The Biggest AI Winners May Not Be AI Companies
Every conversation about the AI boom starts with the same question: which model wins? OpenAI, Anthropic, Google, Meta. It's the obvious question, and it's also the wrong lens if you're looking at where the money actually flows.
Every one of those companies has to spend billions with the same underlying industries, whether their model wins the race or not. Those industries get paid regardless of who's on top. Here are six of them.
Six Industries Every AI Company Has to Pay
1. Semiconductors
AI runs on chips, memory, and advanced manufacturing capacity that takes years to build. Every AI lab and every hyperscaler is competing for the same limited supply. Companies to know: NVIDIA, AMD, and TSMC.
2. Data centers
Training and running AI models takes enormous physical computing capacity, and that capacity has to live somewhere. Data center operators are leasing space and building new capacity as fast as they can. Companies to know: Equinix, Digital Realty, and CoreWeave.
3. Electricity
Data centers cannot expand without reliable power, and AI's electricity demand is growing faster than the grid can currently supply. Utilities and power generators with capacity to sell are becoming AI infrastructure plays in their own right. Companies to know: Constellation Energy, Vistra, and GE Vernova.
4. Cooling and electrical systems
More computing density creates more heat and more power demand, and both have to be managed at scale. The equipment that keeps data centers running, cooling systems, switchgear, backup power, is a bottleneck of its own. Companies to know: Vertiv, Eaton, and Trane Technologies.
5. Cybersecurity
More AI means more data, more identities, and more attack surface to protect. Every company adopting AI is also expanding what it has to defend. Companies to know: Palo Alto Networks, CrowdStrike, and Cloudflare.
6. Construction and skilled trades
Someone still has to pour the concrete, run the electrical, and physically connect the AI economy to the grid. Construction and specialty trades firms are in the middle of a multi-year buildout of data centers, power plants, and transmission infrastructure. Companies to know: Quanta Services, EMCOR, and Comfort Systems USA.
My Investment Thesis: Picks and Shovels
I'm less interested in guessing which AI model wins than in understanding which industries get paid across many possible winners. That's the picks-and-shovels approach: look for industries that collect revenue no matter which AI company comes out on top. If ten labs are racing to build the best model, the industries above get paid by all ten.
FAQ: Investing in the Physical AI Economy
Is this financial advice?
No. This is an educational framework for thinking about the AI economy, not a recommendation to buy any specific stock. The companies named above are examples of businesses operating in each category, not investment recommendations. Do your own research, or talk to a licensed financial advisor, before making any investment decision.
Why look at industries instead of individual AI companies?
Picking the single AI company that wins is a bet on one outcome out of many possible outcomes. Industries that supply chips, power, data centers, cooling, security, and labor to every AI company get paid regardless of which model or which lab ultimately wins the race.
The companies named above are examples of businesses operating in each category, not a recommendation to buy their shares. I am less interested in guessing which app wins than understanding which industries can be paid across many possible winners.
Jenny