Two Moats in Enterprise AI Software — Palantir vs. Alphabet

TL;DR After the hardware layer, capex lands in the software and intelligence layer — where enterprises and governments actually use AI. I hold two names here: Palantir and Alphabet. Both ride the AI software wave, but their moats run in opposite directions. Palantir is narrow and deep — government, defense, and critical infrastructure. Alphabet is wide and massive — $243B cloud backlog, $175-185B in 2025 capex, and a vertically integrated stack that produces capex rather than just consumes it. Owning only one misses the story.

After the hardware layer, there's a second place capex lands — the software and intelligence layer, where enterprises and governments actually use AI.

This is where I hold two names: Palantir and Alphabet. Both sit in the AI software layer, but the way revenue flows and the character of their moats are completely different. One is embedded deep inside government, defense, and critical infrastructure. The other runs a hyperscaler cloud backlog and its own models — manufacturing capex cycles rather than just receiving them.

I hold both together because enterprise AI arrives through two fundamentally different mechanisms. To understand either one clearly, you have to see it against the other.

Palantir — The "Operating System" Embedded in Government and Defense

Let's be honest about Palantir first. Volatility has been high. There's been valuation noise, and there still is. So why am I not flinching? Because the contracts are real.

They recently won a contract with the FAA. They're already embedded deep inside the Department of Defense. The AIP platform is becoming, effectively, the operating system for AI-driven enterprise and government operations. This isn't "selling AI services." It's closer to "selling a core system you can't easily rip out once it's in."

Government and defense moats have two properties. First, replacement is hard. Once a system is installed, swapping it requires security clearances, retraining, and massive switching costs. Second, contract lengths are long. There's no quarterly churn like in consumer SaaS.

That's why selling Palantir on short-term noise misses the compounding window. The real value here isn't "next quarter's revenue guide" — it's "how deeply do the currently installed contracts root themselves over the next ten years."

Alphabet — The Most Underappreciated AI Compounder

Alphabet is one of the most underappreciated AI compounding machines in the market right now. And the street is finally starting to recognize it.

A few specific numbers. Google Cloud backlog just hit $243 billion. 2025 capex is running $175–$185 billion. Gemini is maturing rapidly and has closed the multimodal gap with competitors.

Each of those numbers matters because Alphabet isn't a company that merely "benefits from AI." It's a company that manufactures the AI capex cycle itself. Its own models (Gemini), its own cloud, its own TPUs — the depth of vertical integration is arguably the deepest across all of big tech.

The $243B cloud backlog figure is already-booked demand. That's a multi-year revenue pipeline. Fulfilling it requires $175B+ in capex, and that capex flows right back into Nvidia, TSM, Micron, and Vertiv. So Alphabet sits in the software layer but simultaneously generates demand for the infrastructure layer.

The Character of the Moats — Contracts vs. Platform Scale

Dimension	Palantir	Alphabet
Primary buyer	Government, defense, critical infra	Global enterprise, consumer
Moat character	Contract depth, security clearance, switching cost	Scale, vertical integration, own model
Revenue visibility	Long contracts	$243B cloud backlog
Stock volatility	High (valuation debate)	Medium (big tech average)
Risk	Valuation, commercial segment scaling	Regulation, search ad structural shift

Both can be summarized one-line as "AI software beneficiary," but the structure underneath is entirely different. Palantir's moat is narrow and deep (government and defense). Alphabet's moat is wide and less deep individually but enormous in aggregate (global platform scale).

So How Do You Blend the Two

Asking which one to pick is a trap. These two companies work in different scenarios simultaneously.

Palantir carries the alpha when AI budget flows into government, defense, and critical infrastructure. Alphabet becomes a compounding machine when AI budget flows into enterprise cloud. Those scenarios aren't mutually exclusive. They're both happening at the same time.

If you can stomach Palantir's volatility, the right construction is a small position + long holding period. Alphabet is comparatively more stable, so you can carry more weight. Yes, the ride is bumpy — but holding both means you're positioned regardless of which path AI software dollars take.

The most underpriced risk I see, restated: the market still hasn't fully escaped the "Alphabet = ad company" framing. The moment that framing cracks, a valuation re-rating follows. Getting positioned before that re-rating is the game at this stage of the cycle.