AI eats Software

"Markets are never wrong – opinions often are."

– Trading Legend Jesse Livermore (as quoted in the book “1929”)

Software.

For years, it was our favorite subsector.

And why not? The business model was almost too good to be true: asset-light operations, subscription revenue, high visibility. Once you've built the platform, every new customer flows straight to the bottom line. Translation: fat margins, gushing free cash flow, and profitability that makes accountants smile. Add in the moats—network effects, switching costs—and you've got yourself a compounding machine.

The poster child?

Microsoft.

It started as a programming language company and built product after product. Then came Office 365, now a subscription staple on every desk. They copied what worked elsewhere—think Zoom, which became Microsoft Teams—and bundled in extras like Co-Pilot and Microsoft Calendar (lifted straight from Calendly's playbook). The goal? Lock you into their ecosystem through switching costs while leveraging network effects: everyone uses it, everyone knows it, so everyone stays on board.

It's no surprise that non-capital-intensive businesses like Microsoft have crushed capital-intensive ones by a massive margin (chart 1) over the past decades.

But as Heraclitus observed 3,000 years ago: the only constant is change.

Enter large language models (LLM).

ChatGPT. Gemini. And now Claude — once a beloved coding LLM, now stepping onto the consumer stage with its latest launch: Claude Cowork.

Suddenly, a new villain appeared.

Software stocks responded with a brutal sell-off, with some names dropping 10%, 20%, even 30% in a matter of days.

Now let's address the elephant in the room.

Is this the end of software?

Chart 1: Capital-light businesses have outperformed capital-intensive businesses over decades

What’s Causing the Software Stock Collapse?

First things first: what's behind the carnage?

While it's hard to pin a prolonged drawdown on a single cause, AI has become the recurring villain in this story.

For years now, companies across the AI landscape have been dismantling the barriers to entry in web development through increasingly capable coding agents. Anthropic's Claude Code is widely considered one of the most advanced, and last week they democratized its power with the launch of Claude Cowork—bringing complex problem-solving capabilities to non-developers for the first time.

This release rattled investors.

A belief now seems pervasive on Wall Street: with lower barriers to entry in web development, companies will become less reliant on third-party software vendors.

Take Intuit.

A business we owned for years. The parent of TurboTax, the tax software that helps millions file their returns in the US—a process that, as we've noted, ranks among life's great bureaucratic nightmares. The company has a formidable moat built on network effects and switching costs. The product is so valuable that people willingly pay for it rather than use the US government’s free alternative. Yes, US tax filing is such a headache that people pay for software to help them do it.

Then came Claude Cowork. One user claimed that work that normally took 40 hours—an entire work week—was compressed into 15 minutes.

The time of a coffee break.

The result?

Intuit's stock dropped almost 20% in days after the Cowork launch (chart 2).

But Intuit isn't alone in facing the "AI eating Software" narrative. We've already highlighted others—Adobe being the poster child, down 70% from its highs over the past months despite firing on all cylinders operationally.

Other household names have suffered too:

• ServiceNow fell roughly 10% in one week
• Salesforce down 13%
• DocuSign down 17%
• Atlassian down 18%
• HubSpot down 15%
• Workday down 9%
• Figma down 19%
• SAP down 7%

Yet, all their "Good Stories" are still fully intact. That means valuations have compressed dramatically.

Take Adobe.

Its free cash flow yield now sits at 7% with a P/E of 13—compared to historical averages of 3% and 27, respectively.

The story is consistent. Although nothing has fundamentally changed yet in these businesses, the market is pricing in a future where AI devours the world—or at least drastically reduces the importance of code, the software built on it, and the products sold at premium prices. The fear is that companies will simply build in-house solutions instead.

Which raises another question: Is the AI threat real?

Or is this the biggest buying opportunity ever?

Chart 2: Intuit, owner of tax software Turbo Tax, shedding almost 20% in days after launch of Claude Cowork (IBD Partner)

Is the AI Threat Real?

Let’s start with a thought experiment.

While Claude’s advances will undoubtedly encourage companies to experiment with more in-house development, the idea that a “vibe-coded” prototype is good enough to rip out and replace mission-critical enterprise software is, in my view, highly unlikely.

Imagine walking into the management team of an SMI company and suggesting the following: They should abandon SAP and migrate to an internally built system created by a handful of developers and Claude Code.

That would mean:

• Migrating decades of data
• Retraining thousands of employees
• Rebuilding compliance, reporting, and controls
• And doing all of this without a single error — because earnings are reported every quarter

Yeah… good luck with that.

It’s an extreme example, but it makes the point clear: critical systems are sticky, and switching them is risky.

And beyond the initial build, any developer knows this truth: large-scale software is never “set it and forget it.” Systems break. Requirements change. Security threats evolve. Regulations shift. Over time, those systems demand dedicated teams, ongoing maintenance, and constant iteration.

At that point, the key question becomes: Is this cheaper than buying from a best-in-class vendor?

And is it better for the teams using it?

Now, none of this means the AI threat should be dismissed.

AI is real. It will lower barriers to entry. It will create new competitors — many of which don’t exist yet. Legacy software companies that fail to adapt will be left behind.

But here’s the other side of the coin.

AI can also strengthen incumbents.

The SAPs, Salesforces, Adobes, and ServiceNows of the world already sit on massive installed bases, proprietary data, and deeply embedded workflows. AI allows them to increase product velocity, ship features faster, and enhance user productivity — all without forcing customers to change platforms.

Take Adobe.

It owns the most valuable and extensive creative content database in the world. It has integrated AI through Firefly. It runs on an affordable subscription model. And millions of users are already trained on its tools.

Yet the stock is priced as if the business is structurally broken.

It isn’t.

The fundamentals tell a completely different story—one in which, if you only look at the fundamentals, they point to a screaming buy (chart 3; for a deeper dive, see here and here).

So, how do we at arvy approach this moment?

Chart 3: Adobe’s quarterly revenues and forward P/E valuation

Markets are never Wrong – Opinions Often Are

We live by this principle, borrowed from the wisdom of trading legend Jesse Livermore.

And while the "Good Story" is undeniably compelling—with valuation alone suggesting 30% to 50% upside—we're staying put.

The "Good Chart" tells us to.

What we see is a tumbling, falling trend with no signs yet of stopping or stabilization (chart 4). Moving averages—clear, objective signals from the market—point south as well.

Simply put: the trend is down.

And as long as these beautifully compelling, cheaply valued stories aren't honored and confirmed by a rising trend, we won't touch them. They're all on our watchlist, but we wait.

We listen to Mr. Market and accept his current verdict. Right now, his narrative is clear: “AI eats software.” And we know from experience that such narratives can dominate sentiment — and stock prices — for months, sometimes years.

So, we don’t try to be heroes. We don’t bottom-fish. We don’t bet on turnarounds. We wait until we see a trend reversal.

Because in the end, as Jesse said, the stock price is what pays your bills.

And markets are never wrong.

Opinions often are.

Chart 4: Selected software stocks over the last year