock-up ends on up to 318.5M ordinary shares C (calculated).²

The managers have more of their own money in this than almost anyone I follow, and they have votes that let them be wrong for a long time.

VFinancials

There are only three audited years, 2023 to 2025, so I cannot show ten. The first financial statements in the prospectus start in 2023, and the company has reported one quarter and six months since. I show those, and I use the pro forma statements in the prospectus to see the three big 2025 and 2026 purchases as if they were owned for a full year.

Revenue grew from $387M to $1,306M in two years, and net income barely moved: $161M in 2023, which included a $65M tax benefit, then $89M and a $0.2M loss in 2025.² Operating income was $278M in 2025, 21% of revenue. The company’s own measure, Adjusted Operating Income, was $613M, or 47%. The difference is $335M, of which $151M is amortization of what it paid above book value for the businesses, $79M is reorganization pay-offs, $85M is deal-related expense and $21M is other items.² I think the first is a fair add-back. The other three are not. They have been present in every year, and they grew with the company: reorganization expense was $14M in 2023 and $79M in 2025.

The pro forma statement in the prospectus adds Eventbrite, AOL and Vimeo to 2025 as if they were owned all year. That gives revenue of $2,608M and an operating income of $509M, after the company’s own acquisition adjustments.² The pro forma interest expense is $368M, so about $152M of income before tax remained, and net income was $22M C (calculated).² That is what a full year of the 2025 portfolio looks like with today’s debt, and it explains why the shares trade on adjusted measures.

Debt

The debt is the other number to look at. At 30 June the principal was $4.98B, with a weighted average rate of about 7.7% on the term loans C (calculated), and cash of $793M. Net debt was $4.09B and the leverage ratio, the company’s own measure, was 2.4 times.³¹² After the quarter it sold $1.1B of stock, borrowed $1.25B and €395M more as term loan B add-ons, and priced $1.25B of 9.625% dollar notes and €500M of 8.625% euro notes due 2032 for settlement on or about 4 November.³¹³¹⁵ I estimate net debt at about $5.0B after the cash spent on Airtable and Miro E (estimate). The bridge is in the valuation.

The owner cash table is the most important one on the page. Over the twelve months to June, Adjusted Operating Income was $1.05B. After the cash costs of reorganizing, the deal costs, interest and everything else, operating cash flow was $457M. After counting stock pay as a cost, the owner’s share is $382M, or 36 cents on each dollar of the company’s favourite measure C (calculated).²³ The TTM covers part-year ownership of the largest deals while carrying interest on all the debt raised for them, so it is lower than a normal year. My 2027 number, with every deal owned for twelve months, is $1.06B E (estimate), and I explain that in the next section.

What I take from the numbers

The margin gains are real: the group is more profitable than the sum of what it bought, and cash flow rose from $59M in 2023 to $290M in 2025.² The rest is cost. There is a regular drag, of reorganization and deal expense, that reported earnings show and the adjusted figure removes, and I treat it as a running cost of the growth strategy. I do not know what the margin would be if it stopped buying, because it has not stopped.

The company earns a lot more than it keeps today. The gap is interest, reorganizing, and deals, and it closes only if the buying slows or the debt falls.

VIValuation

How this goes wrong

Before any numbers, I ask what would most likely hurt an owner. I see three ways. The first is the credit window. The company funds deals with debt at 8% to 9.6%, and its latest secured notes pay 9.625% in dollars.¹⁵ If lenders pull back or rates rise, it cannot buy, and the rest of the structure carries $8.3B of gross debt by my count E (estimate). The second is the products. It reports net revenue retention of 94%, so the existing base loses about 6% a year before new customers, and organic growth was 3% in Q2.²³ If AI tools take customers faster from Evernote, Vimeo, Miro and the like, cost cuts cannot save the revenue. The third is the people at the top. Four founders control the votes. If a large deal goes wrong, no one outside the four can force a change.

What an owner can take out

The price is set on Adjusted Operating Income, so I start there and subtract what the company does not. In the last twelve months the owner’s cash was about $382M against $1,055M of the adjusted number C (calculated).²³ The gap is reorganization pay, deal costs, interest and stock pay. I treat the first three as costs of the strategy, because the strategy needs them every year. For capital needed to stand still, the software needs almost nothing, with capital spending of $4M in a year.³ But the strategy needs a lot to grow. In my model each dollar of enterprise value bought carries 8 cents of reorganization and deal costs, a ratio close to the 2025 and 2026 record E (estimate).²

What I cannot judge

Four inputs are unknowable from here. I do not know how Airtable and Miro earn today. The company gives their annual recurring revenue, $480M and $600M, but no profit, and the pro forma statements stop before them. I do not know whether the cuts that raised margins elsewhere will work on companies that sell to large customers through sales teams. I do not know the cost of the next debt it raises. And I cannot measure the stock’s sensitivity to markets: over 86 trading days its daily beta against the S&P 500 was 0.28 and its weekly beta was 2.11, with annualized volatility of 114%, which is noise C (calculated).¹⁷ I show ranges in place of false precision. My discount rate is a judgment.

