Tech Giants Are Spending $725 Billion on AI This Year. Here's Who Gets Paid.

Here's a number that should make every investor pay attention: $725 billion. That's how much Alphabet, Amazon, Microsoft, and Meta plan to spend on AI infrastructure in 2026 alone. It's 77% more than the $410 billion they spent last year.

And this week, the number got even bigger. Microsoft told investors it expects to spend $190 billion this calendar year, about $38 billion more than Wall Street expected. Meta bumped its range to $125 to $145 billion, citing higher memory costs and the scramble to lock up land and power. Google guided $180 to $190 billion and said 2027 spending would increase "significantly."

Bank of America and Evercore now project total hyperscaler capex will blow past $1 trillion in 2027. Goldman Sachs puts the three-year total (2025 to 2027) at $1.15 trillion. These aren't projections from startups hoping for growth. These are the biggest companies on Earth writing actual checks.

Every dollar of this $725 billion flows through a supply chain most folks never think about. And understanding that chain is where the real investment insight lives.

Layer 1, Chips: It all starts with a chip order. NVIDIA (NVDA) is still the clear leader here. Revenue hit $216 billion over the trailing twelve months, with $120 billion in profit. At $198.45 a share and a $4.82 trillion market cap, it's the most valuable company on Earth. The stock is up about 8% this year and 72% over the past 12 months. Earnings come May 20.

Broadcom (AVGO) took a different path, building custom AI chips designed specifically for Google, Meta, and now Anthropic and OpenAI. The stock closed at $421.28, up roughly 22% YTD and 128% over the past year, with a market cap pushing $2 trillion.

AMD is the No. 2 GPU challenger. CEO Lisa Su locked in multi-year deals with Meta and OpenAI, and the data center segment is growing 60%+ annually. The stock closed at $360.54, up 65% YTD and roughly 270% over 52 weeks. Earnings are May 5.

Worth noting: Global chip sales are on track to hit $1 trillion in 2026 for the first time ever, after a 25.6% jump in 2025. The hyperscalers are the single biggest reason why.

Layer 2, Data Center Infrastructure: Chips don't run themselves. They need cooling, power management, and physical buildings. Vertiv (VRT) makes the thermal systems that keep AI racks from overheating. Q1 2026 sales came in at $2.65 billion, up 30% year over year. The stock closed at $328.31, up 86% YTD and 258% over the past year. Eaton (ETN) supplies the transformers and switchgear, sitting near all-time highs at $410.77. And Quanta Services (PWR) builds the actual networks and power lines, at $631 after strong Q1 earnings.

Layer 3, Energy: All this compute needs power. Not intermittent power, but reliable 24/7 baseload energy. That means nuclear, natural gas, and uranium.

Constellation Energy (CEG) owns the largest nuclear fleet in the U.S. and signed a landmark deal with Microsoft to restart Three Mile Island Unit 1. The stock sits at $308, down 12.87% YTD amid regulatory headwinds, but still up 28% over the past year. Vistra (VST) combines nuclear and gas generation at $155.80, down modestly this year but up 13% over 12 months. And Cameco (CCJ), the world's largest publicly traded uranium producer, has been the standout energy play at $120.60, up 32% YTD and 165% over the past year as long-term uranium contracts lock in.

There's a natural question here: is all this spending sustainable? The honest answer is, it depends on whether you think AI workloads are real or hype.

The earnings calls this week gave a pretty clear answer. Google Cloud revenue beat expectations, growing faster than both Azure and AWS. Meta said its AI-powered ad targeting drove a 33% revenue jump to $56.3 billion in Q1. Microsoft said it'll stay "capacity constrained" through 2026 even after spending $190 billion. These companies aren't spending because they feel like it. They're spending because their customers are demanding more compute than they can physically build.

But here's where it gets more nuanced. Not every stock in this trade has moved the same way. The chip makers (NVIDIA, AMD, Broadcom) have seen the most dramatic gains. The infrastructure plays (Vertiv, Quanta) are close behind. The energy stocks are a mixed bag, with Cameco up 165% on uranium demand, but Constellation and Vistra still working through regulatory uncertainty and power pricing questions.

The clearest read on whether NVIDIA's AI GPU dominance is starting to crack. Watch data center revenue and the Helios GPU update closely.

The single most important event for the AI trade this quarter. Forward guidance will set the tone for every stock on this list through Q3.

Constellation reports amid nuclear hype and its Calpine acquisition. Will management confirm the hyperscaler power deal pipeline?

The AI spending boom isn't a prediction anymore, it's a confirmed, funded, multi-year industrial buildout.

$725 billion in 2026. A trillion-plus in 2027.

The money flows through a clear supply chain: chips first, then data center infrastructure, then energy. You don't need to pick the next AI app winner.

You just need to follow the money.

And right now, it's going to the companies that build, power, and cool the machines.