Editor’s note: Today, we’re giving you a special look inside the August issue of Disruption Investor. It tackles one of the biggest questions in markets right now: Is the artificial intelligence (AI) buildout getting ahead of itself?

We’re also running a special Labor Day sale on Disruption Investor. Join today and you’ll get our Q4 Disruption Playbook, with Stephen and Chris Wood’s top stocks for the rest of 2026.

Here’s Stephen with more…

***

“We taught sand to think.”

Venture capitalist Marc Andreessen recently said this on “The Joe Rogan Experience.” No other single quote captures the scale of what we’re actually doing when we talk about the AI buildout.

We take silicon (sand) and melt it, purify it, grow it into crystals, and slice it into wafers. We etch patterns into those wafers so tiny they’re measured in atoms.

Then we repeat the steps 1,000X over.

Run power through at exactly the right voltages, and the sand can reason through problems… write software… and make connections in research that no human has ever made.

Computers have been able to process calculations for decades. But before AI, computers didn’t actually “think.”

A spreadsheet can’t reason through problems. A Word document just holds your thoughts. You still have to write them. Pre-ChatGPT, computers were just tools. We still had to do the work.

The latest AI models can work for more than 12 hours. And that duration is doubling every few months.

Set them on a task and they’ll plan out the work, break tasks into separate subtasks, spot their own errors, and start over on their own. No human needed.

All that from sand and electricity.

From the beach to the data center

To see AI’s “sand story” in action, look no further than our own portfolio…

“Semi-cap” stocks build the machines that shape the sand. Making a single AI chip takes hundreds of individual steps, and if a single step goes wrong, the chip is useless.

Each of these steps needs machines that work at atom-level precision. And the more steps in the process, the more equipment you need.

We built our semi-cap “ETF” to capture four separate layers of this process.

Chipmakers like Taiwan Semiconductor (TSM) are upgrading their factories to make the most complex AI chips. Buy just one semi-cap stock, and we’d miss out on the full wave of spending on all the equipment these chipmakers need.

TSM is the one actually shaping sand. About 70% of the chips that end up in AI data centers started as wafers in TSM’s foundries.

Those wafers leave Taiwan and become Nvidia (NVDA) GPUs. Nvidia is TSM’s largest customer, and it’s the company that turns shaped sand into the chips that actually “think.”

Nvidia’s chips run essentially every frontier AI model on the planet. But the processor can only think as fast as the memory chips that supply the data. Micron Technology’s (MU) high-bandwidth memory is designed to do exactly that.

We also own the companies that wire that thinking together. [Redacted] keeps the signals clean as they travel between chips. [Redacted] keeps the whole thing powered and cool enough to stay online.

Every stock in our Disruption Investor portfolio is a link in the same chain.

On the scale of human history, there are only a handful of moments like this one. AI is a civilization upgrade on the same level as fire, the wheel, and electricity.

And it’s spreading faster than any technology in human history.

Electricity took about 40 years to reach most households. The internet buildout took only 12 years to reach most households.

AI crossed the 50% threshold in just three years. Half of US businesses now pay for AI, up from 5% in early 2023.

And now everyone wants to know, “Is AI a bubble?”... 

Every bubble in history died the same way…

Oversupply.

Railroads laid too much track before there was enough cargo demand to run on those rails. Telecoms wrapped the world in cables before the world was online to use it.

Buildouts collapse when supply races past demand. With AI, we see the exact opposite. Demand is still far beyond what AI companies can supply.

For the data center operators, it’s impossible to overbuild right now. Whether it’s a “free” model like Kimi K3 or the best Anthropic has to offer, every AI model needs two things that the world has already run out of: wafers and watts.

AI companies can’t add chips when TSM is rationing wafers. Its advanced packaging is sold out through 2026, and its backend plants are reportedly booked into 2027. Memory chips are also sold out through the end of next year.

And you can’t just add more power when the queue to connect to the grid is eight years long.

Natural gas turbines, which power most new data centers, are sold out into the 2030s. Transformer lead times run for more than 160 weeks.

If AI compute were overbuilt, hyperscalers’ cloud revenue would be collapsing. Instead, cloud revenue and profits are accelerating.

To keep up with AI demand, there’s only one lever left to pull: If you can’t build more, get more out of what you have.

Every shortage we’ve tracked so far has been a shortage of silicon: Nvidia’s GPUs, TSM’s wafers, Micron’s memory.

For the first time in the AI buildout, the answer isn’t more silicon. Silicon does almost every job in the AI stack. But it’s a terrible light emitter.

Just as we taught sand to think, we need to make light carry thoughts.

Light has always been a faster, more efficient way to send data than electricity. It’s why we buried thousands of miles of fiber-optic cables across the ocean floor.

This month’s recommendation is the company that’s manipulating light to do what silicon can’t.

Disruption Investor members can catch up on the rest of the issue here.

If you’re not a member and would like to keep reading, upgrade to Disruption Investor here. In it, you’ll learn about the company best-positioned to profit from the shift from silicon to light… and how its lasers are making this transition possible.

If you’ve been thinking about joining, this is the best time to do so. We just updated our Disruption Playbook with our highest-conviction stocks to close out 2026.

As part of our Labor Day sale, you’ll get the new playbook the minute you join. Details here.

Stephen McBride

Chief Analyst, RiskHedge