Hold Forever, Sell Never

Constellation Software has made roughly 1,000 acquisitions since 1995. Bending Spoons has made north of 100 in the last twelve years. Neither company has ever sold a portfolio asset. In Constellation’s case the founder, Mark Leonard, calls the one early exception “the deal I most regret.” In Bending Spoons’ case the CEO, Luca Ferrari, puts it more directly: “we have never sold an acquired business and we’re confident we’re the right long-term owner.”

These are unusual sentences in a market that has spent forty years optimising for the opposite. The dominant private equity model is buy-fix-flip on a 5–7 year clock, with the exit baked into the entry IRR before the closing dinner is over. Constellation and Bending Spoons inverted the sentence. Buy, fix, hold forever. The asset class their existence implies is larger than the unicorn class, and almost nobody is pricing it correctly.

The Constellation Model

Software engineers will recognise the Constellation Software architecture as a pattern. It is decentralisation taken so seriously it becomes a religion.

The numbers, before the philosophy. Constellation trades at roughly C$100B market cap on around C$10B in revenue and $2B+ in annual free cash flow. Headcount sits near 125,000. Headquarters staff is approximately twenty people. That ratio — 1 HQ employee for every 6,250 operating employees — is not a typo. It is the entire thesis. The CAGR of the share price since the 2006 IPO has been roughly 35% compounded over nearly two decades, which is the kind of return that makes spreadsheets stop and ask whether the formula is broken.

What Mark Leonard built was a permanent capital vehicle for vertical market software. Vertical is the operative word. Constellation does not buy general-purpose SaaS. It buys the boring software that the world’s gymnastics federations and regional dental supply distributors and Australian car dealerships actually run their operations on. Markets too small to attract a venture round. Software too essential to switch off. Pricing power constrained by the customer’s revenue base — typically less than 1% — but defended by the fact that ripping out the system would cost the customer far more than the licence ever did. Recurring revenue above 70%. Capital-light to grow. The financial profile of a utility wearing a software mask.

The acquisition discipline is the part that doesn’t generalise easily. Constellation screens roughly 5,000 companies a year. It deploys 88% of operating free cash flow into new acquisitions. It targets 30% cash return on invested capital on smaller deals and accepts 15–20% on the larger ones, where capital is harder to put to work. Most importantly, it never centralises functions across the portfolio. There is no shared sales team, no shared HR, no shared marketing automation. Each subsidiary operates as a standalone unit. Leonard’s argument for this — he was rational about it, almost stubborn — is that the diseconomies of scale in software businesses are larger than the savings, because the moment you centralise, you flatten the local pricing power and operational knowledge that made the asset acquirable in the first place.

The compensation structure is the closing argument. 75% of senior executive bonuses are paid in cash that the executive is then required to use to purchase Constellation stock. That stock is escrowed for up to four years. The result is roughly 100 employee-millionaires inside the company who became wealthy by being long the same compounding vehicle as their shareholders, on the same time horizon. Leonard himself has flown economy and stayed at modest hotels for twenty years, on the explicit grounds that he didn’t want to “freeload on CSI shareholders.” The prose style is dry. The behaviour is religious.

He stepped down as president in October 2025 for health reasons. Mark Miller leads now. The compounding hasn’t stopped.

The Bending Spoons Counterpoint

Bending Spoons is the version of the same idea built for consumer technology rather than vertical B2B software. It is also, in important ways, the opposite of Constellation.

Founded in Milan in 2013, Bending Spoons reached an $11B valuation in October 2025 on roughly $700M of 2024 revenue, with 2025 projecting to clear $1.2B. EBITDA margins sit near 35%. Monthly active users touch 300 million across the portfolio. Paying customers pass 10 million. The portfolio reads like a roll call of internet brands the venture economy gave up on: AOL ($1.5B in October 2025), Vimeo (announced September 2025), Brightcove (a $233M take-private in 2024), Evernote (2023), WeTransfer (2024), Meetup (2024), StreamYard (2024), Issuu (2024), Komoot (2025), Harvest (2025), Remini, plus a long tail of smaller acquisitions. Roughly a hundred products in total, run from a single engineering and operations platform out of Milan.

