The Software vs. Hardware Divide
One of the defining investment narratives of early 2026 has been the divergence between software and hardware stocks. While semiconductor and chip-related companies have surged — buoyed by insatiable demand for AI infrastructure — software stocks have broadly underperformed. The iShares Expanded Tech-Software ETF (IGV) has lagged significantly behind the semiconductor index (SMH), creating a stark visual divergence that has rattled investors. Yet this sell-first-ask-questions-later approach to software may be throwing the baby out with the bathwater.
The Oversimplification Trap
Markets have a tendency to oversimplify complex technological transitions, and the AI revolution is no exception. The prevailing fear is straightforward: if AI can write code, companies can build their own software in-house, eliminating the need for expensive SaaS subscriptions. This reasoning, while intuitive, fundamentally misunderstands what makes enterprise software valuable and how deeply embedded it is in business operations.
The reality is far more nuanced. Enterprise software — the modern incarnation of what was once simply called "business software" — is the backbone on which companies operate. It is not a nice-to-have; it is the infrastructure of daily business life. The question investors should be asking is not whether AI can produce code, but whether that capability actually threatens the specific moats these companies have built.
The Moat That Matters: Core vs. Icing
Not all software companies are created equal in the face of AI disruption. The critical distinction lies between companies that provide the meal — the core operational platform businesses depend on daily — and those that provide the icing — supplementary tools that are convenient but ultimately dispensable.
Microsoft sits firmly in the "meal" category. The vast majority of the business world runs on Microsoft's ecosystem: Exchange for email, Excel for data, PowerPoint for presentations, Word for documents. Years of accumulated data, workflows, and organizational knowledge are stored within these platforms. No amount of vibe coding is going to replicate that entrenched position overnight.
Salesforce occupies a similarly fortified position. Companies run their entire sales operations through Salesforce, with proprietary data stored in formats deeply integrated into their business processes. The switching costs are not merely financial — they involve migrating years of customer relationship data, retraining entire sales organizations, and risking operational disruption during the transition.
Google commands its own enterprise stronghold through Google Workspace and its cloud infrastructure, serving as the operational backbone for millions of businesses worldwide.
The software companies that should worry are those whose products fall into the "icing" category — tools that businesses periodically evaluate and ask, "Do we really need this, or can we get along without it?" These are the names most vulnerable to AI-driven replacement.
The Vibe Coding Fallacy
The notion that a company manager or owner could simply use AI to rewrite enterprise software in-house betrays a fundamental misunderstanding of what enterprise software entails. Writing functional code is only a fraction of the challenge. The real questions are: Do you truly understand the complex business logic you are trying to replicate? Can you maintain, secure, and scale that software over time? And most critically, can you migrate the vast stores of proprietary data that currently run in formats designed for existing platforms?
For most companies, the answer to these questions is a resounding no. Enterprise software is not just lines of code — it is an ecosystem of data formats, integrations, compliance frameworks, and institutional knowledge that has been refined over decades.
AI as Augmentation, Not Replacement
There is a meaningful distinction between AI as a productivity enhancer and AI as a wholesale replacement for human workers and existing software systems. The hype cycle tends to collapse this distinction, painting a picture of a near future where machines take over entirely and white-collar workers become obsolete.
This vision ignores the significant holes that still exist in current AI models. Having worked in AI development, I can say with confidence that these systems are not ready to replace human judgment. They can augment human capabilities — making workers more productive, automating routine tasks, and surfacing insights from large datasets — but they remain tools in service of human decision-making, not substitutes for it.
The Investment Opportunity
The market's indiscriminate punishment of software stocks has likely created pockets of genuine value. The key to identifying opportunities lies in asking the right questions: Is this software core to business operations or merely supplementary? Does the company hold proprietary data in formats that create high switching costs? Is the product deeply embedded in daily workflows in ways that would be extraordinarily difficult to replicate?
Companies that pass these tests — the Microsofts, Salesforces, and Googles of the world — are likely to emerge from this period of uncertainty not as victims of AI disruption, but as its primary beneficiaries, integrating AI capabilities into their existing platforms and deepening their competitive moats in the process.
The old advice applies well here: don't cry until you're actually hurt. And for the giants of enterprise software, the pain the market is pricing in may never actually arrive.