When Gutenberg invented the printing press in 1455, he basically went broke. His financier? That guy made a fortune. Same story plays out every time: someone builds the revolutionary thing, someone else owns the supply chain and prints money.
Last week, Microsoft, Google, Amazon, and Meta just confirmed which side of that equation is about to get very, very wealthy.
The $700 Billion Spending Spree Is Real
These four tech giants collectively committed over $700 billion in 2026 capital spending to build out AI infrastructure—and they’re already signaling that number’s going up in 2027. This isn’t hype. This is the largest capital investment cycle in tech history, backed by real revenue, real margins, and real customer commitments.
Microsoft’s AI business hit $37 billion in annualized revenue, growing 123% year-over-year. Azure is growing at 40%—its fastest pace in years. Microsoft’s spending $190 billion on infrastructure in 2026 alone, with $627 billion in already-contracted revenue waiting to be recognized. Translation: they’re not slowing down.
Google Cloud just crossed $20 billion in revenue, growing 63% year-over-year. Their backlog nearly doubled in one quarter to $462 billion. Half a trillion dollars in contracted revenue sitting there. And here’s the kicker: the bear thesis that AI would cannibalize Google’s search business? Dead. Search queries are at all-time highs, and AI-powered results are driving more commercial searches. Search revenue grew 19%.
Amazon’s AWS grew 28% year-over-year on a $150 billion revenue base—which shouldn’t be possible at that scale, yet it’s happening. But the real story is Amazon’s custom chip business. Their Trainium AI chip offers 30-40% better price performance than Nvidia GPUs. Amazon’s already got $225 billion in Trainium revenue commitments, with Trainium3 nearly fully subscribed and Trainium4 already mostly reserved.
Meta’s spending $125-145 billion on CapEx in 2026, up from their previous guidance. They admitted they’ve “continued to underestimate” their compute needs—a remarkable confession from a company that’s been aggressively ramping capacity for two years straight.
Follow the Money
When these companies spend $700 billion-plus, that money flows through the entire AI hardware stack: GPUs, custom chips, networking equipment, semiconductor manufacturing gear, power systems, thermal management. Every company on the receiving end of that spending wins.
The obvious plays: Nvidia (still the primary GPU supplier), Broadcom (custom chips for Google and Meta), Marvell (custom chips for Amazon and Microsoft), AMD (gaining share in AI training), and Micron (benefiting from memory price surges).
But don’t sleep on the infrastructure layer. Bloom Energy, Eaton, and Vertiv are positioned at the center of the biggest bottleneck: power. U.S. data center energy demand is projected to nearly double from 80 to 150 gigawatts by 2028—equivalent to adding another Spain to the American grid in three years.
Then there’s the networking layer: Credo Technology (copper interconnects), Lumentum, and Coherent (optical interconnects for the long term).
The Bottom Line
This isn’t a momentum trade built on narrative. It’s a fundamental trade anchored in the largest capital investment cycle in tech history. As long as hyperscalers keep spending—and they just emphatically confirmed they will—the economy expands, the market rises, and the AI infrastructure names keep working.
The picks-and-shovels trade is visible. But what matters just as much is what gets built on top of it. That’s where the next phase of AI fortunes get made.