The playbook is brutal and the playbook is consistent. Step one: identify a venture zombie — a company with a large established user base, brand recognition, recurring revenue, and a board that has quietly given up on the next round. Step two: acquire at a valuation the previous investors are now desperate to take. Step three: lay off the majority of the staff, consolidate operations into the Milan platform, and shut down the geographically distributed engineering and support teams. Step four: aggressively raise prices on the existing user base, restrict the free tier, optimise the data and ad monetisation. Step five: never sell.

The Evernote case is the canonical example. Pre-acquisition, Evernote was running roughly $100M ARR on 250 employees, unprofitable for years, spiritually adrift after a decade of failing to become the second brain of every knowledge worker on earth. Bending Spoons acquired it in 2023, eliminated nearly all the US and Chile-based staff, moved core engineering to Europe, and approximately doubled the premium subscription tier to €129.99 / €169.99 annually. Twelve months later, Evernote was profitable for the first time in its institutional history. The user base hated some of it, the customer reviews on Trustpilot are still bruised, and the company is healthier than it has been at any point in the last decade. Both things are true at once.

What makes the model work at scale is the platform underneath the portfolio. Bending Spoons runs roughly fifty internal tools shared across all its products: a single recruiting pipeline, a single payments stack, a single marketing automation layer, a single behavioural analytics rig, a single experiment framework. The semi-fixed cost of running a hundred internet products approaches the marginal cost of running one well-instrumented one. This is the part that does not show up in the headline EBITDA margin. It is the entire reason the EBITDA margin exists.

The funding stack is worth a glance. The October 2025 round added $270M of equity at the $11B mark, on top of a $2.8B debt facility led by JPMorgan, BNP Paribas, HSBC and Goldman Sachs. The cap table includes the obligatory celebrity investors — Eric Schmidt, Bradley Cooper, Andre Agassi, The Weeknd, Xavier Niel, which is an oddly Wes-Anderson-cast assemblage — but it also includes the institutional growth pile: T. Rowe Price, Baillie Gifford, Fidelity, Durable Capital, Cox Enterprises. Ferrari has said an IPO is “more likely than not” in the not-too-distant future, with NYSE the probable venue. Whether that happens or not, the company is now structurally large enough that the hold forever claim is the load-bearing element of the thesis. If the wall ever cracks — if a single asset gets sold under duress — the whole architecture loses its premium.

Two Models, One Asset Class

The two companies look like opposites until you ask them the same question, and they give you the same answer. Why do you never sell? Both will tell you, in slightly different vocabularies, that the entire mathematical edge of compounding only works if the compounding period is allowed to be very long, and that the LBO industry’s structural insistence on a 5–7 year exit window destroys roughly half of the value the underlying asset can produce, because the half that compounds most aggressively only exists in years 8 through 25. Holding forever is not a sentimental claim about loving your assets. It is a quant claim about where the integral lives.

This is also why you cannot easily replicate either model with hired-in talent on a fund-of-funds structure. Both companies have founder-operators who have publicly precommitted to never selling, and whose entire personal wealth is tied to the same equity that LP wealth is tied to, on the same indefinite horizon. The moment you delegate the holding decision to a CIO with a four-year contract, you reintroduce the time pressure the model was designed to escape. Permanent capital has to actually be permanent, not just permanent-looking. Both Leonard and Ferrari understood this and built the cap table around it.

Venture Zombies as an Asset Class

The interesting investment thesis embedded in both companies is not about the companies themselves. It is about the supply curve.

The venture power law produces, by Andrew Dumont’s estimate, roughly 80% “failed” outcomes — failed in the sense of did not return the venture fund’s preferred multiple, not failed in the sense of bad business. These are the venture zombies. Cash-flow positive or near-positive. Real customers. Real product. Real brand equity. Sub-scale relative to the venture fund’s required outcome. Stagnant growth. Tired team. Frustrated cap table. Approximately worth nothing to the venture LP and approximately worth several hundred million dollars to a permanent-capital acquirer who can run the asset on a different time horizon and a different cost structure.

