Is SaaS still a compounding machine? Yes — but only for the few, and the ground is moving.

The book opened on the compounding machine and the winter of 2021, when software was worshipped and then cruelly re-examined. So we end where we began: is SaaS still a compounding machine? The honest answer is more complex than either 2021’s euphoria or 2022’s despair. Yes, it can be — but true compounding never belonged to every company wearing the SaaS label, only to the few with high-quality retention, a deep moat, real cash flow, and pricing power. Recurring revenue isn’t high-quality revenue; high margin isn’t real return; a good company isn’t a good stock. Telling the rare machine apart from the many that merely charge like one is the whole craft of this book — and that was true in 2021, is true today, and stays true in the AI era.

But the machine’s foundation is being shaken more deeply than at any point since it was born, because AI is hitting all three of its pillars at once: the near-zero marginal cost behind its high margins (inference has a cost), the per-seat pricing model (why sell ten seats when one agent does the work?), and the entry-point position that anchors its moat (what’s defensible when everyone can call the same model?). When one technology reshapes a business model’s cost, pricing, and moat simultaneously, it isn’t a tailwind — it’s rewriting the rules.

The title says it: software is starting to do the work, not just help with it. That flip dissolves the “person” that seat pricing was anchored to, and it redistributes where value lands — because “software is more powerful” and “the software company is worth more” have always been two different things. So in the AI era, what is the truly scarce software asset? After the long detour, the answer returns to the word this book has circled from the first page: the moat — but specifically the moat AI can’t easily copy or bypass. Deep proprietary data and feedback loops, solid distribution, deeply embedded workflow, hard-won trust. The shallow moats that hid under “high growth” and “high margin” will have nowhere to hide under AI’s light; the genuinely deep ones become more precious for their scarcity.

This book won’t tell you which stock will rise — and shouldn’t; any specific prediction goes stale. What it hopes is that when you close it, facing any SaaS company, you can ask your own questions: is its compounding machine really turning? Is AI thickening or thinning its moat? Is the price making it a good company, or a good stock? What evidence would prove me wrong? When you can answer those independently, with discipline, you no longer need anyone else’s answer — because you already own the ability to find it. And in an era where software starts doing the work, that ability is the scarcest, most un-stale asset there is.

This is the conclusion of “Compounding in the Cloud: The Business of SaaS, and How to Invest in It.”