Bending Spoons,
read like an analyst
One study session, worked through question by question: what the company actually is, how its acquisition machine runs, and how to read the numbers in its Nasdaq prospectus without being sold a story.
How to read this page
- Definition a term explained plainly the first time it matters.
- Red flag a place where a smart reader slows down and gets suspicious.
- Highlight a sentence worth memorising — the kind that changes how you read every other number.
What is Bending Spoons' economic engine?
I want to understand how this company — the one that bought AOL and Vimeo — actually runs financially.
Bending Spoons is a Milan-based conglomerate that went public on the Nasdaq with a pop, briefly topping a $25B market cap — roughly double its last private valuation of $11B. It reported $1.31B revenue in 2025, and its portfolio served over 500M monthly active users and more than 9M paying customers as of March 2026.
The model resembles private equity with one twist: it holds forever and has never sold an acquired business. It buys established digital brands — Meetup, Eventbrite, WeTransfer, Evernote, Vimeo, AOL, Issuu and others — and makes them more profitable through technology and AI, but also through price hikes and layoffs.
The standout is efficiency: revenue per core employee ("Spooner") rose from $1.12M (2023) to $2.57M (2025). The core team is only ~620 people; in 2025 it hired 286 out of ~800,000 applicants.
Most of these figures come from the company's own IPO filing. The article gave revenue but not net profit or debt — and those are exactly what you need to judge whether the engine is healthy.
Is it building a platform to rival TikTok?
"Hold forever" — so the goal is a complete system? Is the platform meant to compete with TikTok? What do the acquired companies actually do?
Almost certainly not TikTok. TikTok is one unified consumer platform bound by a recommendation algorithm. Bending Spoons' products — notes, ticketing, file transfer, video hosting, a portal — are unrelated, and it has no intention of fusing them into one consumer app.
The "complete system" it wants is an internal operating machine. Its own filing says the founders decided "to treat the operating machine itself as our most important product." It's a centralised back office — engineering, billing, marketing, support, pricing, analytics — onto which each acquired product is migrated.
The right comparison is Constellation Software or a "Berkshire Hathaway of bits," not TikTok. One critic notes it does not build a super-platform; it holds unrelated assets that share only the operating machine.
What each acquired company does
- Meetup — organising interest-group events (in person / online)
- Eventbrite — event ticketing & registration
- WeTransfer — large-file transfer
- Evernote — note-taking / knowledge management
- Vimeo — video hosting (creator/pro-focused, ad-free)
- AOL — legacy internet portal: email, news/media
- Issuu — digital publishing (turns PDFs into online magazines)
Plus Remini (AI photo), Brightcove (enterprise video), komoot (route planning), Harvest (time tracking), StreamYard (live streaming) and more — around 50 deals in total.
Isn't sourcing deals a VC's job? How do they get data?
Screening 2,500 targets, analysing 200, closing 6, tracking 1,000 more — isn't that venture capital's work? Do they build their own analysis platform? How do they get first-hand data and do market research?
Your instinct is half right: sourcing and diligence look like an investment firm. But the model is a PE-style serial acquirer, not VC.
VC buys minority stakes, bets on early growth, and exits for returns. Bending Spoons buys outright, operates forever, never sells. Early on, VCs literally didn't get it — the founders say they got a lot of "you're crazy" reactions.
How they price and research
- Cash-flow modelling. Because they buy-and-hold, they model future cash flows to decide what they can pay — and they've been the highest bidder on every deal they won. Purchase prices run ~1–3× revenue.
- First-party data comes after the deal. They target products with real customer behaviour; once owned, they enter a weeks-to-months "learning phase" using actual usage data to decide what to change and how to price.
- Experimentation, not surveys. They use "sophisticated data tracking, analytics infrastructure and an experimentation toolkit" — essentially large-scale A/B testing — for pricing and feature calls.
What powers the efficiency — whose algorithms?
The core is running companies efficiently. Which mature sub-platform algorithms or models do they rely on?
Honest answer: the specific algorithms aren't disclosed — the operating machine is their prized secret. But per the F-1, execution rests on three pillars: people, proprietary technology, and proprietary data.
- Talent density — ~620 Spooners, hired at hedge-fund-competitive pay from a brutal funnel.
- A centralised "universal OS" — every acquisition migrates onto one Milan back office, cutting large amounts of legacy overhead.
- Deep rewrites + AI — e.g. Evernote's backend was moved from a monolith to microservices; Remini was rewritten line by line.
The filing confirms they build on top of third-party models from Anthropic, Google and OpenAI (via API or self-hosted), plus open-source models, and add their own narrow-purpose models. So "fine-tuning on others' AI" was partly right — they don't train a large model from scratch.
