I spend too much of my waking hours questioning the impact of AI on the software industry. Currently, I’m playing with the thought experiment that, just like coffee, software is a commodity. What I find interesting in that framing is that, unlike coffee, the software price equilibrium was defined by a scarcity in production capabilities. The main driver of this production scarcity was the high capital requirements needed to produce and maintain software.

The pre-AI equilibrium

If software was a scarce commodity with high production costs (developer salaries, time) and infrastructure/maintenance as the carrying cost, we have a market where supply is constrained, prices are high, and there's significant unmet demand.

In that market, organizations ration software. They build what they need, maintain backlogs of things they wish they could build, and pay premium prices for producers. The software carrying cost (technical debt, infrastructure, on-call) acts like warehousing costs for physical goods.

In that market, software is slow to build and expensive to maintain. The price tags are high and margins are fat.

The supply shock

Agentic software development turns that world upside down. It makes software creation quicker and cheaper than ever before. By dramatically lowering the costs of software production, it has the potential to create a massive positive supply shock in the software world.

I see two immediate effects of this supply shock. They are already happening.

First, software enters a deflationary period and prices collapse. A race to the bottom ensues. SaaS investors have a hard time. Everyone panics and says “SaaS is dead”. On a personal level, I highly doubt it.

Then, producer margins get compressed. Engineers who were previously paid for the scarcity of their production skill see that premium erode. The value shifts from problem-solving to maintenance (curation, quality assurance, architecture). This is the equivalent of shifting from farming to supply chain management. Businesses that don’t have a strong moat will feel the heat of the competition.

The new software proliferation

If agentic software development continues improving, we might see a demand explosion for software. If we dare apply the Jevons paradox to this context, then as software gets cheaper, there will be stronger demand for it. Just as when coal became cheaper to burn, people didn't burn less coal, they burned vastly more.

Organizations that previously couldn't afford to build custom software now can. The total volume of software in the world increases dramatically. Unmet demand starts to be addressed and software continues eating the world at an even higher pace.
We witness a new wave of software proliferation. This wave scales existing software carrying costs to new heights. Everything changes, yet nothing changes.

This is where my thought experiment ends. Whether software is a commodity or not is not the real question. It’s a pretext to navigate the changes that our industry is going through. The biggest question now becomes: how can a software business succeed in this economic paradigm? That’s the question I’ll explore next.