The discount rate

I use 11.0%. The 10-year Treasury yielded 4.94% on 30 October, and Damodaran’s implied equity risk premium was 4.20% on 1 October.¹⁸¹⁹ A beta of 1.45 gives 11.0%. I chose the beta for a company with net debt near 2.5 times earnings that depends on acquisitions, and it can’t be tested against a record, because the record is 86 days. As a check, the company’s own secured notes pay 9.625%, so equity must earn more than that. Spotify and others on this site use 10.0%. I add a point for the leverage, the dependence on new deals and the missing history E (estimate).

The three cases

My base case starts from the portfolio the company owns today: the businesses in its Q3 guide, which implies about $3.0B of annual revenue and $1.56B of Adjusted Operating Income, plus Airtable and Miro at the $1.08B of recurring revenue the company reported.³⁴⁵ Existing businesses grow 3% a year, then 2%, and keep a 52% to 53% margin. Airtable and Miro start at a 30% margin, which I assume, and reach 48% by 2030. Then I let the company buy $2.5B to $3.0B of businesses a year, at 5.5 times the profit it expects them to make, paying 8% on top for reorganization and fees. Net debt is held to 3.0 times Adjusted Operating Income. Cash beyond deals is paid out to owners. After 2036 the business grows 2.5% a year E (estimate).

The bear case has customers leaving faster than the cuts can offset: existing revenue falls 1% to 2% a year and margins drop to 44%, deals shrink to $0.8B to $1.2B a year at 7.5 times, and debt costs 9%. The bull case has organic growth of 3% to 5%, margins of 57% in the existing group and 55% in the new, and deals rising to $7B a year at 5.0 times. Bull reaches $13.6B of Adjusted Operating Income in 2036, which is 6.8 times 2027 E (estimate). The weighted value, at 25%, 50% and 25%, is $39.

Base value is $36.88, which I round to $37. 93% of it is the value after 2036, because the company pays out little before then. That is a feature of a compounder, and it is also why the discount rate matters so much. Owners receive $6.98 a share in cash by 2036 in the base case E (estimate).

What the price assumes

At $45.52 the market value is $31.5B and the enterprise value is $36.5B, about 19 times my 2027 Adjusted Operating Income C (calculated). My base cash flows give an 11% return only at $37. At the price they give 10.0% a year. If the company stopped buying today and paid everything out, I would value it at $25 a share. So about 46% of the price is a payment for deals not yet done.

Volume alone cannot close the gap. In my model, spending more a year than $6B stops adding much, because the debt limit binds and the company can only buy what its cash flow and borrowing allow. What closes it is the price paid per dollar of profit. At 4.4 times, not 5.5, the base reaches $45.52. Or the discount rate: at 10% it does too.

What closes the gap to $45.52

Each of these closes the gap alone, with all else at my base. Pay 4.4 times for deals: the company paid about 2.3 to 2.7 times annual recurring revenue for Airtable and Miro, and I see that as about 5 to 6 times the adjusted profit it hopes to earn there. Or accept a 10% return, which is a point under my rate and equal to the 9.6% secured notes plus half a point. Or raise the existing businesses’ margin by 21 points, to 73%, which I do not believe E (estimate). The last shows that margin growth in the old businesses is not the thing the price relies on. The price relies on the machine.

Quality and price

These are two questions. As a business the machine is unusual: it has bought and rebuilt more than fifty companies and the margins say it works. As a purchase at $45.52 it asks for more than I can give it. My base is $37, the price is 23% above it, and the margin-of-safety price is $16, which is my bear case divided by 0.75, so that the price has room for the bear case to be right. The other things I could do with the money are in this site’s pages: Berkshire Hathaway’s page has a base of $467 against a price of $542.19, so it too sells above my value, and Prosus has a base of $49 against $38.18. A 10-year Treasury pays 4.94% with no risk to the value.¹⁸

The case needs many things to go right together: lenders keep lending, sellers keep selling at about five times, cuts keep working without losing customers, AI does not erode the old products, and the founders stay disciplined with 83% of the votes. If I give each an 80% chance and treat them as independent, which they are not, the chance of all five is about one in three. The price does not need all of them at full strength, but it needs most.

Where this could be in 5 and 10 years

The base value reaches $62 at the end of 2031 and $97 at the end of 2036. This includes about $13B of deals in the first five years and $28B in ten, at 5.5 times, with the existing businesses growing 2% to 3% and net debt kept near 3 times Adjusted Operating Income at first. For the base to hold at five years the company must have turned Airtable and Miro into a 46% to 48% margin, with Adjusted Operating Income near $4.1B by 2031. For it to hold at ten, it must still find $3B of deals a year at the same price. The value grows about 9% a year between those dates; the rest of the 11% rate comes back as cash paid out. I have only three audited years behind this, so the ten-year path rests on my assumptions, not on a record.

