Software stocks are down nearly 20% in 2026. AI hype is at an all-time high. Venture capitalists are pouring record sums into AI-native startups while traditional SaaS companies watch their valuations crater. The market is telling you something, and it's not subtle: the era of software as a guaranteed growth engine is over. But this isn't a death sentence. It's a repricing. And if you understand what's actually happening, the picture is more nuanced, and more interesting, than the headlines suggest.

The numbers tell a clear story

The S&P Software & Services index is down nearly 20% year to date. The broader Nasdaq is only down a modest 2-3%. That gap is enormous, and it's not random. By mid-March 2026, an estimated $2 trillion in market capitalization had been erased from the enterprise software sector. Forbes reported that software stocks lost over $1 trillion in the first weeks of the year alone. Industry stalwarts like Salesforce, ServiceNow, and Microsoft saw sharp single-day drops of 3-6% on multiple occasions. Meanwhile, AI infrastructure plays are booming. Optical networking companies like Lumentum and Ciena, data center firms like Vertiv, and storage companies like Seagate are posting massive gains. The money hasn't left tech. It's moved to a different part of it. Perhaps the most striking data point comes from SaaStr: software forward P/E multiples have now fallen below the overall S&P 500. Not at parity. Below. This has never happened before. Not in 2022 when rates spiked. Not in 2008. Not even during the dot-com crash.

The market's thesis: AI commoditizes custom software

The selloff isn't just about sentiment. It reflects a genuine structural concern: what happens to SaaS businesses when an AI agent can do what their product does? The catalyst that crystallized this fear was Anthropic's launch of new tools for its Claude Cowork product in early 2026. Within hours, global software stocks were in freefall. CNBC reported that IT stocks across the U.S., Asia, and Europe all posted declines as investors processed the implications. The logic, whether fully correct or not, goes like this: if AI can automate legal workflows, customer service interactions, data analysis, and code generation, then why would an enterprise keep paying per-seat licenses for dozens of specialized SaaS tools? The value proposition of many software products, doing a specific task in a specific workflow, starts to look thin when a general-purpose AI can handle the same task and ten others. J.P. Morgan Research pushed back, calling the selloff driven by "broken logic." Their argument: high-profile AI products from major software companies are growing rapidly but still account for only a small fraction of overall sales. The disruption is real but slower than the market is pricing in. But the market doesn't wait for disruption to finish. It prices in the trajectory.

SaaS metrics were built for a different world

The traditional SaaS playbook relied on a few key assumptions: high switching costs, predictable expansion revenue, and sticky user bases. The metrics that defined success, ARR, NRR, net dollar retention, all assumed that once a customer was in, they stayed in and spent more over time. AI is quietly undermining each of these assumptions. Switching costs are falling. When AI can replicate core functionality, the moat around any individual product narrows. Bain & Company noted that while average gross retention is still around 90% or better for now, investor confidence has clearly shifted. The market is pricing in what retention will look like in two years, not what it looks like today. The valuation math has changed accordingly. Private SaaS companies now trade at 3x to 7x ARR, with a median around 4.5x. The days of double-digit revenue multiples for private companies are, as one analyst put it, "not coming back." Investors have shifted focus from pure ARR growth to capital-efficient growth, making NRR the primary driver of SaaS valuation multiples. And here's what makes the AI-native alternatives particularly threatening: they don't need the same business model at all. AI-native B2B SaaS companies have a median gross retention rate of just 40%, according to Gainsight research. That sounds terrible, but it doesn't matter as much when customer acquisition costs are near zero and the product improves with every interaction. The old metrics don't map cleanly onto the new economics.

The winners are becoming "AI-native workflow companies"

Not every software company is getting punished equally. Reuters reported that companies with proprietary data are seen as having a key defense against the AI threat. Salesforce, with its deep customer relationship data, and Oracle, with its enterprise resource planning data, are better positioned than companies whose data is more standardized and replicable. The companies navigating this well are making a fundamental shift: from selling software to selling outcomes. SaaS companies are rapidly moving from per-seat pricing to hybrid or outcome-based pricing models. They're embedding AI deeply into their products, not as a feature, but as the core value delivery mechanism. But most can't make the jump. The challenge is that developing and supporting AI features requires significant investment that squeezes profit margins. You need to spend more to transform your product while your stock price is falling and investors are demanding efficiency. It's a brutal catch-22.

