Last week, Google (GOOGL) briefly overtook Nvidia (NVDA) as the world’s most valuable company.

Nvidia reclaimed the throne within hours, but the moment captured something important about the artificial intelligence (AI) boom.

One of the world’s biggest AI spenders (Google) and the world’s biggest AI winner (Nvidia) are now fighting for the same crown.

In the three and a half years since ChatGPT launched, my advice has always been to own the companies cashing the AI checks.

That’s how we made multi-bagger profits in Nvidia, Micron Technology (MU), and several other AI plays in Disruption Investor.

And I’m confident this strategy will keep working. Spending on AI data centers will blow past $750 billion this year—more than double 2025 levels.

But...

• Not all AI spending is created equal.

Most AI dollars are still spent on the chips needed to train and run ChatGPT. But it’s no longer the fastest-growing slice.

As AI models grew bigger and more capable, they needed massive amounts of high-bandwidth memory to move data faster. Without it, GPUs would sit idle waiting for work. Nvidia recently revealed that memory alone accounts for roughly 60% of the manufacturing cost of its Blackwell AI chips.

Memory used to be a small slice. Now it’s right up there with GPUs. We were early to this shift and have more than tripled our money on Micron.

We’re hunting for the next opportunities in the slices that are still relatively small but growing rapidly.

• “The most undervalued AI stocks…”

That’s what I called optical stocks back in December.

Applied Optoelectronics (AAOI) has surged 700% since January. Lumentum Holdings (LITE) jumped 300%. And Ciena Corp. (CIEN) has nearly tripled.

Inside massive data centers, hundreds of thousands of AI chips are constantly “talking” to each other.

More than 90% of an AI model’s response time can be spent moving data between chips. If the pipes between GPUs are slow, the chips sit idle. A $40,000 GPU doing nothing is the most expensive paperweight in tech.

These chips used to be able to communicate with each other through copper wire. But copper is inefficient at these workloads. And it generates excessive heat over distance.

That’s why AI builders like Google, Microsoft (MSFT), and Amazon (AMZN) are rapidly shifting toward optical networking, which transmits data as pulses of light through fiber.

Optics are roughly 3X more efficient than copper over short distances and as much as 20X more efficient over long distances.

If you want tens of thousands of GPUs to behave like one giant brain, optics is the only way. The growth here is massive for three reasons:

First, AI data centers are getting dramatically larger every year.

Second, data speeds are doubling roughly every 18 months. Each upgrade is a full rip-and-replace cycle of all optical gear and a new revenue wave for the optics companies.

Third, Nvidia is pushing the industry toward “co-packaged optics,” where optical components sit millimeters from the GPU instead of inside separate networking systems.

Combined, these changes are opening a whole new optics market that didn't exist 12 months ago.

• Ever hear of power semis?

They’re a type of computer chip that regulates and converts electricity flowing through AI systems. And they’re fast becoming one of the most important parts of the AI buildout.

Traditional server racks consumed roughly 5–10 kilowatts of power.

Nvidia’s latest Blackwell systems consume 10X to 20X more energy!

Before electricity gets to expensive AI chips, it must be carefully regulated every step of the way. Too much power, too little power, or unstable voltage can damage hardware and reduce performance.

That requires a lot more power semi equipment throughout the system.

So demand is rising from two directions at once: AI data centers getting bigger, with more racks packed into each facility, while each individual rack is consuming far more electricity.

• My #1 slam dunk AI trade over the next 12 months is…

The companies who make the machines that produce chips.

Think of these as the ovens that bake the world’s latest and greatest chips.

The largest chipmaker in the world—Taiwan Semiconductor (TSM), or “TSMC,” for short—just announced it will spend a record amount of money expanding its factories and buying new equipment this year.

This is TSMC admitting it can’t make chips fast enough.

The other two chipmaking giants—Samsung and Intel Corp. (INTC)—are building new factories, too.

• Chip factories are insanely expensive.

A modern, leading-edge fab now costs close to $30 billion to build from scratch. Besides nuclear power plants, they’re the most expensive structures on Earth.

Between 70% and 80% of the total cost of a chip factory goes into the machines inside it. The tools that actually print, shape, and inspect the chips.

These machines costs so much because chipmaking requires extreme precision.

Chips are absurdly tiny. A 2-nanometer chip is about 40,000X to 50,000X thinner than a human hair.

And as chips get smaller, the number of manufacturing steps roughly doubles. More steps mean more tools are required to produce the same number of chips.

A modern 2-nanometer chip can pass through more than 1,000 individual process steps before it’s complete. Each step must be nearly perfect. A single microscopic defect can ruin the entire chip.

So when global demand for AI chips explodes, it doesn’t just require more factories. It requires more of the equipment inside those factories.

• Keep buying the AI winners.

Big spenders like Google will likely do very well.

But I prefer to own the companies who continue to cash the AI checks.

The important thing to watch for is where the hot ball of AI money is moving next.

My favorite opportunity for the rest of 2026 and beyond is the chipmaking machine stocks.

Invest accordingly.

Stephen McBride

Chief Analyst, RiskHedge

PS: One of the easiest ways to spot where investor money is moving next is through price action. And that’s exactly what my colleague Justin Spittler will be breaking down during his ChartMania Live! event on Monday, May 18, at 4 PM ET.

You can submit any stock you want—AI leaders, speculative moonshots, or whatever else is on your radar—and Justin will chart it live and tell you if he’d buy it and how he’d trade it.

If you’d like to participate, you can submit your stock here.