The "Berkshire of bits," static companies, and the data question
Explain "Berkshire of bits." Are these all static, non-growing companies? Aren't they really buying data to analyse user behaviour? And do Google/Microsoft's data not overlap with these small firms'?
Berkshire of bits
Berkshire Hathaway is Buffett's holding company: it uses cash flow (historically insurance "float") to buy good businesses and hold them almost forever. (Buffett stepped down end-2025; Greg Abel became CEO on 1 Jan 2026, with Buffett as chairman.) The analogy captures the buy-and-hold-and-compound structure — but Berkshire lets subsidiaries run themselves, whereas Bending Spoons intervenes deeply.
Static or declining?
More precisely: past-peak, stalled or underperforming — not zero-growth. Ferrari deliberately avoids AI companies ("already winning") and wants businesses "not already winning," because stalled-but-solid assets leave room for his cost-cut-and-reprice playbook.
Is it "buying data"?
This overstates it. Its edge is operational turnaround + financial engineering, not harvesting data for AI. Crucially, the datasets are unrelated — a notes app and a ticketing app don't pool into one behavioural model. The moat is operating discipline, not data.
Google/Microsoft overlap
Big tech's data dwarfs theirs — which is exactly why data isn't Bending Spoons' moat. In-product behaviour (how you take notes, how you send files) is first-party and product-specific, but Bending Spoons isn't competing with Google on data or ads. It makes money by operating single stalled products more profitably.
A/B testing, leverage, and the revenue reality
Explain large-scale A/B testing and how the leverage (~2.7× EBITDA) actually works. How much revenue did the transformed companies create versus before — did they revive, or keep declining? Did they try to grow market share?
Users are randomly split into groups that see different versions (price, paywall, onboarding). The business measures a metric (conversion, retention, revenue per user) and rolls out the winner. "Large-scale" means hundreds of experiments across millions of users at once — turning "what price?" from a guess into an engineering problem.
"~2.7× EBITDA" means net debt is ~2.7× annual EBITDA. Per the prospectus, each deal is financed with debt equal to the lower of (a) 85% of the acquisition's enterprise value, or (b) the max debt the target's projected free cash flow can fully repay within 5 years. In plain terms: the acquired company's own cash flow pays off the loan used to buy it — a classic leveraged buyout, then the freed cash funds the next deal.
Revived or declining? The data says: neither — it's "extraction"
- Revenue $387M (2023) → $1.31B (2025), +95% in 2025 — but almost entirely from bolting on new acquisitions (about half of revenue now comes from AOL).
- Organic growth was only 13% in 2025 (outside estimates put it as low as 3–5%).
- Net revenue retention (NRR) is 91–95% — below 100%, meaning the existing customer base's revenue is net shrinking. Without new deals, the portfolio would flatline or slip.
Growth comes from price hikes and shifting one-off purchases to subscriptions (komoot, Mosaic raised prices 2–4×), not from winning market share — the filing explicitly says it favours returns over organic growth. Individual results are mixed: Evernote revived (profitable in ~18 months); Remini grew 5× users / 9× revenue; but Brightcove revenue is still declining on subscriber losses.
Can't you extract the "playbook" by analysing many products?
With first-party data across many similar products, couldn't anyone extract the same behavioural patterns?
Yes — and this is the key insight. Pricing elasticity, paywall placement, onboarding funnels, churn triggers are generalisable and extractable. But precisely because the pattern is extractable, the pattern is not the moat.
- Knowing "a 2× price rise lifts revenue" is the easy half; executing it on millions of live users without collapse is the hard half.
- Generic patterns leak and get copied.
- What's hard to copy is the combination: buying stalled assets cheaply + elite talent + the willingness to make unpopular moves (mass layoffs, steep price hikes) backed by a low-cost capital structure.
Your reasoning self-consistently defeats the "data is the moat" thesis: extractable ⇒ not a moat ⇒ the moat is execution and capital discipline.
Are the founders cashing out and offloading risk?
Is the testing real, and how are users chosen — could two people see different prices? And are the founders rushing to IPO to offload risk and pay debt, then take the money and leave? Predict the ending.
Testing is real; users usually don't know they're in an experiment; assignment is by ID hash / cohort / geography. Two people can see different prices (price testing), but new users typically bear experiments while existing users are "grandfathered." Group-level price testing is legal; pricing by protected characteristics is not.
The "cash out" thesis — half true, half wrong
Supports it: real high leverage (~$4.4–6B debt; interest of $143M in 2025 ≈ half of operating income); ~23.6M secondary shares were sold; going public shifts risk to public shareholders and lenders and paves the way for more leverage.
Against it: founders kept control via dual-class shares (80%+ voting power) — they're staying, not leaving. Most of the raise was primary (~$1B into the company for acquisitions), and use-of-proceeds is new deals, not debt repayment. Their ~$8.9B is paper wealth, likely under a ~6-month lock-up.