The cloud transition parallel

We've seen a version of this before. When cloud computing emerged in the late 2000s and early 2010s, on-premise software vendors faced an existential question: adapt or get repriced. Many couldn't pivot. Companies that had built their entire business around selling perpetual licenses and maintenance contracts struggled to shift to recurring revenue models. Their stock prices suffered for years during the transition, even when their products were still widely used. The ones that made it through, like Microsoft under Satya Nadella, emerged stronger. The ones that didn't became acquisition targets or irrelevant. The AI transition rhymes with this, but it's faster and the stakes are higher. Cloud computing changed how software was delivered. AI changes what software is. A cloud migration meant moving the same product to different infrastructure. An AI transformation means rethinking whether the product needs to exist at all.

What this means for developers and startups

For developers, the paradox is sharp. Building software has never been easier. AI coding assistants, vibe coding, and no-code platforms mean that a single developer can ship in days what used to take a team months. But building a software business has never been harder. When anyone can build an app quickly, the barrier to entry disappears. The competitive advantage shifts from technical execution to distribution, data, and taste. The question isn't "can you build it?" but "can you build a moat around it?" Venture capital is following this logic. AI startups accounted for 41% of the $128 billion in venture dollars raised on Carta last year, a record-high share. Crunchbase reported that it will be "very difficult for a SaaS company without native AI/agentic capabilities to find VC dollars at any stage." The money is flowing toward AI-native companies and away from traditional SaaS, and that shift is accelerating. AI-native startups like Cursor, Lovable, and StackBlitz crossed $10 million in ARR within 18 months of launching. That kind of velocity was almost unheard of in the previous SaaS era. It signals something important: the next generation of software companies will look fundamentally different from the current one.

The counterargument: maybe the market is wrong

It's worth considering that the market might be overreacting. J.P. Morgan thinks so. Many analysts see the current selloff as indiscriminate, punishing good businesses alongside vulnerable ones. The reality is that people use more software than ever. Global SaaS market revenue is still growing and is projected to reach $315 billion by 2026, heading toward over $1 trillion by 2032. Enterprise software spend is forecast to rise at least 40% by 2027. The market for software isn't shrinking. It's the market for software stocks that's repricing. There's a meaningful difference between a product being replaced and a business model being disrupted. Most enterprises won't rip out Salesforce or SAP tomorrow, even if an AI could theoretically replace parts of what they do. The installed base is massive, the data is entrenched, and the risk of switching is high for complex workflows. So maybe software is just on sale. Maybe the smartest trade of 2026 is buying great software companies at a 25% discount while everyone else panics about AI. But even the optimistic case requires acknowledging that something has structurally changed. The premium the market once placed on software, the assumption that recurring revenue and high margins made these businesses inherently more valuable than the broader market, that premium is gone. For the first time ever, software trades at a discount to the S&P 500. Even if the selloff reverses, that shift in perception won't.

What's actually being repriced

The Nasdaq isn't saying software is dead. It's saying something more specific: the default assumption that software companies deserve premium multiples is dead. For two decades, software was the closest thing to a guaranteed growth trade. High margins, recurring revenue, low capital expenditure, and expanding markets made it the darling of growth investors. AI doesn't eliminate any of those qualities, but it introduces a new variable: the possibility that your product's core functionality could be replicated by a general-purpose system at a fraction of the cost. That possibility, even if it's years from materializing for most companies, is enough to reset the risk premium. And that's what we're watching: not the death of software, but the death of software exceptionalism. The companies that survive and thrive will be the ones that recognize this shift and build accordingly. They'll own proprietary data, deliver measurable outcomes, and treat AI as their operating system rather than a feature to bolt on. The rest will learn what on-premise vendors learned a decade ago: the market doesn't wait for you to catch up.

References