This is the largest under-priced asset class in private markets. The reason it is under-priced is structural, not informational: the people who own the assets (venture funds) have a return profile that requires unicorn outcomes, and the people who could pay realistic prices for the assets (permanent capital vehicles) are vanishingly few. The bid-ask spread is enormous. Bending Spoons exists inside that spread. Constellation exists inside an adjacent version of the same spread. Andrew Wilkinson’s Tiny — now part of Metaplanet — was an earlier and smaller bet on the same observation, focused on bootstrapped internet businesses that never took venture in the first place. SaaS.group out of Germany is doing a micro-version. Calm Capital, Arising Ventures, a handful of others are wading in around the edges.

The capacity of the asset class is, roughly, the entire venture power-law tail. Which is to say: it is enormous, and the constraint is not the supply of assets but the supply of patient capital and operating muscle willing to be the buyer.

What This Means For ACAL — and For Sidechain

The implication for any European fund is uncomfortable in its simplicity. The traditional exit pathways for portfolio companies — strategic sale, secondary, IPO, sponsor-to-sponsor trade — are not going away, but they are becoming less attractive at the margin as the permanent-capital buyer pool grows. Bending Spoons is now a credible bid in any process where the target has a large user base and a stagnant growth profile. Constellation is now a credible bid in any process where the target is a small vertical software business with sticky revenue. Both will pay realistic prices on long horizons. Both should be on every European fund’s exit list, alongside the traditional names.

For Sidechain Power specifically, the alignment is even cleaner. Energy software fits the vertical market software definition with embarrassing precision. Niche vertical (battery aggregation, virtual power plant orchestration). High recurring revenue once the asset base is built (FCR-D and aFRR participation, day-ahead optimisation, the dispatch layer). Capital-light scaling characteristics once the platform exists. Pricing power defended by switching costs that are operationally enormous. Customer base that grows as the energy transition does its slow Swiss work. This is, almost line by line, a description of what Constellation buys.

The signal to operators is therefore the same as the signal to founders: build your company so it could be owned, comfortably, by a permanent-capital vehicle. That means recurring revenue, defensible margins, a lean cost structure that does not depend on growth heroics, and a culture that can survive a change of ownership without the institutional knowledge walking out the door in a single quarter. None of that is incompatible with venture funding. All of it is incompatible with the grow at all costs until you flip operating model that the last decade rewarded.

Why This Happens In Toronto and Milan, Not New York

A small geographic note before closing. It is not an accident that the two most successful permanent-capital software vehicles in the world are based in Toronto and Milan rather than San Francisco or New York. The American private equity industry has spent forty years building incentive structures that punish the long hold and reward the rapid flip — carry crystallisation on exit, fund vintages on a clock, LP return expectations calibrated to a specific J-curve. None of those incentives exist for a founder-operator who built the vehicle from the inside. Both Leonard and Ferrari designed their cap tables to escape the LP-clock pathology specifically. Both were able to do so because they were operating outside the gravitational field of the American PE machine, which made the unconventional compensation structures and unconventional reporting cadences locally tolerable in a way they probably would not have been in Greenwich.

Europe and Canada are not better at capital allocation than the United States. They are simply less optimised, in a way that occasionally allows the unoptimised path to win. The most interesting permanent-capital story of the next decade may continue to come from the parts of the financial map that the American compounding machine considers peripheral. That is its own kind of moat.

The Closing Note

Mark Leonard’s most quoted line is the one about wanting to build “a permanent capital vehicle for the vertical market software industry, where you didn’t have to buy and sell a company — you could keep it forever and build real businesses that last a lifetime.” Read it again, slowly. Notice that the sentence is not about returns. It is about what kind of company you get to build if you remove the exit requirement from the design constraints.

That is the trade. Hold forever, and you give up the carry crystallisation and the IPO bell-ringing and the LP-letter heroics of the flip. In exchange, you get to compound at 35% for thirty years, you get to keep the operating asset, and you get to keep the people. Most of the private equity industry will never make this trade because the incentive structure prevents it. The two companies in this piece made the trade because they built the incentive structures themselves, from scratch, before the industry could install the pathology.

That is the lesson. The buildings of the next compounding decade will be designed by the people who refuse to sign the standard lease.