Google's $462 billion cloud backlog is a number worth sitting with

Google Just Quietly Became the Most Dangerous Company in Tech

Let me tell you what happened last week that most people completely missed.

While the financial press was busy writing the same AWS-vs-Azure horse race story they've been writing since 2019, Alphabet's CFO Anat Ashkenazi sat down at a microphone and said something that should have stopped every portfolio manager in America cold.

She said 2027 CapEx will 'significantly increase' above an already-raised $180-190 billion range, and that those bets are based on 'tangible demand signals.'

Read that again. Not 'we believe.' Not 'we anticipate.' Tangible. Demand. Signals.

Wait. Strike that. The quote stands on its own. A CFO with a fiduciary duty to shareholders is telling you, in the driest possible language, that the returns are real and they are accelerating into them. That's not a vision statement. That's a capital allocation decision backed by signed contracts.

Here's the number that should haunt every AWS and Azure bull: Google Cloud's backlog nearly doubled in a single quarter to $462 billion. Not year-over-year. Quarter-over-quarter. And AI solutions revenue inside that cloud business grew 800% year-over-year.

Eight hundred percent.

This is where the market is still asleep at the wheel.

Google announced it is selling TPUs externally to enterprise customers. This is the sleeper story of the entire earnings season. For years, Google's custom AI chips were an internal advantage, a moat you couldn't buy. Now they're selling access to that moat. Which means Google Cloud isn't just competing with AWS and Azure on compute. It's competing with Nvidia.

Think about what that $462 billion backlog actually represents now. It was already large. But if enterprise customers are baking Google's TPU architecture into their AI builds, the switching cost just went from 'annoying' to 'existential.' You don't rip out your chip architecture on a Tuesday morning.

Sundar Pichai has been a punching bag for years. 'Google can't execute.' 'They'll lose search to AI.' 'They're going to miss the cloud wave.' I've heard it all. I've probably written some version of it myself.

But look at what's actually happened. Search grew 19%, not despite AI, because of it. Cloud hit $20 billion in quarterly revenue. And the company is now building a full-stack AI offering (models, chips, cloud infrastructure, developer tools) that nobody else can fully replicate today. Not Amazon. Not Microsoft. Not anyone.

Microsoft is extraordinary. AWS is re-accelerating. But Google? Google is running a different race entirely.

The stock trades around 21x forward earnings as of this morning. For a company with a $462 billion cloud backlog, an AI hardware business the market hasn't priced in, and a CFO who just told you returns are real and they're doubling down?

That's not a valuation. That's an opportunity.

✦

Quick Takes

Three more stories worth knowing about

Amazon's $50B Chip Secret Is the Most Underpriced Story in Tech

Andy Jassy mentioned almost in passing that Trainium (Amazon's custom AI chip) could be valued at $50 billion as a standalone business. Almost nobody reported on it. AWS is now, by Amazon's own assessment, a top-3 data center chip company. The market is pricing Amazon as a retailer-with-a-cloud. It's closer to a chip company, a satellite company, a logistics empire, and a media conglomerate that also happens to run the world's biggest cloud. AWS grew 28% on a $150 billion run rate, the fastest pace in 15 quarters. Re-acceleration on that scale doesn't happen by accident.

Microsoft's OpenAI Deal Just Became a Long-Term Cost Advantage

Everyone focused on the drama of Microsoft 'renegotiating' its OpenAI partnership. Here's what actually happened: Microsoft locked in royalty-free model IP through 2032 while simultaneously turning OpenAI into a paying Azure customer. They converted their most expensive partnership into a durable cost advantage. Meanwhile, Azure grew 40%, Microsoft 365 Copilot hit 20 million paid seats growing 250%, and AI annual recurring revenue hit $37 billion growing 123% year-over-year. Amy Hood also flagged the shift from per-seat to per-seat-plus-consumption pricing, the move that turns Microsoft's 600-million-user installed base into an open-ended revenue engine.

Apple Dropped a New Product Category and Nobody Noticed

Apple posted its best March quarter ever: $111.2 billion in revenue, up 17%, EPS up 22%, double-digit growth in every single geographic segment simultaneously. That last part is genuinely rare and almost never happens. But the detail I can't stop thinking about: Tim Cook casually mentioned something called 'MacBook Neo' as captivating customers worldwide. That name has never appeared in any Apple filing or product line before. Apple's PR team does not use words accidentally. A new product category is being assembled in plain sight, and the $100 billion buyback tells you management thinks the stock is cheap relative to what they know is coming.

✦

Overlooked Stock

SKYT

SkyWater Technology, Inc.

Everyone is fighting over TSMC and Samsung. Meanwhile, SkyWater Technology is steadily becoming the most strategically important semiconductor company that most retail investors have never heard of.

Here's the setup: SkyWater is the only pure-play, ITAR-compliant semiconductor foundry operating entirely on U.S. soil. Their Bloomington, Minnesota fab is cleared for defense and aerospace work that cannot legally go to Taiwan or South Korea. With the U.S. government spending tens of billions through the CHIPS Act to reshore semiconductor manufacturing, SkyWater isn't a beneficiary. It is the infrastructure.

They serve aerospace and defense, advanced computation, bio-health, and automotive sectors. Their specialty is analog, mixed-signal, MEMS, and radiation-hardened chips, exactly the categories the Pentagon cannot afford to offshore.

This isn't a story about competing with Nvidia. It's a story about the U.S. government needing a trusted domestic foundry and there being essentially one place to call. That's a moat built by geography and security clearance, not marketing. The stock is small, illiquid, and ignored. That's usually where the real money is made.

The Contrarian Take

The AI Infrastructure Boom Isn't Slowing Down — It's Just Getting Started, and the Bears Are Wrong

Every few weeks someone publishes a piece arguing that hyperscaler CapEx is unsustainable, that AI ROI is imaginary, that we're in a bubble. I've read them all. They're wrong, and here's why.

Look at what just happened in earnings season. Microsoft's Azure grew 40%. AWS re-accelerated to 28% on a $150 billion run rate. Google Cloud's backlog doubled in a quarter to $462 billion. These aren't projections. These are signed contracts, consumed workloads, and real cash changing hands.

Anat Ashkenazi at Google doesn't raise CapEx guidance to $180-190 billion and then signal 2027 will be 'significantly higher' because she's optimistic. She does it because enterprise customers have told her, in writing, what they're going to spend.

The bears are confusing 'we don't understand the use cases yet' with 'the use cases don't exist.' They existed in 1997 too, you just couldn't see them from where you were standing. The infrastructure being built today is the 1990s internet backbone. The applications haven't been written yet. That's not a warning. That's the opportunity.

Next briefing

Middle East Mayhem + Warby Parker's 32% confidence score: The canary or the false alarm?

Previous briefing

Google just told you the ROI is real. Listen.