Wilkinson argues that the traditional moat for software—the high cost and difficulty of hiring programmers—has vanished. He compares it to a machine that makes perfect pizza cheaply: consumer quality rises, but business margins plummet. Lasting value must now come from other sources like brand or distribution.

The primary threat of AI to software isn't rendering it obsolete, but rather challenging its growth model. AI will make it harder for SaaS companies to implement annual price increases and will compress valuation multiples, creating stress for over-leveraged firms from the zero-interest-rate era.

AI is making core software functionality nearly free, creating an existential crisis for traditional SaaS companies. The old model of 90%+ gross margins is disappearing. The future will be dominated by a few large AI players with lower margins, alongside a strategic shift towards monetizing high-value services.

The primary threat of Large Language Models to the SaaS industry isn't that they will build better software, but that they will enable the creation of 50 to 100 competitors for every existing player. This massive increase in competition will inevitably compress profit margins for everyone.

The market's downturn in legacy SaaS isn't primarily about AI automating jobs within those companies. The core fear is that new competitors can now use AI to build feature-complete products at a fraction of the cost, creating intense pricing pressure and margin compression for incumbents.

Software has long commanded premium valuations due to near-zero marginal distribution costs. AI breaks this model. The significant, variable cost of inference means expenses scale with usage, fundamentally altering software's economic profile and forcing valuations down toward those of traditional industries.

Unlike cable or power companies that benefit from regional monopolies, AI intelligence is a globally competitive, frictionless market. This dynamic is 'so much worse' for business because it allows for perfect arbitrage, driving the price of intelligence toward zero and making it incredibly difficult to build a sustainable, high-margin business on the infrastructure layer.

As AI makes software creation accessible to everyone, Silicon Valley's historical edge—knowing how to code—disappears. The new defensible moats are assets like proprietary data, trust, or network effects, not the software itself, threatening the region's dominance.

As AI tooling advances, building complex applications becomes trivial, commoditizing software development. Defensibility can no longer come from technical execution. Companies must find moats in business models, distribution, or data, as simply 'building what customers want' is no longer a competitive advantage.