Key Takeaways

  • Fleeing software for physical industries, defense, and nuclear creates heavy capital dilution and execution hurdles rather than guaranteed defensibility.
  • Historic venture returns concentrate in a tiny cluster of business models: social ad flywheels, retail scale economies like Zepto, capital moats, and embedded transaction rails.
  • AI code generation commoditizes feature builds, but enterprise switching costs protect core systems of record from displacement.
  • Products like NetSuite and QuickBooks maintain high retention despite poor user experience because workflow integration outweighs interface complaints.
  • Hybrid monetization, such as Toast pairing front-end workflow software with payment processing volume, builds durable revenue lines that raw software clones cannot touch.

The Dilution Trap in Deep Tech

Market panic over generative AI has sparked predictions of an absolute SaaS apocalypse. Investors are rushing away from pure software to chase defense tech, nuclear power, and manufacturing hardware. Will Robbins views this mass migration as an overreaction that ignores basic capitalization math.

“If I had to take a bit of a contrarian pessimistic view on something, it's that there's been too much overcorrection away from pure software business models towards the deep tech, the nuclear, the defense, the physical,” Robbins noted. “These are fundamentally extremely dilutive. They're fundamentally very hard to do.”

Hard tech requires massive capital expenditure cycles before reaching cash flow generation. Every successive equity check dilutes early investors and founders. While building a nuclear reactor or autonomous defense drone creates an obvious engineering barrier, the capital required to reach commercial scale often erodes fund-level return multiples.

Why Systems of Record Outlast Feature Clones

Software defensibility was never about the difficulty of writing code. It has always rested on business model structure, workflow gravity, and high friction around vendor replacement.

“Most of the returns have come from a very small number of business models,” Robbins stated. “To be clear, business model is different than sector.” Outside of ad-driven consumer networks and retail scale operations like Zepto, the most durable software companies anchor themselves directly into financial workflows or payment pipelines.

Robbins points to vertical platforms like Toast, which distribute operational software to restaurants and monetize through payment processing rather than high subscription seats. When software controls the checkout terminal, point of sale, and back-office accounting, an AI model that generates equivalent interface code poses little threat to the underlying cash collection engine.

This structural lock-in explains why legacy enterprise accounting tools retain dominant market share year after year. “NetSuite is one of the great businesses of all time, and all the QuickBooks style accounting systems have done a really remarkable job over the past 20, 30 years in the markets,” Robbins said. “They're sticky and the most hated, unusable products maybe in the entire world.”

CFOs do not rip out core general ledgers or ERP systems simply because a competitor launches a cleaner user experience or an AI-generated interface. The switching costs, data migration risks, and audit compliance requirements create an effective barrier to entry. AI will rapidly collapse the cost of building commodity software features, but it does not remove the organizational friction of replacing systems of record.

Why It Matters

Capital allocators rotating out of enterprise software into capital-heavy physical sectors are taking on balance sheet risk to avoid code commoditization. The durable software winners in an AI-heavy market will not be thin workflow wrappers; they will be platforms that command mission-critical data, run transactional rails, and enforce high organizational switching costs.