The Picks and Shovels of the AI Revolution

How to Participate in the AI Trade Without Buying More Mega-Cap Growth

In the spring of 1848, a man named Samuel Brannan walked the streets of San Francisco waving a glass vial filled with gold dust.

Brannan was not a miner. He was a merchant, a newspaper publisher, and above all an opportunist. Before he ever shouted the words that would ignite the California Gold Rush, he had quietly done something far more strategic: he bought up every pickaxe, pan, and shovel he could find between San Francisco and the Sierra Nevada foothills. Then, and only then, did he take to the streets crying, "Gold! Gold! Gold from the American River!"?

By mid-June, three-quarters of San Francisco's male population had left for the mines. Brannan's store sold $150,000 worth of goods per month. He purchased tin pans for twenty cents apiece and resold them for fifteen dollars each.? Within two years, he was California's first millionaire.

The irony is now a proverb: during a gold rush, sell shovels. The prospectors who arrived with dreams of striking it rich mostly didn't. The people who supplied them—the toolmakers, the clothiers, the freight operators—built generational businesses. Levi Strauss didn't pan for gold. He sold riveted denim pants to the men who did. Come on, Brenda,* you know how that turned out.

We are living through another gold rush right now. The AI revolution is real, transformative, and accelerating. But the market has priced it as though it belongs to a handful of companies.

In reality, they have more options than they think. Beneath the software, there is an industrial scaffolding being built that is vast, essential, and in many cases reasonably valued.

Power: The Unsexy Foundation

Every revolution has an energy story. The Industrial Revolution ran on coal. The twentieth century ran on oil. The AI revolution runs on electricity—and the amounts are staggering.

According to the International Energy Agency, global electricity consumption from data centers is projected to more than double by 2030, reaching approximately 945 terawatt-hours annually—roughly equivalent to Japan's entire electricity consumption.? In the United States, data centers could consume between 9% and 17% of total electricity generation by 2030, up from approximately 4–5% today.4

The utility sector—long associated with retirees and dividend checks—is suddenly at the center of the most important technology story of the decade. Natural gas is the bridge fuel keeping data centers running while the grid catches up. Nuclear energy is experiencing a quiet renaissance as hyperscalers scramble for reliable, carbon-free baseload power. The AI trade is, first and foremost, a power trade.

Real Estate: The Physical Footprint of the Cloud

"The cloud" is a masterful piece of marketing. In reality, the cloud is a warehouse: a massive, climate-controlled, water-cooled warehouse sitting on real land with real construction costs.

The data center construction boom is one of the largest industrial build-outs in modern history.7 The average data center consumes approximately 300,000 gallons of water per day for cooling.?4 As AI workloads intensify, the thermal management problem grows exponentially. Data center REITS represent one of the most direct ways to own the physical layer of this revolution. The cloud lives in buildings. And buildings are an asset class.

Raw Materials: What AI Is Actually Made Of

Every digital revolution is, underneath, an analog one. AI is made of rocks and wire before it is made of code.

Copper is the wiring of the AI age. A Microsoft data center in Chicago used 2,177 tons of copper—roughly 27 tons for every megawatt of applied power.? S&P Global projects copper demand will swell to 42 million metric tons by 2040 (a 50% increase) while global production is poised to peak in 2030 at just 33 million metric tons—a "systemic risk for global industries."8

The Grid: Rebuilding the Backbone

The U.S. electrical grid was not designed for this. Goldman Sachs estimates approximately $720 billion in grid spending will be needed through 2030.6 But there's a bottleneck almost no one is discussing: transformers.

Wood Mackenzie projects a 30% supply shortfall for large power transformers in 2025, with lead times averaging 128 weeks—nearly two and a half years.?? Since 2019, demand for generator step-up units has surged 274%.?? Infrastructure is never exciting until it fails. The investors who understand the grid's role in AI are investing in the thing that makes everything else possible.

The Revolution Beneath the Revolution

The physical infrastructure required to support the AI economy is being built right now, at a scale not seen since post war electrification. And the companies doing that building are available at valuations reflecting the market's persistent blind spot: the assumption that the AI story is a software story.

It is not. It is an everything story. And the shovels have never been more valuable.