Listen to the audio version of this article (generated by AI).
Tom Yeung here with today’s Smart Money.
Two gas stations – Hamilton Market and Exxon Mobil – sit on opposite corners of the same intersection in Spokane, Washington. Same fuel, same pumps, same dull branding that was decided on by committee. The only thing that ever changes is the number on the big sign out front.
Last December, that number turned into a weapon.
Hamilton Market dropped its price by a penny as a holiday gift to the neighborhood – and Mobil answered right back by matching it.
The dueling gas stations cut prices all day, all the way down to 59 cents per gallon. The high demand led to so many cars filling up their tanks that fuel trucks had to be called in to refuel the pumps. About 12 hours later, Hamilton Market station lost a few thousand dollars, while the Mobil station lost around $25,000.
When you sell the exact same thing as the guy across the street, you have no power to charge more – and profits get squeezed until there’s nothing left.
This is what I call a “bad business model.”
And today, a similar pattern is happening in artificial intelligence.
In this Smart Money, I’ll show you where AI profits are already getting squeezed, and which companies to stay away from.
Then, I’ll share how to spot companies that have something more valuable than a commodity.
Let’s jump in…
Where AI’s Commoditization Is Already Showing
Today, one of the clearest instances of a “bad business model” in action is GPU rental companies – also known as “neoclouds.”
These are like two gas stations on opposite corners.
For instance, CoreWeave Inc. (CRWV) and Nebius Group N.V. (NBIS) are both neocloud rental companies. And both companies buy the same fuel – Nvidia Corp. (NVDA) GPUs, electricity, and power systems – and produce the same AI computing power.
That makes it hard to stand out, and the financial results show the pressure. CoreWeave’s adjusted operating margin was just 5% in the most recent quarter, down from 16% in 2025, and Nebius’ was negative. The reason is simple: Switching is easy. Most customers don’t care if their AI workloads run on a CoreWeave server or a Nebius one. The decision usually comes down to price.
Now compare that with the Big Tech giants like Microsoft Corp. (MSFT) and Alphabet Inc. (GOOGL). These hyperscalers can charge far more for their services, of which they offer more than just computing power. There’s specialized software (Azure), built-in AI models (Gemini), and custom-designed chips to run certain models faster (Maia, TPU 8). So, they can charge far more for their services.
“Bad business models” in AI also exist in other (temporarily) red-hot areas:
1. General purpose AI models.
Many Chinese AI labs producing open-weight models like Z.ai and MiniMax (both traded in Hong Kong) are surprisingly interchangeable.
2. Routine services.
Companies like Veritone Inc. (VERI) and SoundHound AI Inc. (SOUN) produce AI audio software products that are replaceable by those from larger players.
3. Independent power producers.
Electricity has long been a commodity, and so high-cost electrical utilities like Clearway Energy Inc. (CWEN) and Capital Power Corp. (CPXWF) struggle to earn high profits even in good times.
When the air goes out of the AI trade, you will see these companies buckle first. We want to be on the other side of the equation…
The AI Winners That Survive the Squeeze
Companies with “good business models” don’t compete on price alone. They sell differentiated products that customers actively seek out, giving them the power to raise prices without destroying demand.
Upscale luxury hotels are great examples. The Oriental Hotel in Milan offers private tours of Leonardo da Vinci’s “The Last Supper,” where guests can view the artwork without any crowds. The Four Seasons of London does the same with the British Crown Jewels.
Eric’s latest addition to his Fry’s Investment Report portfolio offers a similarly powerful example.
It is a major global pharmaceutical company that uses AI to advance its lifesaving drug development program – an industry where a single dose often costs more than a night at a five-star hotel. Drugs can cost multiple billions of dollars to develop, and this firm is funding this expensive research with robust cash flows from its existing drug business.
In his September monthly issue, released last Friday, Eric notes that the company has overseen one of the fastest-growing product launches in its history, taking only 12 weeks to reach the first million prescriptions and just four weeks to add the most recent million.
To access all of Eric’s latest research on this company, learn how to join Fry’s Investment Report here.
Of course, there are even more great AI businesses in the Fry’s Investment Report portfolio – companies with pricing power that will survive the eventual squeezing of the AI industry.
Click here to discover more about these compelling business models today.
Until next time,
Thomas Yeung, CFA
Market Analyst, InvestorPlace
