The great software IP writedown

Open original

Martin Alderson argues that AI agents have collapsed the scarcity that used to justify high software margins, so the value of existing software IP, not just the cost of building it, may be falling sharply. He sees SaaS as the most exposed, since cheap bespoke software and easy cloning erode both the moat and the premium paid for acquired products.

  • Software's high margins came from scarcity of people who could build it, not from monopoly, network effects, or corruption. As agents erode that scarcity, prices should fall.
  • Jevons paradox means far more software will get built, but much of it will be bespoke and serve small markets, which weakens the case for SaaS businesses that sell one product to millions.
  • Cloning is now cheap (agents can recreate apps from screenshots, and switching costs are dropping as data export gets easier), so R&D spend may be a running cost rather than a lasting asset, and acquisition premiums for tech may be overstated.

Martin Alderson's core argument is that software has been unusually good at generating high margins, but not for the reasons people usually assume. He tests the three classic sources of outsized profit (monopoly, network effects, and corruption or regulatory capture) and finds they explain only a few big tech firms, not the typical small or mid-tier SaaS company with high gross margins. His explanation is scarcity: there were never enough people able to build software to meet demand, so whoever could build it captured the value. That scarcity is now collapsing as AI agents do a large share of the engineering, and he expects margins to fall accordingly. He pushes back on the idea that Jevons paradox will rescue the industry. Yes, cheap software means far more software gets built, but much of it is bespoke, built for one company's exact needs, so it can't be sold to thousands of customers the way classic SaaS could. Add Baumol's cost disease, where the slow parts of the economy become relatively more expensive as others speed up, and the bottleneck shifts to sales and marketing, which he says is as slow as ever for B2B deals. That undercuts the "software factory" pitch, since building hundreds of apps doesn't help if you can't sell them. He also points to how easy it now is to clone products, with the ArtCraft suite cited as an example, and to how agents can export data and reverse-engineer web apps, which weakens switching costs. Taken together, he thinks software IP is hard to defend and that the venture model of pouring cash into SaaS to scale fast looks even more fragile. The last section turns to finance. Loss-making software firms have long argued that R&D is an investment that produces a durable asset, the product. Alderson suggests that if the product can be copied cheaply and bespoke software has a smaller addressable market, that asset may be worth much less than the books assume, and a writedown of software IP could follow. The practical stakes are for investors and accountants who have been valuing these companies on the premise that their engineering spend builds lasting value.

Ask about this article