SHOTS FIRED

(GOOG), (META), (AAPL), (MSFT)

In the 1970s, I was crawling through back alleys in Bangkok chasing stories about guerrilla fighters. 

Forty years later, I’m crawling through earnings transcripts chasing stories about AI. Some things change. But the skill set, spotting what others miss, never goes out of style. 

That’s how I ended up looking deeper into Alphabet (GOOG). 

Everyone’s busy asking if Gemini can beat ChatGPT—wrong question. The real story isn’t the chatbot. It’s that Google now controls the full stack: custom chips, enterprise adoption, and Cloud margins that are climbing faster than anyone expected. 

They’ve quietly built a vertically integrated AI empire while the rest of Silicon Valley was busy issuing press releases. And to top it all off, it’s already profitable.

The Google Cloud business has gone from punchline to powerhouse. What was once a money-losing side gig is now minting cash at a scale even the bulls underestimated. 

Just look at the numbers from their Q3 2025 report: 65% stock growth over the past year, a $73.6 billion free cash flow engine, and more than $1 billion in new cloud contracts signed in just the first nine months of the year. 

But the real gem isn’t confined to the numbers. It’s the foundation: a vertically integrated stack with friction so high, customers don’t just sign deals, they dig in for years. 

Enterprise workloads are now built on Google’s frameworks, custom AI agents are handling mission-critical ops like marketing automation and search optimization, and 70% of customers are using Google AI tools on a recurring basis. If that’s not a moat, then I don’t know what is.

Part of that moat is paved in silicon. Ironwood, Google’s seventh-generation TPU, is the kind of chip that’s destined to move markets. It’s 10 times faster, four times more efficient, and links up 9,216 TPUs per super pod – the largest AI compute clusters commercially available. 

Ironwood powers everything from Bard to Gemini to custom enterprise applications, and it’s been winning over clients that don’t typically place bets lightly. Meta (META) just signed on for $10 billion worth of access. Anthropic is loading up a million TPUs. Even Apple (AAPL), notoriously tight-lipped and allergic to outside tech, is training foundation models on Google’s clusters. 

Of course, it helps when your in-house model isn’t just “competitive” but actually embedded. Gemini now boasts 346 million monthly active users and isn’t just gunning for ChatGPT; it’s showing up everywhere from Pixel phones to Google Workspace to Search itself. 

It’s also enabling things that are frankly more monetizable than casual chat: call center automation, predictive marketing, and internal workflow orchestration. 

The brilliance of Gemini isn’t that it’s smarter – it’s that it’s everywhere, and it keeps Alphabet’s AI revenue from being a single-product story. The more Gemini shows up in workflows, the more Cloud usage compounds. 

Yes, the stock has run. It’s trading at nearly 30x forward earnings, and the market is no longer discounting Alphabet’s AI potential. But it’s not priced for perfection, either. 

Forward revenue is still projected to grow 14% annually, earnings climbing alongside as Cloud margins expand. And Alphabet’s CapEx discipline puts it ahead of the pack: spending 49% of operating cash flow versus Microsoft’s (MSFT) 65% and Meta’s 77.5%. 

This isn’t reckless reinvestment. This is a company turning scale into operating leverage, and doing it while most investors are still arguing about which LLM is wittier.

There are risks, of course. Scaling AI infrastructure isn’t just a chip problem anymore. It’s now a power problem, a data center real estate problem, and a talent allocation problem. 

Multi-cloud hedging among large clients is real, and Alphabet will need to keep earning its seat at the table. But as long as it keeps pushing the stack forward – chips, models, integration – it won’t just keep its lead. It might even grow it.

So for now, I’m holding my position and letting the story play out. I’m betting that in five years, we’ll look back and realize that Alphabet didn’t win the AI race by running fastest. It won by quietly building the road the others are forced to run on. 

And while I traded guerrilla warfare for balance sheets, I haven’t forgotten the lessons from the battlefield. Control the terrain, and the rest tends to follow.