Most software invested in over the last twenty years was held inside a structure that made one assumption: a good company could be identified, capitalized, scaled, and exited inside a decade. The structure worked because the assumption was mostly true. AI is breaking the assumption. The structure has to follow.
Why the AI curve is different
A SaaS company built in 2010 reached most of its enterprise value within ten years of founding. The product matured, the go-to-market settled, the gross margin profile stabilized, and the curve flattened — at which point a fund could mark, exit, and move on. The fund-cycle structure matched the underlying compounding curve almost perfectly.
AI businesses do not compound on that curve. They compound on a different one.
- Models keep improving. A product built on top of intelligence rides three years of free capability gain for every year it sits in the market. That gain shows up as feature surface, retention, and unit economics — none of which would have arrived in a SaaS company at the same age.
- Data accumulates. Every interaction is signal. Companies that capture and structure that signal get smarter faster than companies that do not. The compounding is not linear; it accelerates.
- Customer-driven feedback loops tighten. AI-native products learn from each customer in a way that traditional software products cannot. Year five looks qualitatively different from year three in a way that the previous era's P&L could not predict.
- Distribution flywheels strengthen. Word-of-mouth on a product that genuinely makes customers more capable compounds at a rate paid acquisition cannot match. By year seven, the cheapest channel is also the most valuable channel.
By the time the curve in a SaaS-era business was flattening, the curve in a comparable AI-era business is still bending up. A fund that has to mark and exit is exiting before most of the value has been created.
The structural misalignment
A fund-structured investor has a fiduciary obligation to its limited partners that forces a return of capital inside a defined window. That obligation is non-negotiable. It shows up, gradually, as exit timing optimized for portfolio markups rather than enterprise-value creation. The misalignment is not malicious. It is structural — built into the legal shape of the vehicle.
The same misalignment shows up in the operator's seat. The founder who is told the fund needs an exit in eighteen months runs the company differently than the founder who is told the holder will be there for the next two decades. The decisions about hiring, pricing, capital efficiency, and customer focus look different. So does the willingness to forgo a marketing-driven growth quarter in favor of a retention-driven one. The horizon shapes the operating discipline.
What permanent capital changes
A permanent-capital holding company does not have the mark-and-exit obligation. It can hold a company through the full compounding curve, which means it can also let the company operate against that curve from the inside. The founder who knows the holder is patient is freer to make the decisions that produce a real business — not the ones that produce a clean exit photo at year seven.
The advantages, in order:
- Time-arbitrage. The biggest dislocation between price and value in private markets right now is between fund-cycle pricing of AI-native companies and their actual long-curve value. Permanent capital is the structure most able to capture it.
- Operating coherence. One holder, one set of incentives, one horizon — across all companies in the portfolio. Founders are not optimizing against three different fund cycles at once.
- Compounding allocation. Cash from one mature company can fund the next venture, the next acquisition, the next bet. There is no return-of-capital event that breaks the chain.
- Reputation effects. Founders self-select toward holders who behave like owners. The best inbound flow follows the longest horizon.
Why this is not the same as old-style holding companies
Permanent capital deployed against AI-native companies is not a re-statement of Berkshire-style holding. The companies are not insurance subsidiaries or consumer brands; they are software organisms that improve continuously and require active operating involvement, especially in their first eighteen months. The structure borrows the patience of the old holding-company model and pairs it with the operating intensity of a venture studio.
That combination is hard to assemble inside a fund. It is natural inside a holding company built by operators who plan to be doing this work in twenty years.
How we underwrite against this
When we evaluate a company, we ask whether the curve we are buying is one that flattens within a decade or one that continues to bend up well beyond it. We are not interested in the first kind for our own portfolio — there are funds that can run that play, and they will run it more efficiently than we ever could. We are interested in the second kind, where the marginal value of each additional year of holding is large and rising.
That is the structural bet. Permanent capital is the right shape for it.