Everyone wants the AI story. Nobody wants to do the AI math. - 7/18

Everyone Wants the AI Story. Almost Nobody Wants the AI Math.

A quick note from Behind the Markets

Everyone wants the AI story. Almost nobody wants the AI math.

Real enterprise AI isn't magic. It's budgets, procurement, compliance, and systems that don't break when regulators show up. So let's talk about the angle Wall Street hates: the boring software that gets renewed.

One honest thread runs through today's issue, and you've heard it from us all month: the market has already bid up most of the obvious "AI infrastructure" and security names to nosebleed levels. So this is as much about what not to overpay for as it is about what to own.

1) Enterprise AI Reality Check: If It Doesn't Save Time, It Gets Cut

In a frothy market, companies "experiment." In a tight market, they demand payback. That's why the real AI adoption curve will look less like a rocket ship and more like a CFO's spreadsheet: Does it cut labor hours? Reduce errors and rework? Speed up revenue collection? Reduce compliance risk?

Here's the contrarian point: plenty of "AI" will be treated like marketing and get defunded the moment budgets tighten. But the tools that touch real workflows — and show measurable ROI — will quietly compound. There's no ticker for this section, because it's the test you apply to every AI name you look at, not a stock itself. When a company claims "AI," ask the CFO's questions: Is there a hard-dollar ROI (hours saved, headcount avoided, errors prevented), or just a slick demo? Is the AI bolted onto a workflow the customer has to run anyway, or is it a standalone "nice to have"? Is it already generating revenue, or still in "pilot"? The answers separate the software that survives budget season from the software that becomes a line-item casualty.

Bottom line: Forget the demo. Follow the ROI. If it can't justify itself, it won't survive budget season.

2) The Underfollowed Winners: Vertical SaaS With Built-In Data + Switching Costs

The mega-cap AI names get all the oxygen. But many of the best small/mid-cap opportunities are in vertical software businesses that already own the workflow, the data, and the compliance headaches. When a vertical SaaS provider adds AI to a must-have workflow, customers don't just "try it" — they renew, because switching is painful and getting it wrong is expensive. The hunting ground: companies selling into regulated or high-cost-of-error industries — healthcare, insurance, finance, industrial safety, government-adjacent workflows.

And here's the one that — refreshingly, for this month — has not run away from you.

Company: Veeva Systems (SYM: VEEV)
The dominant cloud-software backbone of the life-sciences industry — the definition of regulated vertical SaaS with a moat

Veeva is the textbook version of this thesis. It runs the mission-critical software that pharma and biotech companies use to manage clinical trials, regulatory submissions, quality, and commercial operations — a workflow so embedded, and so tangled up in FDA compliance, that customers essentially never leave. That's ~$3.2 billion in revenue growing in the mid-teens, ~$1.25 billion in annual free cash flow, roughly 1,550 customers including nearly every major drugmaker, and a genuine AI angle in its new "Falcon" agentic-AI platform aimed squarely at automating high-volume, compliance-sensitive work (medical/legal/regulatory content review). This is "boring subscription software with pricing power" that happens to be a real AI beneficiary.

Now the part that makes it interesting right now: unlike almost everything else in AI-software, Veeva is down roughly 18–21% over the past year and trades near $190 — well below its 52-week high (~$310) and its ~$244 average analyst target (~27% below). The reasons are real and worth respecting: it's migrating its huge CRM customer base off a legacy Salesforce partnership to its own platform (execution risk if migrations stumble), Salesforce/IQVIA are competing harder, it's concentrated entirely in one industry (life sciences), and at ~19x forward earnings it's cheaper than its history but not "cheap." Still — a wide-moat, cash-gushing regulated-SaaS leader that's fallen while the rest of software ripped is a rarer find this month than another all-time-high chart.

Bottom line: The best AI businesses won't always "look like AI." They'll look like boring subscription software with pricing power — and occasionally, like Veeva, they go on sale.

3) AI and the Compliance Trap: The Rulebook Is a Moat (If You Can Survive It)

Wall Street complains about regulation. Smart operators use it. As AI moves from novelty to production, the compliance burden rises — data privacy, audit trails, explainability, security requirements, vendor-risk controls. That's bad news for fly-by-night tools and good news for serious platforms that can document, control, and secure their models. The market often underprices compliance as a catalyst, but compliance is exactly how "experiments" become procurement-approved systems with multi-year budgets.

The obvious "compliance moat" beneficiaries we've discussed in prior issues have largely re-rated. But the screen is durable, so use it. Favor platforms where compliance is built in and billable — audit logs, access controls, model governance, and security that the customer is contractually or legally required to maintain. Favor recurring subscription revenue over one-time project work. And favor incumbents with the balance sheet to fund the (rising) cost of certifications and security infrastructure — because that cost is itself the moat that keeps smaller rivals out. Veeva (above) is one embodiment; the enterprise-identity and data-governance names we've featured before (Okta, Datadog) are others — most just aren't cheap today.

Bottom line: The AI gold rush ends when the auditors arrive. Own the businesses built for that world — but don't pay a fantasy price for the privilege.

4) The "Capex Shadow": AI Is Forcing a New Hardware Cycle (Even Outside Data Centers)

AI doesn't just drive GPUs. It forces upgrades in networking, storage, endpoint security, and data governance — creating second-order winners in the infrastructure and IT-services ecosystem, especially the firms selling critical infrastructure rather than hype. The right questions for small/mid-cap work: Is this company tied to mandatory upgrades or optional experiments? Is revenue recurring or project-based? Do customers have to keep spending just to stay secure and compliant?

Here's the uncomfortable truth again, because it's the theme of the whole issue: this trade has been found, and it's expensive. The natural candidates have gone vertical — NetApp (AI-storage) is up ~60% in a year and sits right at its average target; Fortinet (network security, the archetypal "mandatory recurring spend" name) has surged ~90% year-to-date to ~$167 and now trades ~20–30% above its average analyst target (consensus target ~$112, and one firm as low as $70). These are quality businesses riding a real, recurring, non-optional spending wave — but the market is pricing them for perfection. So rather than hand you a chase-here ticker, we're hand you the discipline: keep NetApp, Fortinet, and their peers (Palo Alto, CrowdStrike, Zscaler) on a watchlist, define the price you'd actually pay, and wait for the inevitable AI-sentiment air-pocket to bring one to you. The spending wave is real; the entry point on offer today is not attractive.

Bottom line: The next AI wave is a spending wave — mandatory, recurring, compliance-driven. Own the infrastructure behind it, but let valuation, not FOMO, set your entry.

Before You Go

AI isn't a religion. It's a tool. The market will overpay for magic and underpay for billing. You want the opposite.

And this month's lesson, one more time: even when you've found the right theme, the crowd may have found the right stocks first. The edge left on the table isn't the idea — it's the discipline to wait for the price.