More accurate framing: not "founders exit" but "founders fuel an already-aggressive, highly-levered machine with public capital while keeping the wheel."
Three scenarios (not a single prediction)
- A — Flywheel keeps turning: cheap targets + cost cuts + reprice → compounding; the bull case.
- B — Engine stalls (your worry): NRR <100% + pricing tapped out + rising rates on the debt → a slow decline in valuation, not a dramatic blow-up. Constellation Software (the comparable) is already down 44% over a year.
- C — Debt blow-up + founder exit: possible but high-barrier; lock-ups and self-inflicted stock damage make a clean escape costly. Scenario B is more realistic than C.
Watch two numbers over time: does NRR fall below 90%, and does interest expense / operating income keep worsening? They'll tell you the flywheel is stalling before any headline does.
What is SaaS, as a business model?
Define SaaS and its operational business.
Software as a Service. Instead of buying software once and installing it, you subscribe (monthly/annually) to software the vendor hosts in the cloud.
- Recurring revenue — subscriptions renew, so revenue is predictable.
- Low marginal cost — one more user costs the vendor almost nothing, so margins scale.
- Key metrics — MRR/ARR, churn, NRR, CAC, LTV.
Bending Spoons is a classic SaaS base: 84% of Q1 2026 revenue was subscription, 12% advertising, 4% other. Predictable, repricable revenue is exactly what makes its playbook work.
How to read the prospectus like an analyst
Walk me through the company's IPO document — I want to learn.
An F-1 is the registration filing a foreign company uses to list in the US (domestic firms use an S-1); the "prospectus" is the investor-facing body. It's a seller-written but legally-liable self-introduction — the skill is separating narrative from hard fact, and knowing where the truth hides.
The anatomy (and how a pro reads each part)
- Cover — ticker, dual-class structure, and "foreign private issuer" status (⇒ lighter disclosure: no 10-Q, no 8-K). Read this and you already know how little you'll get to see.
- Letter / Summary — read as marketing; extract the hard metrics hidden inside (e.g. return hurdles of 65% levered / 25% unlevered; AI now writes >90% of code).
- Use of proceeds — does the money go to the company or to selling shareholders? (Here: mostly to the company, for acquisitions.)
- Summary financial data — highest signal density; the truth lives here.
- Risk factors — skip the boilerplate, hunt the company-specific red flags.
| Line | 2023 | 2024 | 2025 | Q1 2026 |
|---|---|---|---|---|
| Revenue | 387,067 | 671,053 | 1,306,404 | 601,321 |
| Operating income | 84,267 | 127,352 | 277,851 | 120,171 |
| Interest expense | 17,173 | 32,605 | 142,601 | 93,184 |
| Net income (GAAP) | 160,643 | 88,997 | (204) | 27,465 |
| Adjusted net income | 95,856 | 229,364 | 375,592 | 205,977 |
The three "tells" in these numbers
- GAAP net income ≈ 0 in 2025, but "adjusted" is $376M. The gap is add-backs — amortisation of acquired intangibles, deal costs, restructuring (layoffs). For a company that always acquires and restructures, those "one-offs" recur, so "adjusted" flatters reality. The "$500M profit" in the press is the adjusted figure.
- Interest expense exploded — $17M → $143M → $93M in a single quarter. That's the leverage bill, worsening fast.
- 2023's pretty $161M net income was propped up by a ~$103M tax benefit (see §12). Strip it out and the underlying year was weaker.
| Cash flow | 2023 | 2024 | 2025 |
|---|---|---|---|
| Operating | 59,154 | 204,895 | 290,600 |
| Investing | (170,294) | (889,654) | (1,851,874) |
| Financing | 110,664 | 664,044 | 1,935,932 |
Small-positive operating, big-negative investing, big-positive financing = the signature of an acquisition machine that can't self-fund its buying and plugs the gap with debt and equity.
A pro's reading order
- Cover — share structure, foreign-issuer status.
- Income statement — size the GAAP-vs-adjusted gap and what's added back.
- Cash-flow trio — self-funding or transfusion-dependent?
- Use of proceeds — company or insiders?
- Risk factors — hunt "material weakness / going concern / litigation / customer concentration."
- One piercing operating metric — here, NRR 91–95%.
Bending Spoons disclosed a material weakness in financial-reporting internal controls — a hard signal for a newly public company that its reporting process isn't yet mature.
Dual-class shares: 1 vote vs 5 votes
How do I read the dual-class structure? What's the difference between the two share classes?
Two share classes with the same economic rights but different voting rights. Each share is worth the same in dividends and liquidation; but ordinary shares get 1 vote while founders' super-voting shares get 5 votes.
