Applied Materials reported earnings this week, and the results were clean. Revenue came in at a record $7.91 billion, up 11% from a year ago. Good quarter, but not the interesting part…

For the third quarter, management guided for revenue of $8.95 billion, a significant rise compared to last year and a big beat on what the market was expecting. To drive the point home, the CEO made his most bullish public statement to date: "We expect our semiconductor equipment business to grow more than 30% in calendar 2026."

Amazon, Google, Microsoft, and Oracle are collectively spending hundreds of billions building out data centres and AI infrastructure. That money doesn't instantly reach every corner of the supply chain: it flows slowly, like water finding its way through rock. First the hyperscalers commit the CAPEX. Then the chip designers like Nvidia get the orders. Then the chip manufacturers like TSMC need to build more capacity. And only then does the equipment that builds the chip factories (Applied Materials' bread and butter) see the full surge in demand. As the CFO put it bluntly: "The growth in AI that Applied has been investing for is now in full force."

This is precisely the "picks and shovels" dynamic we've been writing about for the past year. While the market chases the most visible names in AI, a small group of less-glamorous companies are quietly supplying the entire industry regardless of which model wins or which chip architecture dominates. AMAT's stock is already up over 60% year to date, and the fundamentals might soon catch up to the share price.

Constellation Software has been weighed down by a single, nagging question from investors: can an empire of old-school, niche software businesses survive the AI era?

If tools like Claude or ChatGPT can help even a non-technical person build a decent piece of software from scratch, what stops Constellation's customers (think a mid-sized municipal utility or a regional dental clinic chain) from simply cancelling their subscriptions and rolling their own solution? This is a legitimate question, one that has knocked billions off the company's market cap without, so far, a single dollar of actual damage to the underlying businesses.

Last week, CSU’s Q1 2026 results beat expectations, with revenue coming in at $3.18 billion and earnings per share of $27.37, well above the $25.28 forecast. But the more important number was zero. Zero customer attrition attributable to AI. Across more than 1,500 software businesses spanning over 150 verticals, management reported not a single meaningful customer loss driven by AI competition.

Understanding who Constellation's customers actually are matters enormously here. These are not Silicon Valley startups. Constellation's typical customer is a small organisation running software that lives on a server in their own building, in industries like healthcare, utilities, or local government, where change is slow and regulations are dense. These customers aren't going to wake up tomorrow and rebuild their systems with AI. Many of them are still relying on workflows that the rest of the world abandoned a decade ago.

As the new CEO noted on the call, the company's defence comes from being intimate with its customers, often serving dozens or a few hundred clients in very specific niches. That intimacy, and the switching costs that come with deeply embedded niche software, is a moat that a general-purpose AI chatbot doesn't easily dissolve.

The AI disruption narrative in software tends to assume a customer base that is eager to adopt new technology and capable of doing so quickly. For Veeva Systems, Salesforce, or Adobe, whose customers are more sophisticated, the reality on the ground may look a lot more like Constellation's: slower, stickier, and far more resilient to disruption than the market currently gives credit for.

Upgrade to unlock full access to the RatedA Portfolio (with weekly updates on every buy/sell and rationale), and the complete database of companies + ratings on www.rateda.co.