Here's a funny thing about artificial intelligence: it might be creating one of the most dramatic market divergences we've seen in a long time. On one side, you have the companies selling the picks and shovels for this gold rush, like Nvidia, which keeps posting record revenues. On the other side, you have a whole bunch of software companies who might discover that AI is quietly sawing off the branch they're sitting on.

According to a report from Citrini Research, AI is setting off a powerful, and perhaps vicious, feedback loop. It works like this: a company deploys AI to cut labor costs. It then takes those savings and reinvests them into... more AI. This accelerates the very disruption that forced the cuts in the first place. The report describes it as a cycle with "no natural brake": "AI capability improves, payroll shrinks, spending softens… capability improves."

This isn't just a theoretical labor market story. This dynamic is starting to reshape the very economics of enterprise software.

The Erosion of Software's Moat

For years, the software-as-a-service (SaaS) model was a beautiful thing. Companies like Salesforce built empires on recurring revenue and high switching costs—once you're in their ecosystem, it's a pain to leave. That gave them incredible pricing power.

But AI, specifically agentic coding tools like Claude and Codex, is rapidly changing the math. The report suggests that a "competent developer" armed with these tools could replicate the core functionality of a mid-market SaaS product in a matter of weeks. Suddenly, the cost of building your own solution plummets.

This is moving from theory to practice in boardrooms. Citrini cites one example where a Fortune 500 company renewed a major SaaS contract—but only after securing a 30% discount. Their leverage? The credible threat of replacing the vendor with internally developed AI tools.

That's a structural threat. It means companies whose business models are built on pricing power and recurring licenses—think Salesforce, ServiceNow, and Atlassian—could be looking at long-term margin pressure as AI lowers the cost of both building and maintaining software.

The Irony of the Infrastructure Play

Here's the ironic twist. The companies enabling this software disruption are, for now, sitting pretty. Even as AI threatens to commoditize parts of the application layer, the firms building the foundational layer are operating at full capacity. The report notes that Nvidia "was still posting record revenues."

This divergence highlights a potential fundamental shift in market leadership. As AI reduces the economic value of certain types of human labor and software intermediaries, the companies providing the underlying compute power—the chips, the cloud capacity, the data centers—may capture a growing share of the profits. All this is happening while the SaaS companies, who helped popularize the cloud-based, subscription model that feeds this infrastructure, face the unintended consequences.

It's a reminder that in a technological revolution, the biggest winners aren't always the ones using the new tools, but the ones selling the indispensable components that make the tools work. For software investors, the question is no longer just about which company has the best AI feature, but whether the entire business of selling software is about to get a lot more competitive—and a lot less profitable.