What If Everyone Is Wrong About Software?
The Most Misread Narrative in Markets Right Now
If AI is going to replace software, why hasn’t it done it already?
That is not a rhetorical question. It is one of the most important questions in markets right now, and the answer is worth a lot of money to the people who get it right.
The prevailing narrative says AI kills software. That foundation models replace enterprise platforms, vibe-coding startups displace incumbents and the world’s largest companies rebuild their operational infrastructure from scratch and legacy software vendors slowly get left behind. I have spent the last few weeks going deep on this theme and I have come away thinking the market may have it backwards.
I think that narrative is wrong because what I am seeing is not AI replacing software. It is AI being absorbed by it.
And if that is right, the winners of the next phase are not who the market currently thinks they are.
The full thesis, the basket architecture, and all fifteen names are below.
Let’s be honest about the backdrop first.
Software has been ripped apart this year. IGV is down more than 20% year to date, and the median software stock is more than 40% below its 52-week high. Sector valuations have compressed to roughly 40% below their 5yr average on forward sales, while free cash flow multiples are more than 75% off the peak. The broader tape is not exactly offering comfort either. As I write this, the S&P is sitting on key support with the moving averages starting to roll over. There is a very real chance this gets worse before it gets better.
I am launching Phase 3 into that weakness deliberately.
This is not a trade for the next three weeks. It is a multi-year position being built at a point of maximum dislocation between sentiment and reality. If the market deteriorates further, the basket will feel pain in the short term. That is the cost of the entry point and I want to be upfront about it.
The question that really matters is whether the thesis is right.
The Thesis
The software selloff has bundled together two completely different stories, and that confusion is the opportunity.
On one side you have weak, sponsor-backed private software businesses from the zero-rate era. High leverage. Weak vintages. Over-earning point solutions. Businesses underwritten in 2021 on aggressive assumptions, thin moats, and cheap capital that no longer exists. Some of those companies deserved to be re-priced. Some of them face real disruption risk. The fifteen names in this basket are not those companies.
These are incumbent platforms that run the operational backbone of global enterprise. They sit inside payroll, finance, planning, compliance, customer workflows, security, logistics, and supply chain. They are well embedded into the machinery of how large companies actually function. And rather than being threatened by AI, I think they are the primary mechanism through which AI gets deployed at scale.
The market is treating all software as one story. It isn’t. The mistake starts with how people think about software in the first place. The bearish case may sound neat in theory. AI writes code, code becomes cheaper, thus startups build faster. Legacy vendors lose their edge but enterprise software is not just code. In many cases, code is the easy part. The hard part is everything around it.
It is architecture. Workflow logic, regulatory scar tissue, auditability, the integration with dozens of other systems. Customer retraining. Reliability under stress. Years of edge cases discovered the hard way. Years of decisions made by people who actually had to keep these platforms running inside live businesses. And that is where the moat lives.
The market keeps talking about foundation models as if they can simply be dropped into the core of enterprise systems and take over. I do not buy that. A payroll system cannot be mostly right. An enterprise resource planning system cannot hallucinate its way through compliance. A financial workflow cannot produce two different answers to the same question depending on how the model feels that day.
The closer you get to real enterprise workflow, the less tolerance there is for probabilistic nonsense. That is why I think the “AI replaces software” narrative breaks down the moment you move from demos to mission-critical systems. These platforms need to be repeatable, auditable, reliable, and deterministic.
The second mistake is assuming that faster code generation somehow collapses incumbent moats. It doesn’t. The market is acting like vibe-coding equals instant competition. It does not. Writing code faster is useful, but only if you know exactly what needs to be built, how it needs to behave under stress, how it fits into the broader workflow, and how it interacts with the dozens of other systems surrounding it.
A startup can generate code quickly, but that does not mean it can recreate decades of architecture, customer trust, workflow depth, compliance readiness, implementation history, and switching-cost gravity.
In enterprise software, the moat is not just what the product does. It is what the customer risks by leaving. And that is where the replacement narrative really falls apart.
The CFO of a global manufacturer is not ripping out an ERP because a startup has a cleaner interface and a better AI demo. They are thinking about data migration, regulatory exposure, integration with forty other systems, retraining thousands of employees, operational downtime, board accountability, and the non-trivial possibility that the entire project becomes an expensive disaster.
Those risks are real. They have always been real. AI does not make them disappear. If anything, it makes the incumbents stronger because the incumbents already own the workflow. That, to me, is the most important part of the thesis and the bit the market is missing.
The incumbents are not sitting still waiting to be disrupted. They are the ones with the installed base, the distribution, the domain expertise, the customer relationships, the workflow context, and in many cases the best data. If AI is going to be embedded into enterprise workflow, why would the value accrue to outsiders rather than the companies that already own that workflow?
That is why I think the winners of this phase are not the people trying to replace enterprise software from the outside. They are the incumbents absorbing AI from the inside.
That is how I think about what is happening now. AI gets pulled into the software stack as an intelligence layer. Not as a substitute for the platform, but as an enhancement to it. The platform still provides the reliability, controls, audit trail, and scalability. The AI layer improves productivity, speeds up decisions, automates repetitive workflows, and expands what the software can actually do.
Together, that creates something much more powerful than either on its own.
And crucially, the platform captures the value because the platform owns the customer relationship, the workflow, and the place where decisions actually get executed.
That is why I think this is not some niche side theme. This is the main way AI diffuses across the real economy.
Every top company runs on enterprise software. That software is now getting smarter. The companies that own those platforms are, in my view, the primary beneficiaries of everything the market has spent the last few years financing further down the stack.
The Valuation Case
Even if you ignore the thesis completely and just look at the numbers, the setup is attractive.
Software as a sector is trading at roughly 5x forward sales versus a 5-year average closer to 8x. That is around a 40% discount to recent history. On a growth-adjusted basis the discount is still material. P/E multiples are well below their medium-term average and free cash flow multiples have been crushed relative to prior peaks.
In other words, this is not a sector trading as if the market sees a coming acceleration in value. It is still trading as if disruption is imminent.
That is the disconnect. Because I do not think the underlying businesses have deteriorated in the way price is implying. In many cases, their strategic position has improved. Their AI products are more credible than they were a year ago. Their role in the stack is more important, not less. Their moats look deeper once you separate real workflow ownership from AI narrative tourism.