Software Development Now Costs Less Than Minimum Wage

TL;DR: When AI drives software development costs toward zero, traditional technology moats collapse — and the surviving moats are distribution, data, and utility-based pricing. Geoffrey Huntley’s argument is the infrastructure-layer version of the AI moat question, and the title is a literal claim he is making about per-hour cost.

What it means

Huntley’s argument is the most uncomfortable read in modern strategy. If writing software becomes near-free, then every moat built on the difficulty of writing software evaporates. Switching-costs from technology lock-in — the historical bread-and-butter of enterprise SaaS — are “provably falsified now.” If a competitor can rebuild your product in a weekend with AI assistance, your codebase is not a moat. It is a budget item. The industry has not internalized this yet, and the people running enterprise SaaS companies are about to find out the hard way.

This is a direct challenge to Helmer’s process-power as applied to software companies. Toyota’s production system took decades to build because manufacturing processes involve physical constraints and tacit human knowledge. Software processes, by contrast, can potentially be replicated by AI that has access to the patterns and best practices. The open question is whether process-power survives the commoditization of its substrate. The honest read is that some software process power survives (the kind that’s about org culture and judgment) and some doesn’t (the kind that’s about codebase complexity).

The new moat landscape that emerges is leaner and harder. Distribution — having the customer relationships and brand awareness to sell — becomes more important when the product itself is easier to replicate. Data moats (per cheating-is-all-you-need) grow in importance. And pricing models shift from seat-based SaaS (which charges for access to software) to utility-based pricing (which charges for outcomes).

The argument

Technology lock-in is dead. The classic SaaS moat — “it would take years to rebuild this on another platform” — dies when AI can rebuild it in days. Enterprise buyers will increasingly demand portability, and the cost of switching shrinks with every improvement in AI-assisted development. This undermines switching-costs as a durable Power for software companies, which is a much bigger deal than most SaaS founders are ready to admit.

Distribution survives. Building software may be cheap, but getting it in front of buyers is not. Sales teams, brand recognition, channel partnerships, and trust built over years — these don’t get automated away by AI. Distribution becomes the primary moat when product differentiation shrinks.

Utility-based pricing. When software costs nothing to build, charging per-seat feels increasingly arbitrary. The shift toward pricing based on value delivered (usage, outcomes, transactions) is both inevitable and strategically important — it aligns the company’s revenue with the customer’s willingness to pay, which is a much more defensible position than charging for access to a thing the buyer could now build themselves.

Data as the durable moat. Proprietary data that improves the product over time is the one moat that AI strengthens rather than weakens. More users generate more data, better data produces better AI features, better features attract more users. This connects directly to cheating-is-all-you-need and the data-moat thesis. The companies that started accumulating data before everyone realized data was the moat are the ones positioned to win this decade.