Rumors of Software’s Death Have Been Greatly Exaggerated

In 2011 venture capitalist Marc Andreessen wrote a famous essay titled "Why Software Is Eating the World”.

In that piece he laid out how software was getting cheaper, faster, and more disruptive to traditional businesses.

Software was the key differentiator that helped Apple beat Blackberry, Netflix beat cable, Spotify win music streaming, and so many more examples.

A decade later his thesis seemed correct, as the value of publicly traded software companies (excluding the Mag-7!) exceeded several trillion in market cap.

Software businesses had some of the highest margins, growth, and valuations in the world.

Although software engineers were expensive and in short supply, well run companies could afford to pay them as they achieved scale.

After all, once software is built, gaining additional customers turns most of the added revenue into pure profit.

It’s a beautiful business model, one of the best ever.

At least it was…

The New AI Threat

As new AI programming tools like Cursor, Claude Code, and ChatGPT Codex have rapidly improved in the past year, investors have begun to question the future of many software businesses.

Recent AI coding model improvements along with new AI agent capabilities have taken AI coding to the next level.

With some proper prompting and setup, these AI tools can quickly build sophisticated apps in mere hours.

Go on YouTube or X lately and you’ll see hundreds of non-technical people bragging about the detailed apps they’ve built purely by “vibe coding” with AI tools.

A handful of entrepreneurs have even bragged that they’ve dumped their existing software providers and replaced them with their own homegrown vibe coded versions.

This has led many people to speculate that existing software vendors could lose customers.

Even if they retain big corporate clients, it’s also possible that software companies could see lower paying users or “seats” within those clients due to AI productivity gains.

The stock market, as it usually does, shoots first and asks questions later.

Software stocks have sold off dramatically on fears that growth will screech to a halt, and that perhaps many software companies will now go into terminal decline.

Even the largest software ETF has underperformed the major stock market indexes significantly in the past few months:

Historically software stocks have traded at large valuation premiums relative to the S&P 500, but this is no longer true in early 2026:

So is this the beginning of the end for software companies?

Or could this selloff be overdone, giving us an incredible “buy the dip” opportunity?

Misunderstanding = Opportunity

Anytime an entire sector sells off this dramatically, we sharpen our pencils and get to work because usually there are at least a few misunderstood gems.

Although the market is treating every software company with skepticism right now, the competitive landscape, AI threats, and growth potential for each one is different.

And we do think there are some good buying opportunities emerging in this sector.

Why won’t AI kill all existing software? Here are 10 good reasons:

• AI tools are expensive

• AI often lacks robust security

• Software migrations can be a costly nightmare

• Software often has many complex features & integrations

• Proprietary data isn’t always accessible to AI

• Code requires ongoing maintenance & upgrades

• Large enterprises move slow

• Companies demand human accountability

• AI features offer new revenue opportunities

• AI productivity can enhance incumbents’ development speed

Although AI will get cheaper, right now it costs hundreds of dollars per month in token costs just to run Claude Code for a few days per month.

That doesn’t even consider the biggest ongoing cost of a software system: maintenance.

Any company building in house solutions will need a team dedicated to maintaining the system, as dependencies break and new features get requested.

Even if a company decides to go this route, there may be applications that they don’t even attempt to replace due to high difficulty and risk.

Think ERP systems, cybersecurity, user authentication, financial reporting, transaction processing, fraud detection, real time video, logistics & routing, backend databases & CDPs, payroll & tax, and more.

Take tax reporting as an example…

While an individual might be willing to use an LLM to guide them through filing their 1040 form this year, almost no company will drop their tax software.

The risk of making mistakes could be millions in penalties plus the embarrassment and cost of restating financial results.

On the flip side, marketing websites, survey or polling tools, to-do list apps, lightweight personal CRMs, and simple dashboards are much easier and less risky to replace.

There’s a lot of software that lives in the middle of these extremes.

For example, marketing automation software.

Some of the functions like contact management and metric dashboards can easily be replaced with AI, BUT there are other functions like analytics tracking and email/SMS delivery that are not easily built.

Plus there are deep integrations between a marketing automation system and a company’s backend databases via bidirectional data flows.

HubSpot and Salesforce may not look that complicated, but there are hundreds of key functions within the software which are maintained by thousands of engineers.

If you’ve ever done a migration from one marketing automation system to another, you know how big of a nightmare this can be.

Just imagine accidentally emailing the wrong campaign to thousands of your customers because an automation workflow broke. Or worse, losing critical new user data for days because of a data ingestion failure.

Even though some software CAN be replaced, it is not likely to be replaced, especially by medium and large organizations.

Most company decision makers want to have specific people to blame when things go wrong. They also need assurances on reliability and security, two areas where AI is still very weak.

Perhaps the biggest under-appreciated angle is that many existing software companies could actually GROW FASTER due to adding new AI enabled features.

For example, Notion, a productivity tool many believe might be at risk of AI coding disruption, has seen its growth accelerate in 2025 due to new AI features:

Software companies that are embracing AI assisted coding may benefit from a faster pace of development on new features, or even see margins increase with AI productivity gains.

What’s the Trade?

Subscribe to EVI Premium to read this post.