How Bending Spoons Built a $20 Billion Compounder | Luca Ferrari

Luca Ferrari describes Bending Spoons as a technology company that buys products it believes can become much more successful, integrates them onto a shared platform, and improves their economics over a roughly 10-year horizon. The company uses the resulting free cash flow to acquire larger businesses, repeating the process. He says Bending Spoons has completed approximately 50–60 acquisitions, including Airtable, AOL, Vimeo, Eventbrite, Komoot, and WeTransfer. Anish Acharya frames the company as an unusual combination of intellectual ambition and operational discipline. Ferrari says the model depends on turning an acquired business’s earnings into more earnings: for example, improving EBITDA from 100 to 200, using leverage responsibly, and buying additional companies. In his view, that can compound far more powerfully than optimizing for the highest possible valuation at any particular moment.

Discussing Marc Andreessen’s distinction between companies that behave like bonds and those that behave like options, Ferrari agrees that companies can reaccelerate after plateauing, but says it is rare. Apple’s iPhone was a watershed invention that transformed the company, while NVIDIA’s move from graphics hardware toward AI was enabled by both invention and strong technological tailwinds. In many alternate versions of history, those breakthroughs would not have happened or would have been captured by someone else. Ferrari distinguishes radical, step-change innovation from incremental innovation. Radical innovation can create a much higher peak quickly, as Google Search and the iPhone demonstrate, but its median outcome is poor and it is difficult to depend on. Incremental optimization is more measurable and reliable. He argues that companies should choose a dominant cultural posture rather than trying to be equally committed to incompatible mindsets, while still avoiding an extreme focus on one at the expense of the other. In Bending Spoons’ case, incremental improvement is especially powerful because it is applied across an acquisition portfolio and then compounded through reinvestment.

Ferrari says investors sometimes miscalculate the relationship between growth metrics, subscribers, profits, and long-term value. However, many sophisticated investors are rationally responding to the fact that other investors may value a company based on a future sale or financing round rather than on the cash it will generate if held indefinitely. The resulting market equilibrium may be globally inefficient, but individual investors can still be rational to play by its rules. For founders, Ferrari recommends first clarifying the objective and its boundary conditions, then listing the available strategies. A founder seeking the highest near-term valuation may need to satisfy investor expectations, while another may prioritize a cash-generative company and ultimately distribute dividends. Bending Spoons chose to optimize the long-run compounding rate of free cash flow even when a different acquisition or management approach might have produced a better valuation immediately. Ferrari’s argument is that the valuation multiple is only one factor: if the underlying cash-flow base becomes ten times larger, a lower eventual multiple can still produce the better outcome. He calls long-term thinking an advantage because people are naturally impatient and easily distracted.

Ferrari says Bending Spoons’ portfolio allows it to make expected-value bets that a single-product company might avoid. For an individual company, the median outcome matters much more than it does for a venture portfolio: a risky bet may improve expected returns while changing the founder’s likely outcome from “I end up with something” to “I end up with nothing.” He therefore does not believe single-product CEOs are necessarily irrationally cautious; they may reasonably dislike the downside of the median case. On debt, Ferrari says leverage is beneficial when the expected return on investment exceeds the cost of debt and the probability of default remains negligible. Bending Spoons keeps leverage below 3x, currently around 2.4–2.5x and historically closer to 2x; it aims to be able to repay all debt before maturity even under severe adverse scenarios, refinances to keep maturities more than five years away, hedges 100% of the debt against interest-rate changes, and relies on diversification, profitability, and strong cash conversion. He warns that diversification is not absolute because assets can become correlated under a particular shock. His rule is to begin cautiously, develop strong models and controls, and accept taking two or three years longer rather than risking a crash.

Ferrari says SaaS valuations were often unjustifiably high before the recent AI reset, but not because AI suddenly made software easy to copy. Many valuations assumed 12% organic growth would persist for 15 years, even though growth usually declines. AI gave investors a reason to revisit those assumptions; Ferrari considers the reset broadly reasonable, within perhaps 20% either way. He argues that most successful software products were already inexpensive to replicate technically. WeTransfer, for example, could perhaps be copied to 80% of its quality by five strong graduates in three months, and many clones already exist. The real defenses are first-mover advantage, brand familiarity, psychological and practical switching costs, network effects, proprietary data, and scale. Eventbrite benefits because 90 million people use it monthly and organizers post events there because the audience is already present. Komoot accumulates unique route information, including observations about muddy paths, rain, dogs, photos, and turns. Airtable’s customers generally pay less than $20,000 per year, so even if an AI agent could eventually reproduce much of the product, retaining a small amount of human accountability would often cost more than simply paying for the software. Ferrari expects AI to affect customer acquisition and margins, but not to make most established products disappear overnight.

Bending Spoons builds most of the tools it uses to run digital businesses, excluding universal products such as Gmail, Slack, and Google Drive. Ferrari says the company has more than 50 internal tools covering data processing, A/B testing, marketing automation, lifetime-value prediction, credential management, and recruiting. The benefits are not mainly vendor-cost savings; the tools are tailored to Bending Spoons’ needs, integrated with one another, and can be deployed as a coherent platform when an acquisition is brought in. Ferrari highlights an internal Slack-based agent called “alt Spooner.” It has access to the same repositories, folders, and systems as the employee using it. A user can ask it to retrieve data, request credentials when necessary, perform an analysis, and return charts and commentary in minutes. He recounts an Evernote manager asking the agent to check whether a bug had been reported, find its root cause, write a fix, and send it to the engineering lead for review; the process completed in a way that would previously have required much more coordination. AI now handles nearly all routine customer-support responses, leaving support managers to design the response process and user experience. Ferrari estimates that AI performs at least 90% of routine data fetching, cleaning, and plotting, and that 94% of code was AI-written at the last measurement, compared with 0% two years earlier. He says good companies should already be somewhere between roughly 70% and 99% AI-generated code, depending on their approach.

Ferrari rejects a strict opposition between instinct and analytics. Data, intuition, creativity, and excitement are all inputs that help generate options, but decisions should ultimately be made through logic and rationality. An analysis can suggest that a product will perform better in France, while an instinct can flag that something is wrong; neither input should determine the decision by itself. Even highly creative ideas should be examined to see whether the initial excitement is justified. Asked when a founder should quit, Ferrari admits he is not a reliable guide: his previous startup, Everteam, spent three years trying to build AI-generated diaries, failed to find usage or paying customers, and repeatedly pivoted. The company eventually ran out of money after an expected Danish government tax refund was denied or delayed, so bureaucracy effectively made the decision. For future entrepreneurs, he recommends intellectual independence, very high ambition, rationality, and a talent-dense team. He stresses that improving talent density is extraordinarily difficult if a company begins with weak talent density, making the quality of the initial team a foundational strategic choice.