The latest rout in global software stocks wasn’t triggered by a profit warning or a messy earnings season. It was sparked by a far more uncomfortable idea: that artificial intelligence may be eroding the economic foundations that software investors have relied on for decades.
This week’s selloff followed the release of a new automation tool from Anthropic, which pushed advanced AI agents deeper into workflows traditionally dominated by enterprise software — from legal review and compliance to research and data analysis. Markets reacted swiftly, wiping hundreds of billions of dollars off the sector as investors reassessed what software businesses can realistically charge in an AI-first world.
According to deVere Group chief executive Nigel Green, this wasn’t a panic about AI replacing software altogether. It was a repricing of pricing power itself.
“The selloff is not about fear of AI — it’s about what software businesses can realistically charge in an AI-first world,” Green said, arguing that subscription-heavy models lose leverage when intelligent systems can replicate outputs instantly and cheaply.
“Investors are reassessing whether decades-old assumptions around recurring revenues still hold.”
From recurring revenue to recurring doubt
For years, enterprise software was prized for predictable subscription income, sticky workflows and high switching costs. Those qualities justified premium valuations and underpinned the assumption that margins would remain durable.
AI challenges each of those pillars at once.
Tasks that once supported long-term contracts and premium pricing — legal research, document review, analytics and process automation — are being compressed by AI systems capable of delivering comparable results in seconds. The result is a growing market question: what exactly is scarce — and therefore valuable — in software today?
As Green noted, investors are no longer rewarding innovation alone.
“The sharp falls in software stocks reflect a market recognising that margins, not innovation, are now the battleground.”
Switching costs start to crack
Another pressure point is customer lock-in. As AI systems improve, the friction that once made it painful to change platforms is weakening. Outputs become more standardised, alternatives proliferate and loyalty becomes harder to monetise.
Much of software’s valuation logic rested on the idea that once embedded, platforms were difficult to dislodge. AI, by stripping complexity out of workflows, threatens that advantage.
“Markets are drawing a clear distinction between companies that genuinely control AI economics and those that simply integrate AI to protect existing businesses,” Green said. “The former can potentially expand margins, while the latter risk seeing cost savings passed directly to clients.
“Markets are, it seems, penalising firms that rely on legacy platforms, high headcount or process-heavy models that can be bypassed entirely.”
A valuation reset, not a verdict
Importantly, this repricing is happening before any widespread earnings damage has shown up, Green noted.
“Investors aren’t waiting for earnings warnings or guidance cues,” he said. “They’re repricing now, because AI accelerates disruption faster than quarterly results can capture.”
That helps explain the speed and breadth of the selloff. Software businesses built around information resale, labour substitution or process-heavy models are being judged against a harsher standard: how defensible are those revenues when AI can bypass them entirely?
At the same time, this isn’t a blanket verdict on all software, Green said. If anything, the market is beginning to differentiate more aggressively.
Companies that own critical data, control AI infrastructure or operate at points of real scarcity may ultimately benefit as value chains compress and returns concentrate, with a small number of companies set to “capture disproportionate gains while a far larger group will struggle to defend pricing power”, he said.
What to watch next
The key question now is whether this week’s move proves to be an overcorrection or the early phase of a longer reset. Software results have not yet collapsed. Customers still rely heavily on digital platforms. And AI itself often runs on — and alongside — software rather than replacing it outright.
But the psychological shift is real. The assumption that digital products naturally enjoy durable pricing power is being challenged in real time.
For investors, the takeaway is uncomfortable but necessary: in the AI era, recurring revenue is no longer enough. The market wants proof that pricing power is defensible — and it’s no longer prepared to wait patiently for that proof to arrive.