The purpose: founders can sell most of the economics to raise money while keeping the control. Illustrative math: if founders own 30 of 100 economic shares but those carry 5 votes, they command 150 of 220 votes — ~68% control on 30% ownership. Bending Spoons' founders hold 80%+ of voting power this way.
- You buy a passenger ticket, not the wheel — you can't vote them out, block a deal, or force a strategy change.
- Upside: management can pursue a long-term plan free of short-term pressure.
- Downside: if they err or self-deal, you have little recourse — hence a "governance discount."
- Peers: Alphabet, Meta, Snap, Berkshire.
When you see dual-class, ask yourself one thing: do I trust these people at the wheel?
What is a tax benefit? Does every company have one?
What is a tax benefit — and does every company have this "one-time right"?
Normally income tax is an expense that reduces net income. A tax benefit is a "negative tax" — it adds to net income. On the income statement the tax line goes negative and props profit up.
The most common source is recognising a deferred tax asset from past losses:
A company (or one it acquired) lost money in the past and piled up net operating losses (NOLs). Tax law lets those historical losses offset taxes on future profits — essentially a "right to pay less tax later."
When the company judges it will "more likely than not" earn enough future profit to use those deductions, it can recognise that right as an asset — and doing so creates a tax benefit on the income statement, even with no cash coming in.
So — does every company have it? No.
- You need a history of losses. A firm profitable from day one has few NOLs.
- Recognition is conditional. If future profits aren't likely, a "valuation allowance" wipes the asset out.
- It's used up, not annual. As profits consume the NOLs, the benefit shrinks — so it's inherently one-off / non-recurring.
- Bending Spoons' twist: it buys once-loss-making companies, inheriting their NOLs, then turns them profitable and uses the deductions (subject to anti-abuse limits like US §382). That's why tax benefits surface in its numbers.
So when a year's net income looks unusually pretty, drop down to the pretax-income and tax lines and check whether the beauty was earned by the business or floated by a tax item. In 2023, stripping out the ~$103M benefit, the underlying business was far less impressive than the $161M net income suggests.
The "income tax" line on the income statement is computed on an accrual basis (including deferred, non-cash portions); it is not the same as the cash the company actually paid in tax that year.
Reading the cash-flow statement
Let's start reading the cash-flow statement.
Its core question: the income statement says you earned a profit — but did the cash actually arrive? Net income is an accounting concept full of non-cash items; the cash-flow statement tracks real money. Hence: profit is an opinion, cash is a fact.
Three drawers
- Operating (CFO) — cash from the core business: does it self-generate cash?
- Investing (CFI) — buying/selling assets, companies, equipment: is money going into expansion?
- Financing (CFF) — debt, repayments, share issuance, dividends: how much outside transfusion?
Lesson 1 — the indirect method
Operating cash flow isn't built from scratch; it starts from net income and works back:
We add non-cash items back because they reduced net income without any cash leaving. This is where last section's lesson pays off: deferred taxes are non-cash, so they're one of the reconciling items between profit and cash.
Lesson 2 — the profit-vs-cash gap at Bending Spoons
GAAP net income in 2025 was ≈ 0, yet operating cash flow was +$290.6M. That ~$291M gap is mostly non-cash add-backs — for a serial acquirer, dominated by amortisation of acquired intangibles (brands, customer relationships, tech booked at purchase), which depress accounting profit but cost no cash. So Bending Spoons looks barely profitable on paper yet generates real cash. The pro's job: decide whether that's healthy (profit suppressed by non-cash amortisation) or dangerous (cash squeezed from stretching suppliers). Here it's the former.
Lesson 3 — the combination speaks
Small-positive CFO (+$291M), big-negative CFI (−$1.85B), big-positive CFF (+$1.94B) = the fingerprint of a serial-acquisition machine — it can't self-fund its buying, so it plugs the gap with debt and equity. A mature company usually shows the opposite: big-positive CFO, small-negative CFI, negative CFF (repaying debt / paying dividends).
The amortisation added back into CFO corresponds to real cash that was spent — buying the companies — but it sits in investing (CFI), not operating. So CFO looks clean while the true cost hides in that −$1.85B. Never judge a serial acquirer on CFO alone; read all three drawers together, and look at cash left after acquisitions.
It's the cash left after keeping the business running — available to repay debt, pay dividends, or reinvest.
For a levered company, watch whether FCF keeps covering the ballooning interest bill (the key ratio: CFO ÷ interest expense) — that's the flywheel's margin of safety.
CFO is one of the most important truth-tellers, but not sufficient on its own. "Worth investing" = quality of cash generation × durability × price paid × balance-sheet risk. Verify CFO isn't propped up by stretching working capital; for a roll-up, weigh it against the acquisition spend in CFI; for a levered firm, weigh it against interest; and no matter how strong the cash flow, a price that's too high is still a bad investment.