The bear differs on the price paid and the deal count, not just on margins. The bull differs on how much it can buy. The numbers that drive it are the price paid per dollar of profit, the margin on the new businesses, the yearly deal budget and the 11% rate. The most likely way the base is wrong is that deals dry up or get dearer as the company grows. Cash flow then piles up at a lower return than I assume, and the value drifts toward the no-deal figure of $25.

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

The deals are cheaper than they look. Airtable’s equity price was $2.25B but its enterprise value was $1.285B, because it had about $965M of cash, and Miro’s was $1.79B against $1.355B for the same reason C (calculated).⁴⁵ The company paid for the cash and got it back. At 2.7 and 2.3 times annual recurring revenue, that is the lowest price for the biggest businesses it has bought.

The second fact is Berkshire’s price. It bought 10M shares at $33 on a contract signed in October, when the stock had closed at $37.86 the day before and then rose to $53.46 on the announcement.⁶¹ Its two purchases came to $930M for about 30.7M shares, an average of $30.30 if the first was at the $29 IPO price, against a close of $45.52 C (calculated).⁶⁷¹

The bull case

The company’s own argument is that it faces more than 1,000 possible targets with about $400B of revenue, that its internal return targets are 25% unlevered, and that AI lets it do more with fewer people.² Analyst price targets exist, but I could only find them in secondhand reports that I could not check at the source, so I do not use them. At the IPO, Baillie Gifford held 15% of ordinary shares before selling, Durable Capital 6.7% and Cox Investment Holdings 9.3%, as of 29 May.² I found no public view on Bending Spoons from Brad Gerstner, Gavin Baker, Cathie Wood, Bill Gurley, Chamath Palihapitiya, David Sacks or Bill Ackman.

Against reality

If the company keeps paying under 4.4 times the profit it expects, and finds $3B of deals a year, the price is possible. If it pays 5.5 times, as I assume, the price is about 23% too high. Today the evidence for 4.4 times is in two deals that have not yet shown their margins. The evidence for $3B a year is a record of $5.4B in ten months, which came with a record IPO and a year of debt raising. Neither says what year five will look like.

What a 10x would take. A tenfold share price is $455. On 693M diluted shares that is a market value of $315B, and about $338B of enterprise value with the net debt in my bull case. At the 19 times the market pays for my 2027 Adjusted Operating Income, it needs about $18B in 2036, or roughly $32B of revenue at a 55% margin. My bull case reaches $13.6B. So a tenfold move needs more than the bull case, bought at the prices it assumes, and a multiple that stays where it is. That is revenue growth of about 26% a year from my 2027 base for nine years. Very few companies that were already worth tens of billions have done that, and Nvidia on this site is the nearest example. I treat it as not plausible as a planning number.

Where I come down

The acquisition prices and the margins moved me. The size of what Berkshire paid did not, since it paid $33 and the shares are $45.52, and one investor’s purchase is not a value. What would raise my values is evidence that Airtable and Miro reach 45% margins within a year at the prices paid, a second year of stable organic growth, and a lower cost of debt. My values stay at $12, $37 and $68.

VIIIRisks, and what would change my view

The risks I weigh most are covered in the valuation: credit, customer erosion and control. Others matter too. The company reports material weaknesses in its financial controls, which it is fixing, and an error in its numbers would hurt a stock held up by an adjusted measure.² About a quarter of revenue goes through the Apple and Google stores at fees of 15% to 30%.² Cloud costs run mainly with AWS and Google Cloud.² Of the staff, 40% is in the US, 21% in Italy and 10% in India.² Dollars earn most revenue but the debt is partly in euros C (calculated).¹²

A change in my view would come from either direction. A deal that misses badly, or a quarter of negative organic growth, would lower it. Evidence that large, sales-led products can be rebuilt as well as the small ones would raise it.

Geopolitical risk, and what it means for the value

The company sits between two systems. It is Italian, listed in New York, and gets 49% of its revenue from the US, 7% from the UK and under 2% from Italy.² Its main exposures are legal and financial, not physical. It depends on no fab, no single chokepoint and no sanctioned supplier that I can find in the filings, and it names export-control and sanctions law as a compliance risk.²

The one in force is CFIUS. The company notified the US government of the AOL purchase in March 2026, after closing in January, and the review is open.² The US Treasury has also asked about Eventbrite. In 2024 CFIUS imposed conditions on about one in eight notices.²⁰ Conditions would cost me a point or two of margin on AOL. A forced sale is possible in law and I rate it very unlikely. Italy’s Golden Power law gives the government a say over deals and over lenders enforcing share pledges in strategic sectors, which can delay a financing. The EU-US data framework that lets European user data go to US servers is under appeal at the EU’s top court, and a loss would cost compliance work across the group.²¹

The largest channel is indirect. A war or shock in Europe or Taiwan would not touch its servers. It would shut or reprice the credit it needs for deals. Its notes pay 9.6% already. The table reruns my model for each scenario. The price already carries some of this, since the notes are priced at spreads that assume an open market. It carries none of the CFIUS or data-framework risk that I can see.