The picks-and-shovels version of the AI trade
The AI boom has a glamorous side, chips and models and training runs, and a far less glamorous side, power, cooling, steel, and copper. Adam O’Dell’s “Tech-Opoly” pitch is an argument that the boring side is the better place to invest. Rather than bet on which chipmaker wins, his framing says you buy the physical infrastructure every data center has to purchase, no matter whose silicon sits inside. It is the classic picks-and-shovels logic, updated for the electricity age.
Why infrastructure beats a chip lottery
Infrastructure has one property that the chip trade does not: it is demand-agnostic at the product level. A switchgear order gets placed whether the customer buys from Nvidia or a rival, and a transformer gets ordered whether the data center trains models or simply serves them. That makes the electrical layer more durable than a bet on a single chip architecture, because it earns regardless of who wins the model wars.
The tradeoff is equally real. Infrastructure names grow more slowly and carry thinner margins than a breakout chip stock, and their upside is bounded by how fast utilities and developers can actually build. Our data center power stocks explainer unpacks where the growth is concentrated and where it stalls.
The three layers of the buildout
The infrastructure story splits into three layers, and O’Dell’s basket reaches into all three. The first is the electrical equipment itself, the switchgear, transformers, and breakers that carry and control power. The second is distribution, the high-volume middlemen who get the gear from factory to job site. The third is the engineering and design work that plans where all of it goes. We walk through the equipment side in our electrical equipment stocks explainer.
None of these layers is a new industry. What is new is the scale of simultaneous demand, as every hyperscaler, every utility, and every developer tries to order the same scarce gear at the same time.
The “opoly” claim, examined
The word “monopoly” does a lot of work in the pitch, and it does too much. There is a real chokepoint here: the electrical layer is genuinely scarce, and backlogs stretch for years. But scarcity is not the same as monopoly. Distribution is competitive and low-margin, engineering is a crowded field, and even the equipment makers compete with giants like Eaton and Siemens. The honest version of the thesis is not “monopoly” but “temporarily scarce supply meeting surging demand,” which is still a good setup, just a less dramatic one.
What this means for the average investor
For most investors, the infrastructure basket is a steadier way to participate in AI than chasing the next chip name, but it is not a free lunch. The theme has already run, several of the names in O’Dell’s list are up sharply since they were teased, and a good thesis can be a bad entry price. The durable takeaway is that the electrical layer is a real multi-year bottleneck, and it will not clear soon because the grid cannot be expanded faster than the equipment makers can build the gear. The buyer’s job is to separate the genuine chokepoint owners from the companies that merely touch electricity. Our data center power demand explainer lays out the demand side in detail.
A framework for sorting the basket
If you want to evaluate any data center infrastructure stock, three questions do most of the work. The first is ownership: does the company make a scarce, certified product, or does it resell someone else’s? Makers own the bottleneck and capture pricing power; resellers earn a thin margin on volume. The second is concentration: how much of the company’s earnings actually comes from the AI buildout? A diversified giant gets a modest lift from data centers, while a focused specialist gets a large share of its earnings from exactly the projects being ordered today. The third is price: how much of the future benefit has the stock already priced in? A great thesis bought after a sharp run can still be a poor investment.
Applying that framework to O’Dell’s basket explains the spread of returns almost perfectly. The equipment makers, which own scarce gear, ran the furthest. The distributors, which resell on thin margins, ran less. The engineering and software names, which touch the theme indirectly, ran the least, and one actually fell. The framework does not predict the future, but it separates a genuine chokepoint from a company that merely stands near one, and that separation is worth more than any single stock pick in the promo.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.