There was a time when software companies could rely on growth rates and compelling narratives about scalable business models. Revenue was paramount, profit an afterthought. A decade ago, investors routinely paid six to ten times annual recurring revenue for SaaS businesses, and that felt entirely reasonable. Anyone who raised concerns about cash flow was dismissed as an old-fashioned investor who did not understand the future. That era has ended.
Software stocks outpaced the broader equity market by more than a factor of two between 2014 and 2020. Since then, sentiment has turned. Following the rate hikes that came after the monetary excesses of the pandemic, investors returned to fundamentals: cash flows, earnings, valuations grounded in tangible metrics. Companies that could show only revenue lost their premium multiples.
Now the sector is facing a second, more severe shock. The market is beginning to fear that artificial intelligence will not save software, but disrupt it. The release of Anthropic’s latest models, Claude Sonnet 4.5 and 4.6, appears to have accelerated that concern. Their strong performance in reasoning and coding has fueled the view that specialized software providers could become redundant.
The result has been a broad selloff, with little distinction between potential winners and losers. That is not entirely misguided, but investors willing to look beyond the immediate casualties may find opportunity.
The AI paradox
Two risks need to be distinguished. The first is the build-it-yourself risk: AI has lowered the barrier to software development, potentially enabling new entrants to challenge incumbents. This risk is overstated. High switching costs—data migration, retraining, and embedded workflows—mean customers do not change vendors easily. Cheaper alternatives have always existed, yet established players have largely held their ground.
The second risk is more serious: the licensing model. As AI agents take over human tasks, companies may require fewer user licenses at renewal. This puts pressure on the traditional per-seat pricing model. One response is a shift toward usage-based pricing. Vendors that charge based on consumption rather than seats could benefit if AI agents use systems more intensively than human users ever did.
Software companies currently operate with gross margins of 75 to 90 percent. Their cost structures are dominated by fixed expenses, particularly personnel in development, sales, and support. These are precisely the costs most exposed to automation. Firms that use AI to reduce headcount in functions that underpin their own business models could see net margins rise structurally and disproportionately, as savings at such high gross margins flow directly to the bottom line.
At current valuations of two to four times revenue, many software stocks are priced as if growth has permanently stalled. Assuming long-term net margins of around 30 percent—the level at which mature players such as Microsoft and Adobe have operated for years—investors are effectively paying six to thirteen times normalized earnings. That is a valuation more typical of businesses with no future. In an AI-driven environment, margins are at least as likely to exceed that level as to fall short.
The market has priced in the risks but largely ignored the offsetting margin expansion. Less growth, more profit—economically, that can be just as attractive for shareholders. That scenario is currently valued at close to zero.
Opportunities in the decline
Where are the more concrete opportunities? Cybersecurity firms appear less vulnerable. When it comes to security, clients prioritize reliability and reputation over cost savings. As investment in data centers accelerates, spending on protecting that data is unlikely to be constrained. Platform-oriented companies such as ServiceNow and Workday, which automate business processes, have embedded AI agents deeply into their systems.
As workflows become more automated, platform usage is more likely to increase than decline. Legacy names such as Salesforce may appear exposed, but their lower valuations create scope for margin expansion. The company once acquired targets at twenty times revenue; today, similar assets can be bought at roughly twenty times free cash flow. That represents a materially stronger negotiating position.
The correction has created opportunities, but not universally and not without conditions.
A price-to-earnings ratio is only a starting point. Historical multiples say little about whether a stock is cheap. The key question is always future cash flow: whether underlying earnings are sustainable, whether margins are durable. “Cheap” is not a definition but a hypothesis. The strong balance sheets of many software companies—net cash positions and limited debt—make current valuations all the more striking.
The correction has created opportunities, but not universally and not without conditions. Companies that use AI to improve efficiency, help clients integrate new technologies, and treat shareholders seriously in capital allocation still have a credible story. Those are the businesses the market is waiting for. The rest are waiting for the next hype cycle, which may not come. This is an entry point, even if it does not feel like one.
Han Dieperink is chief investment officer at Aureus Asset Management. Earlier in his career, he serves as chief investment officer at Rabobank and Schretlen & Co.
.